Category: Matrix

  • Custom Keyboards Worth Watching: Weird Themes, Retro Innovation, and Electrocapacitive Builds

    Custom Keyboards Worth Watching: Weird Themes, Retro Innovation, and Electrocapacitive Builds

    Keykobo Emoji

    At the end of March, KEI published the GB post for Emoji, a keycap set inspired by chat emojis. “EMOJI elements have become part of everyone’s daily conversations now, and I found them especially charming, so I wanted to incorporate them into a keycap design.”

    When it comes to designs inspired by “emoji,” they are actually not uncommon in the world of keycaps. Most of the time, though, they appear only as decorative elements — novelty keys, or a small number of special alphas. Examples include Alpaca’s UwU Alphas or novelty concepts like 404. But building an entire keycap set around emojis as the core theme is still relatively rare.

    In terms of color scheme, the Alpha section uses dark gray as its base, paired with lighter gray legends, while the sublegends introduce highly saturated yellow accents. The Mod section uses gray as the base with black legends. Some extension keys and novelty accents use yellow as the base color with black legends. Normally, when people think of emojis, large areas of yellow are the first thing that come to mind. However, KEI’s design instead uses grayscale as the foundation, with yellow-green accents layered on top. This approach also gives the set a much broader compatibility range. The combination of restrained cold-toned bases with bright, high-contrast symbols feels somewhat similar to colorways like GMK Nerve — using subdued foundations to carry more visually aggressive saturated elements, making the focal points stand out while avoiding an overly chaotic overall appearance.

    If the Base kit alone doesn’t immediately communicate a strong connection to emojis, then the novelty and extension kits fully unleash the theme.

    Visually, the set adopts a rounded “face-like” design language, somewhere between early pixel-style symbols and modern emoji expressions, while faintly resembling Pac-Man-esque shapes. At the same time, the design also incorporates quite a few memes and references rooted in the Chinese internet culture.

    The PAD section takes an even more direct approach, using variations of the “👀” symbol from different angles to replace traditional numpad legends. While preserving functional zoning, it also reinforces the playful personality and recognizability of the entire set. In addition, translucent novelty keys appear as accent elements — something that has basically become standard practice in modern novelty design. Still, the absence of classics like the “sweating yellow bean” or the “Huaji” meme feels a little disappointing to me.

    From a pairing perspective, the core of Emoji lies in its grayscale foundation combined with bright accent highlights, which means there are fairly clear boundaries in terms of compatibility.

    It works best with keyboards that lean toward colder tones and cleaner structural designs, such as dark gray, black, or silver neutral-colored cases. These kinds of base colors naturally complement the grayscale system used in the Alpha and Mod sections, helping the overall appearance remain cohesive while allowing the yellow accents to become the visual focal point.

    For keyboard kits whose design language is relatively restrained — for example, cases without excessive chamfers or aggressive decorative elements — this keycap set can effectively brighten the overall look and add personality without breaking the existing aesthetic. But if paired with keyboards that are already highly colorful or carry a strong thematic color scheme of their own, I personally don’t think the result would work particularly well.

    GMK Gregory 2

    At the end of April, pancake published the GB post for Gregory 2, a parody-themed keycap set inspired by the classic doge reaction meme. While most keycap designs try to establish some sort of theme or worldbuilding, this set feels more like a form of community performance art. Its inspiration does not come from films, art movements, industrial design, or cultural symbols, but from a “doge meme” printed on a Taobao desk mat. Most people make desk mats to match their keycaps — they fell in love with the desk mat first, then decided to turn that face into a keycap set.

    And the set genuinely carries a uniquely internet-meme kind of energy: starting from a place of absurdity, stupidity, and even a slightly nonsensical emotional tone, then retroactively rationalizing that feeling into something cohesive. Precisely because of this, it gained extremely strong traction on Geekhack and Reddit, where many users joined in and started riffing on the joke themselves.

    The R2 Base kit continues the iconic “ugly-cute” aesthetic of the original generation: large areas of black-and-yellow coloring, exaggerated facial expressions, graphics that almost resemble low-resolution texture maps, and, of course, the instantly recognizable Gregory face.

    That said, despite being a meme set, the color palette itself is actually fairly traditional. Whether in industrial warning systems, engineering equipment, or street signage, black and yellow naturally create strong visual impact and high recognizability — similar to classic colorways like Serika.

    As for novelties, there are dog paws, red accent keys intended to replace the pupils, and even red-and-blue Matrix-style pills. There’s also a flesh-pink add-on kit inspired by the rosy-cheeked “dimple face” expression. The entire set may be absurd, but it never feels cheap. Of course, because of the theme itself, this set works especially well with TKL layouts, where the full Gregory face can be displayed more completely.

    For pairings, smaller layouts work well with silver narrow-bezel cases, while larger layouts can experiment with yellow or black boards. WKL and TKL are recommended; HHKB is not. In reality, Gregory isn’t the kind of traditional themed keycap set where people can still appreciate the aesthetics through color or artwork even without understanding the background. It also inspired later meme-driven sets such as Silly Goose-style designs. Opinions on the set are quite polarized, though — some people consider it one of the most community-driven and lively GMK projects in recent years, while others see it as nothing more than “turning a dead meme into keycaps.”

    MW Philosopher’s Stone

    At the beginning of March, York published the IC post for Philosopher’s Stone, an alchemy-themed keycap set.

    “The design draws inspiration from medieval alchemical manuscripts, distilling the symbols, instruments, and philosophical core of the alchemical world into each tiny keycap. It also incorporates elements from the anime Fullmetal Alchemist.”

    The most interesting aspect of this set is undoubtedly its color design. Unlike many alchemy-themed sets that instinctively go for a straightforward “gold + black” combination, Philosopher’s Stone instead chooses a brownish-yellow tone reminiscent of aged parchment as its primary foundation. The Alpha section evokes the appearance of ancient books, scrolls, and oxidized metal surfaces, paired with slightly darkened legends. From the very first glance, the set carries a distinct sense of “age.”

    Meanwhile, the burgundy wine-red color serves as the most important visual anchor throughout the entire design.

    Whether on larger keys, the arrow cluster, or the novelties, it feels like the “result of transmutation” itself — intentionally extracted from the otherwise restrained and humble base tones. This color logic aligns closely with the concept mentioned in the original post: “base metals → aqua regia → Philosopher’s Stone.” The main body remains grounded and primal, while the reds and golds symbolize the higher-order material produced after the transmutation is complete. In addition, the set does not use conventional sublegends, instead incorporating a large number of alchemical symbols and cipher-like characters. Because these symbols naturally carry strong ritualistic and totemic qualities, they are able to establish a clear thematic atmosphere even if we do not fully understand their exact meanings.

    Under warm lighting, the entire set takes on the character of old books, ceremonial tools, and brass artifacts blended together.

    The novelty section is built directly around the alchemical system itself: the four classical elements, water/fire/earth/air, the stages of alchemical transformation (nigredo, albedo, citrinitas, rubedo), transmutation circles, the Emerald Tablet, the Kabbalistic Tree of Life, and various instrument symbols. It feels fairly unconventional and uncompromising in its approach.

    For pairings, warm gray, champagne gold, and silver cases are recommended, while white and highly saturated-colored cases are less suitable. Overall, rather than feeling like straightforward Fullmetal Alchemist fan work, this set leans much more toward a rustic and traditional interpretation of alchemy itself. It does not intentionally explore the darker side of alchemical themes, instead remaining relatively restrained. The transparent yellow elements and semi-transparent structures essentially simulate liquids mid-transmutation and glass laboratory vessels. However, because the overall brightness level is fairly low, pairing it with darker-colored cases can easily make the entire setup feel even dimmer. Personally, I think the barrier for achieving a good-looking build with this set is still relatively high.

    MW Gesha

    At the end of March, KENT published the IC update post for Gesha, a keycap set inspired by hand-poured coffee.

    “As a corporate workhorse myself, coffee every morning is absolutely essential (otherwise I’d genuinely fall asleep). I started out ordering coffee delivery, then moved on to drip bags. As I gradually learned more about coffee flavors, I began trying hand-pouring myself, and eventually it turned into a daily hobby.

    “Although there are already plenty of coffee-themed keycap sets on the market — GMK/DCS Cafe, SP SA Espresso, MW Barista, and so on — there still doesn’t seem to be a design specifically centered around hand-poured coffee. So, combining that with my own interests, I began developing the concept for MW Gesha.

    “First of all, the name Gesha actually comes from a fairly famous coffee bean variety. While thinking about the novelty designs, I wanted to include botanical elements inspired by coffee cherries (which, coincidentally, really do resemble cherries). So I borrowed this widely recognized bean variety as the theme name. Combined with illustrations of brewing tools commonly used in hand-poured coffee, the foundation of this design gradually took shape.”

    Coffee-themed keycaps are actually quite common within the community, and every designer tends to have their own interpretation. Most still revolve around the classic “milk white + coffee brown” formula — essentially white alphas with coffee-colored legends, or coffee-colored bases with white legends. Some attempt espresso-like gradient effects as well, though those can easily go wrong if not handled carefully. Gesha’s color philosophy, however, leans more toward expressing the process itself: using the gradual color transition of coffee beans from raw green beans to roasted beans as the central visual narrative, while translucent accent kits metaphorically represent the brewed coffee liquid, extending the concept from raw ingredient to finished drink.

    The main typing area uses a warm light beige as its primary base color, paired with soft light coffee-colored legends, creating a gentle texture reminiscent of lightly roasted coffee beans. The Mod section shifts toward a dark brown base with off-white legends for contrast, corresponding to deeper roasting stages. Transitional accents appear on keys such as Enter, where lighter coffee-colored bases are paired with off-white legends. Meanwhile, some extension keys use reddish translucent dark coffee-colored bases with off-white legends to represent brewed coffee itself.

    The novelty designs are one of the more interesting aspects of the set. The reddish tones of raw coffee cherries and brewed coffee are used as silk-print accents, while brewing tools such as moka pots and manual coffee grinders appear as graphic elements.

    Overall, the reddish silk-print accents do successfully function as a kind of “finishing touch,” preventing the design from becoming trapped within a monotonous brown palette. However, when looking closely at the illustrations themselves, some elements feel slightly underdeveloped. For example, the proportions and structure of the kettle handle, as well as the linework of the hand holding the coffee, carry a certain awkwardness. They are neither fully realistic nor stylized enough to establish a clear artistic direction, which slightly weakens the overall consistency and refinement of the design.

    For pairings, warm gray, off-white, and silver cases are more strongly recommended, as these neutral warm-toned housings naturally complement the beige and coffee-colored layering of the keycaps themselves, creating a softer and more cohesive overall appearance. If you want to emphasize contrast and layering further, dark brown or deep coffee-colored cases approaching black can also work well, though pure black is probably not the best choice.

    KBS Psychedelic House

    At the end of March, York published the IC post for Psychedelic House R2, a psychedelic-themed keycap set.

    “The inspiration comes from psychedelic mushrooms, blending hallucinatory visual effects and distorted consciousness into a bold aesthetic.”

    In terms of color design, compared to the blue-and-pink palette of R1 inspired by holographic film, the R2 colorway is clearly far more aggressive. The set collides extremely saturated fluorescent green, purple, and orange together. The Base kit completely abandons the traditional idea of “restrained balance,” instead directly creating intense contrast through purple-blue and fluorescent green. Even the legends themselves are rendered in bright neon green, giving the entire set an overwhelmingly aggressive presence from the very first glance.

    The novelty section is filled with repeatedly layered wave patterns, spirals, mushrooms, distorted symbols, and visuals resembling motion afterimages. It gives me the same feeling as electronic music spectrums or psychedelic concert posters — almost like the attack effects in Kirby after absorbing one of those laser-eye enemies.

    Personally, I think this set is extremely taste-dependent. There is almost no visual resting area anywhere across the entire design. At first glance, it immediately grabs your attention, but over longer periods of time, it inevitably becomes somewhat exhausting on the eyes. At the same time, the design heavily uses inverted fluorescent green and orange treatments, where identical graphics continuously switch between different background colors. This high-frequency visual collision further reinforces the theme of “mental contamination.”

    To me, it feels like York intentionally gave up compatibility with everyday desktop setups in favor of pursuing a purer, more subcultural form of expression.

    Personally, I think this set works best as part of a heavily stylized desk setup, a display-focused keyboard build, or as a collector-oriented piece for specific subculture enthusiasts. Pairing it with translucent PC or acrylic cases in unconventional colors would probably produce excellent results. However, for cases that already feature strong contrasting colors or highly stylized designs of their own, I don’t think the combination would work very well — because this keycap set already completely dominates the visual center of the build.

    GMK Arcade

    In mid-March, Rassles published the IC post for GMK Arcade, a keycap set inspired by arcade gaming.

    “I grew up playing video games and spent a lot of time in arcades, so this project gave me an opportunity to reinterpret those classics in a fresh way. GMK Arcade is a tribute to classic themes, rebuilt with vibrant colors and packed with references to iconic games, aiming to capture that nostalgic arcade atmosphere. A hit of pure nostalgia.”

    In terms of color design, the moment I first saw this set, it immediately reminded me of the color palette from GMK Miami Nights years ago. The original Miami is undeniably iconic, and countless later designs have drawn inspiration from it. The base uses pure black, likely intended to simulate the glow of CRT monitors inside dark arcade rooms. With ambient light suppressed, the colors on-screen seem to float within the darkness itself. The main typing area adopts a cyan tone reminiscent of Tiffany Blue, though deeper and more saturated, a color extremely common in arcade game interfaces throughout the 1980s and 1990s. The Mod section uses highly saturated magenta tones, while certain extension and novelty keys introduce bluish-purple accents, filling the hue gap that previously existed between the cyan and pink elements.

    Compared to the soft neon reflections of Miami — like lights diffusing through humid evening air at dusk — Arcade’s overall color palette feels far more vivid and direct.

    The novelty section revolves around classic early-era games such as 3D Pinball, The King of Fighters, and vertical shooters, with the overall execution remaining fairly straightforward.

    Personally, I think this set plays things relatively safe. The black-base neon formula has already been repeatedly proven successful through Miami Nights and countless derivative designs. Dark gray or black cases feel like the most natural extension of the theme. White cases raise the overall brightness and shift the aesthetic toward a cleaner, more futuristic vibe, and transparent PC cases create a similar effect. However, extra attention should be paid to RGB lighting implementation — if the lighting effects become too chaotic, the already highly saturated color palette can easily start to feel cheap rather than cohesive.

    That said, while there are not many brand-new Miami-inspired releases these days, there is still plenty of stock circulating, including in the aftermarket. Picking up a novelty kit alone could actually be a pretty fun option.

    KeyBoy40

    In mid-March, Niuniu published the GB post for KeyBoy40, a keyboard inspired by the Game Boy.

    “The Game Boy is an iconic milestone product in the history of handheld consoles. Previously, the GB34 from the same lineup was designed by Teacher Earth and became beloved among many 40% keyboard enthusiasts. Two years later, we’ve finally brought out what can truly be considered its upgraded successor.”

    Custom keyboards themed around the Game Boy are actually not uncommon — examples include Tetris60, AM RGB 65, GAME1989, and others. The predecessor to KeyBoy40, the #34, attracted a huge amount of attention largely thanks to its unbelievably aggressive pricing. While this new version has become noticeably more expensive, it still firmly belongs in the “affordable enthusiast toy” category around the thousand-yuan range, and both the overall configuration and build quality have also received meaningful upgrades.

    The exterior incorporates a number of sculpted details inspired by the Game Boy itself, such as the signature angled cut on the bottom-right corner, decorative grooves, accent lines across the top, anti-slip side textures, and a rear decorative plate. Overall, the design remains fairly minimalistic. Structurally, it uses a shell-style construction paired with either an O-ring or gasket silicone bean mounting system. This setup works particularly well on compact products like 40% boards, while also eliminating the annoyance of screw-based assembly. As for typing feel and related discussions, I imagine people who regularly buy Niuniu’s products are probably not choosing them primarily for that reason anyway.

    One of the major selling points of the KeyBoy40 is its dual compatibility with both mechanical and electrocapacitive switches. A single PCB supports both electrocapacitive switches and mechanical switches — though they cannot be mixed together simultaneously. “By flashing different firmware, you can freely switch between electrocapacitive and mechanical modes. For example, if you install mechanical switches, you simply flash the mechanical firmware either before or after assembly following a straightforward tutorial, and the KB40 will work normally. The same applies to electrocapacitive mode.”

    In terms of layout, the 40HHKB version offers both ortholinear and staggered configurations. The staggered layout uses a 7u long spacebar along with split-space options in 2.25u and 2.75u configurations, making it relatively keycap-friendly.

    The current group-buy version achieves compatibility with both mechanical switches and TEC electrocapacitive switches through the plate design, with additional support for screw-post and PCB mounting. Meanwhile, the more traditional electrocapacitive version used on the HX40 will later be offered separately as an add-on purchase within the customer group.

    As for colors, it still carries Niuniu’s signature approach: custom spray-painted finishes are available for an additional fee. Among the anodized options, I personally think the light gold finish looks especially good. Over the past two years, 40% keyboards have noticeably gained more visibility within the Chinese custom keyboard scene, gradually evolving from an extremely niche enthusiast category into a relatively established subculture of its own. Although the number of people who truly use 40% layouts as a long-term daily-driver solution remains small, these keyboards have instead found a more stable role as objects meant to be appreciated, collected, and played with. And part of that evolution undoubtedly comes from organizers like Niuniu, who have continued supporting the 40% format for years. Perhaps it is precisely through this kind of persistent experimentation and continued supply that what was once a fringe branch has gradually become visible to more enthusiasts. Hopefully, more and more people will discover these fun little creations in the future.

    Key Boy Advance

    At the beginning of March, Galo published the IC update post for KeyBoy Advance.

    “Many studios have already released custom keyboard products inspired by the classic Game Boy handheld, experimenting with various proportions and form factors. Inspired by this, I started wondering: why not move beyond a single reference prototype and explore a wider range of handheld console designs through reinterpretation? My first choice became the Game Boy’s successor — the Game Boy Advance.”

    Right after discussing the KeyBoy40 above, another GBA-inspired product immediately follows.

    In terms of overall form, KeyBoy Advance does not directly replicate the horizontally stretched proportions of the original GBA. Instead, it returns to the keyboard’s own rectangular foundation. Through adjustments to the layout and outer boundaries, the four corners are visually pulled inward, creating an “embedded zone” structure surrounding the typing area, while each corner retains its own dedicated region for detailed treatment.

    A recessed engraved line extends from the front edge toward the back of the chassis, forming a segmented relationship reminiscent of a battery cover panel. Meanwhile, both sides of the bottom edge rise slightly upward, giving the overall silhouette a subtle floating appearance.

    The right side concentrates most of the decorative elements: an independent indicator-light structure sits at the top, while below it, a recessed split-piece structure creates the impression of a speaker module, with color separation further enhancing the layered effect. The left side is comparatively restrained, preserving only a lightly engraved text mark at the lower-left corner, while the upper-left corner introduces an offset light-strip structure to rebalance the visual center of gravity.

    Structurally, the keyboard adopts a traditional top-and-bottom shell construction. Extremely small chamfers are used along the seams to recreate the appearance of injection-molded handheld casing joints. The sides incorporate split-piece and color-separated elements to reconstruct forms resembling palm rests. On the underside, the original prototype’s complex curved surfaces are simplified and reorganized into upward-curving arcs on both sides, visually echoing the upward lift of the front bottom edge and further reinforcing the floating aesthetic. In addition, the chassis introduces pressable shoulder-button structures on both sides, preserving the interaction points of handheld consoles while supporting customizable functionality.

    If previous Game Boy-inspired keyboards mostly stopped at surface-level visual imitation, then KeyBoy Advance offers much more of the interaction and collectible charm associated with the handheld gaming era itself.

    Internally, the board uses a fairly conventional TOP mounting structure. The layout is based on a modified dimple-style configuration combined with a recessed spacebar arrangement and HHKB layout. Spacebar options include 6U, dual 3U, and dual 2.75U + 1U (steel plate only). However, if you want a more symmetrical layout configuration, sourcing compatible dual 3U and dual 1.75U Shift keycaps is still fairly troublesome.

    In addition, the project also offers dedicated contrasting-color accessories, along with optional universal PC add-on components for creating mixed-color combinations. There are quite a few possible configuration combinations available. I’m still curious how the actual anodized finishes will look in real life, though.

    As for pricing, the kit is positioned around the 1.7k–1.8k RMB range. For a “small-scale themed product,” that pricing is admittedly not especially approachable, particularly for users who are attracted primarily by the design alone, since the barrier to entry becomes relatively high. However, considering the structural complexity and the number of individual parts involved, the production cost is clearly not low either. After all, Niuniu’s KeyBoy40 uses a much simpler integrated shell construction. Personally, I still hope this project successfully reaches production — because among recent works of this type, I think this is one of the more fully realized executions we’ve seen in quite a while.

    Matter 65

    At the beginning of April, Chq published the IC post for Matter 65, a 65% keyboard kit combining PC and copper materials. From the front, the kit follows a fairly conventional 65% layout, while the side profile takes the HHKB Line concept and reshapes it into a subtly curved arc. The rear also adopts a relatively standard two-stage design. But what truly makes this keyboard interesting is that it does not place its design emphasis on the outer silhouette. Instead, it uses the PC material to construct an “interior meant to be observed.”

    Structurally, it uses fairly standard O-ring and silicone-particle gasket mounting systems — safe and conventional choices overall.

    The copper component on the underside features wing-like cuts on both sides. The lightweight PC shell wraps around dark metal that has been carved, exposed, and visually emphasized, creating a very deliberate contrast between interior and exterior. At the same time, the center area overlays two completely different typographic styles: the clean and modern “natural,” alongside the more handwritten and emotionally expressive “artificial.” The entire kit uses contrast as a design language throughout — from materials and structure to graphical elements, everything continuously reinforces this sense of internal tension.

    That said, when viewed in the context of today’s market, this design also cannot really be considered particularly radical or groundbreaking.

    On one hand, the combination of PC and visible metal internals is no longer especially novel within the custom keyboard scene over the past two years. Whether it’s projects like Sha 65, Protagonist, or various other attempts at transparent-shell designs, many have already explored the idea of “internal visualization” to varying degrees. On the other hand, the mounting structure itself also falls into the category of relatively common “safe configurations,” offering users a level of optional flexibility that feels somewhat noncommittal.

    As for the micro-arc oxidation process mentioned in the IC post, I don’t think it needs to be overly romanticized. First, the moment an aluminum top case is chosen, the design’s core “internal visualization” concept is already weakened significantly, causing the product to quickly revert into something comparatively ordinary. Second, micro-arc oxidation itself is hardly new technology. Earlier projects experimented with similar treatments as well — products such as Bozi’s Cod67, for example, already used comparable processes. Its advantages lie more in differences in surface hardness and texture rather than any supposed “dramatic improvement” in typing feel or sound performance.

    A more practical issue is that micro-arc oxidation places heavy limitations on color options. The available range is nowhere near as versatile as traditional anodization, which is also one of the major reasons why the process has gradually faded from mainstream custom keyboard design in recent years.

    Finally, returning to the renders themselves: the official images depict the PC material as extremely transparent, but based on real-world experience, CNC-machined PC rarely achieves that level of visual clarity — especially when dealing with thicker structures and more complex internal geometry. Fogging and light diffusion issues are almost unavoidable to some extent. As a result, the final physical product will most likely appear more hazy than the renders suggest, and the visibility of the internal structure may not end up nearly as crisp or idealized as shown in the promotional images. Pricing has not yet been announced.

    Venus

    In mid-April, TRY published the IC post for Venus, a 40% keyboard kit inspired by sculpture.

    When talking about Greek-themed custom keyboard projects — such as Melgeek’s MG WAHTSY, Createkeebs’ Thera75, or Fotu’s Elysium — most of these designs still approach the theme through “symbols” and “imagery.” Their names, color palettes, and localized decorative details are essentially additive layers placed onto already familiar keyboard forms.

    Venus takes a slightly different path. Rather than emphasizing external symbolism, it starts directly from “form” itself, breaking sculptural language down into relationships between volumes and curved surfaces. It is not about adding sculptural elements onto a keyboard, but instead attempting to turn the keyboard itself into a sculptural object.

    TRY mentioned a particularly important point in the IC post: in traditional sculpture, the “front” is the side facing the viewer and establishing a relationship with the ground. But keyboards are different. Although they also have a “ground-facing surface,” our actual viewing angle while using them mostly falls somewhere within a 70–90° frontal perspective. So on Venus, the designer directly treated the “front” as the primary design reference point, organizing the entire form through continuous curves spanning the front, sides, and underside, allowing users to perceive a relatively complete shape from a normal usage angle alone.

    This idea directly influenced all subsequent details. Instead of prioritizing how the structure should be layered or assembled, the design follows a much more “form-first, structure-second” approach. Overall, it feels closer to designing an object meant to be observed.

    Following this logic, Venus also reinterprets the HHKB Line. The original line, which traditionally runs parallel to the bottom case, is interrupted and transformed into an angled break that becomes part of the side profile’s overall transition. To make the proportions feel more balanced, this transition point is intentionally lowered, shifting the visual center of gravity downward and toward the front while also filling the previously empty areas on both sides of the HHKB layout.

    In terms of execution, the design avoids typical chamfers or rounded corners, instead using a continuous concave surface to complete the transition. This surface extends from the front all the way to the side and then down toward the underside of the keyboard, with the curvature gradually changing throughout. Because the transition point is lowered, the side profile creates a visible “break point” near the bottom instead of maintaining the usual top-and-bottom symmetry. This treatment also echoes the “fracture” imagery mentioned in the design concept itself.

    The front-facing details are comparatively restrained, though two additional elements were still introduced. At the layout blocker position, a nameplate with a curved outline is slightly raised and tangent to the front surface, effectively extending the curved surface language upward by another layer. In addition, the very front edge features a rounded contour with changing curvature that corresponds to the concave surface above, creating an inward-contracting tendency through the center area. This section sits directly in front of the spacebar, making it the exact area where the thumbs naturally come into contact during use. As a result, there is also a subtle tactile design consideration involved here, somewhat reminiscent of Domikey’s approach to spacebar-area shaping.

    The side profile overall still maintains upper-and-lower symmetry, using two curved sections to connect the form and preserve the integrity of the volume. The rear is comparatively more restrained, relying on carefully controlled proportions within the weight design to keep the composition visually stable. The underside is divided into four separate weights surrounding a central recessed region. The concave lines here continue the same curved-surface language established earlier, and it actually reminded me somewhat of Lily’s bottom-case design. However, compared to the restrained and continuous curvature expression on the rest of the keyboard, the underside treatment feels noticeably more “filled.” The four weights combined with the recessed contours occupy almost all available visual space, resulting in a denser and more information-heavy appearance. Internally, the board uses a fairly standard silicone-ring O-ring mounting system.

    Venus will be available in anodized, coated, and PC variants, while the weights and decorative pieces can be configured in coated aluminum alloy, raw stainless steel with machined texture, PVD, or stonewashed finishes. Overall, I think the execution quality of this project is relatively high. Starting from the idea of “designing from the front viewing angle,” it consistently derives its curves, transitions, and break points through every level of detail, resulting in a fairly unified design language. In terms of pricing, the higher-end configurations sit slightly above the 2K RMB range, while the standard version starts at just over 1K RMB. If you’re interested in more sculptural keyboard designs that strongly emphasize curved surface language, this is definitely a project worth paying attention to.

    Do.25

    In mid-April, ises published the IC update post for Do.25, a retro-inspired keyboard kit rebuilt from the WS 785 platform.

    Unlike many retro-style keyboards that stop at recreating vintage visual language, Do.25 starts directly from the layout itself, compressing and restructuring the original Digital layout.

    At first glance, the layout somewhat resembles a combination of FKL and HHKB, though the actual differences are concentrated mainly within the alpha section. On traditional keyboards, the number row, Q row, A row, and Z row follow a staggered progression of 0.5U, 0.25U, and 0.5U offsets. This uneven arrangement largely exists as the result of historical and structural compromises. Do.25, however, standardizes all inter-column spacing — including the number row — to a uniform 0.25U. Without completely abandoning conventional typing habits, it simplifies the entire system into a more unified structure.

    Building on that foundation, the lengths of the modifier keys are also redistributed. Backspace is compressed down to 1U, while Shift is shortened to 1.75U on the left and 2.25U on the right. These adjustments help control overall key density while reducing reliance on uncommon keycaps, allowing the layout to remain compatible with standard base kits without significantly increasing the barrier to entry.

    The spacebar area is handled more aggressively. Although the left-side Ctrl key is retained, it is compressed down to 1U in order to preserve the original 0.25U offset relationship between the WKL blockers. This not only maintains the proportions of the blockers themselves, but also creates enough room to fit a 10U spacebar purely for the sake of “dumplings wrapped around vinegar” — though of course, standard spacebars remain supported as well.

    In terms of appearance, Do.25 inherits the foundational design language of WYSE, but rather than simply recreating it, the case structure is rebuilt around this unique non-standard layout system.

    Because the vertical proportions of the modifier section are not fully uniform, the front-facing negative space distribution naturally becomes uneven. Do.25 addresses this by re-dividing the case frame using additional lines. Without altering the overall outer silhouette, it reconstructs visual order by segmenting the rectangular structure itself.

    More specifically, the offset areas to the left of the number row and Z row are visually weakened or partially stripped of their frame structure, then transformed into layered depth through recessed treatment. In the upper-right corner, stepped segmentation creates a distinct isolated area, preventing repetition with surrounding elements. At the same time, the split lines along the side and bottom edges are shifted inward from their original centered positioning, while the relationship between outer frame width and top-and-bottom case boundaries is standardized, resulting in more stable overall proportions. Without relying on extra decorative elements, this approach gives what would otherwise feel like a relatively loose layout structure a much clearer external framework. Compared to directly emphasizing retro symbols, this method feels much closer to reconstruction.

    During later prototyping stages, several structural experiments were also tested and revised. For example, the early detachable F-row top cover introduced problems in assembly and interrupted line continuity, eventually leading to its removal. The USB-C port area was also redesigned: instead of using a more direct segmented cut, it now transitions through separated structural pieces and engraved recessed lines, extending the visual relationship toward the rear of the board. Meanwhile, some previously exposed structural elements were moved internally in order to reduce interference with the overall form. Of course, a separate top case and PCB will also be produced for the retro-accurate Doo version.

    Overall, Do.25 feels much more like a design project built around “layout” as its starting point — the exterior serves the layout, while the structure serves the exterior. The overall logic feels relatively unified, with a clear design direction and a certain degree of experimentation. At the same time, however, a design built upon a non-standard layout naturally demands more adaptation from the user, making the barrier to entry somewhat higher. Personally, I’m quite fond of these kinds of retro-innovation product designs. The standard anodized/RAW version is priced at 1888 RMB, while the coated version comes in at 1950 RMB. Definitely worth keeping an eye on if this sort of design interests you.

    MetaPulse

    In mid-April, Charles from MetaKeebs published the GB post for MetaPulse, an electrocapacitive accessory ecosystem designed for the custom keyboard market.

    When people talk about electrocapacitive keyboards, the common perception is often inseparable from ideas like “expensive” and “closed-off.” For a long time, electrocapacitive products were almost entirely dominated by original manufacturers, and combined with existing patent barriers, they never truly became part of the open custom keyboard ecosystem in the same way MX-style switches did. Agar previously attempted to launch its own electrocapacitive solution as well, but ultimately had to take it down due to patent disputes.

    Looking back at MetaPulse, however, its arrival was not something that happened overnight.

    Last year, Charles first developed his own electrocapacitive rubber dome molds and launched them through a group buy, taking the first step toward an independent electrocapacitive ecosystem. This was later followed by group buys for electrocapacitive PCBs, the EM60 electrocapacitive keyboard kit, electrocapacitive “crater” keycaps, and a series of related products. By the time MetaPulse arrived, it was no longer just a single standalone accessory, but rather an entire fully integrated electrocapacitive customization platform.

    MetaPulse covers most of the core components of an electrocapacitive keyboard, including the PCB, sliders, switch housings, rubber domes, and stabilizers. Compared to traditional electrocapacitive systems, its biggest characteristic lies in its expanded compatibility.

    The PCB follows the GH60 standard, making it compatible with mainstream tray-mount and O-ring-based cases, while the accessories support both crater-profile and MX-compatible keycaps. The switch housings use a cuttable design, allowing users to avoid interference with internally mounted screws, O-rings, and stabilizers.

    As for the rubber domes, two different feel profiles are offered. The Red series focuses on a moderately rounded tactile experience, somewhat similar to OEM Hybrids Snow. Meanwhile, the Blue series takes inspiration from OG BKE domes, pursuing a much heavier and more dramatic tactile event. Across both series, there are twelve different weight options ranging from 25g to 75g, with preset dome sizes covering 1u to 4u keys, allowing users to assemble keyboards ranging from 60% to 80% layouts without needing to cut the domes themselves.

    Personally, I’m genuinely happy to see electrocapacitive keyboards gradually moving toward a more open ecosystem. For many years, electrocapacitive boards felt almost like an “isolated island” outside the broader custom keyboard community — they had their own unique tactile philosophy, along with rich sound and rebound characteristics, yet always lacked a solution truly designed for custom keyboard enthusiasts. MetaKeebs has now provided that option, and hopefully more and more people will be able to experience electrocapacitive keyboards more easily in the future.

    Jahre 65

    At the end of February, Jingvv published the IC post for Jahre 65, a 65% custom keyboard kit inspired by the Porsche 911.

    When it comes to keyboards inspired by the 911, many people immediately think of the Singer 80 from a few years ago. But unlike Singer 80, which focused more heavily on expressing the theme through the backplate, Jahre 65 chooses to integrate the design language of the 911 directly into the keyboard’s silhouette itself.

    “The design inspiration for Jahre 65% comes from the first three generations of the Porsche 911 — the 901, 930, and 964. Even after decades of evolution, these cars still possess incredible charm today. Among them, the most iconic is undoubtedly the special-edition 964 model released in 1994 to commemorate Porsche’s 30th anniversary — the ‘30 Jahre 911.’

    “Compared to the standard Carrera 4, the ‘30 Jahre 911’ adopted the Turbo model’s widebody design. The fuller fenders, smoother waistline transitions, and more layered rear intake grille together created one of the most recognizable silhouettes in the history of the 911.”

    On the front side of the keyboard, the upper case introduces subtly tapering spindle-shaped curves along both sides. This makes the entire board appear visually thicker and fuller, similar to the horizontal tension created by the widebody proportions of the 964 Turbo. At the same time, the side profile incorporates an additional crease line that echoes the curvature of the bottom case, preventing the side view from appearing overly flat while instead creating a more fluid and layered surface transition.

    Many previous “sports car-themed” keyboards tended to focus on elements such as nameplates, engraved backplates, or directly borrowing highly recognizable features like headlights and spoilers. Jahre’s approach feels much closer to industrial design itself, without deliberately emphasizing “this is a Porsche reference.”

    A similar philosophy can also be seen on the underside. The curved bottom surface references the waistline contours of the 911’s rear fenders, while integrating the classic front-hood intake vent elements into the bottom structure itself, giving the underside much stronger visual layering. A black sandblasted nameplate sits below, finished with secondary flat polishing and anodization treatment, adding a touch of refined detail similar to automotive badging within the otherwise understated design.

    Internally, the weight system consists of two stainless steel weights whose overall form references the rear intake grille structure of the 911. The larger weight uses a RAW-machined surface finish, while the smaller weight employs mirror-polished PVD to emphasize metallic reflectivity.

    Structurally, Jahre 65 adopts a Leaf Spring Gasket system this time, along with a genuinely useful “dual-mode” silicone sleeve design.

    By flipping the orientation of the silicone sleeve, users can switch between “firm support” and “soft flex.” When the narrower side faces downward, the Leaf Spring region gains more suspended space, resulting in a bouncier and more elastic typing feel. When the wider side faces downward, the structure provides fuller support, creating feedback that feels firmer and more stable.

    In addition, Jahre 65 also offers a traditional Poron gasket configuration, effectively accommodating different user preferences regarding sound profile and flex characteristics.

    The mechanical-switch PCB uses a 1.6mm black-core board, supports 8K polling rate, and comes in both soldered and hot-swap versions. At the same time, it is also compatible with Venom’s 65% magnetic-switch PCB. As magnetic switches gradually continue entering the custom keyboard market, we will probably see more and more teams moving into the magnetic-switch ecosystem as well — it feels increasingly inevitable.

    As for colors, the current lineup includes green, gold, blue, silver, gray, ice blue, and orange. Personally, I prefer the orange version the most. Compared to the more restrained gray, silver, and darker tones, it carries much more of that classic sports-car “performance” feeling.

    In terms of pricing, Jahre 65 starts at 2399 RMB.

    Within today’s custom keyboard market, that price point already sits in an extremely competitive range. On one hand, users now have increasingly high expectations for build quality, structure, and detailing. On the other hand, there is certainly no shortage of products emphasizing value and aggressive specifications.

    What I think makes Jahre relatively special is that it never feels like a project focused purely on spec-sheet competition.

    Whether it’s the exterior lines, the bottom structure, or the design language built around the 911 theme itself, everything feels thoughtfully developed rather than simply collaging automotive elements together. This kind of genuine thematic cohesion is actually something many projects tend to lack.

    Overall, I personally think Jahre 65 is a highly refined project. Over the past few years, making keyboards more rounded, emphasizing curves and flowing surfaces, has no longer been anything particularly new. But Jahre gives me a feeling of being “soft within strength.” You can sense a certain power through its crease lines, metal components, and widebody-inspired silhouette, while the extensive use of rounded corners, curved surfaces, and waistline shaping prevents it from ever feeling excessively cold or rigid.

    Aepex 60

    At the end of April, KBDfans published the GB post for Aepex 60, a nature-inspired custom keyboard kit.

    “We are drawn to the quiet weight of ancient stone, and fascinated by the cliffside contours carved over time by wind, water, and the passing of ages.”

    The kit adopts a wraparound-style design where the top case sits slightly higher than the bottom case. The edges are softened with fine rounded corners, making the front-facing silhouette appear gentler overall and creating a subtle stepped transition. Combined with the rounded elements, the entire keyboard gives off a visually “enclosed” feeling.

    The treatment of the bottom case is especially distinctive. Large rounded corners are used throughout, but the side profile is not simply a smooth curve. Instead, it incorporates multidimensional cuts that gradually taper inward from top to bottom, forming a streamlined contour. This design becomes especially noticeable when holding the board or viewing it from the side — it does not feel like a straightforward rectangular block, but rather like a continuously shifting curve. The curvature also affects the perceived thickness from side angles, making the Aepex 60 appear visually heavier and thicker than most typical 60% keyboards.

    The backplate is where the design language becomes most concentrated. KBDfans incorporates mountain-like contours here, using continuous undulating lines and textured engraving to simulate cliffs shaped by erosion from wind and water over time. These lines are not flat decorative graphics layered onto the surface, but instead integrated directly into the three-dimensional curved structure of the bottom case itself. The contrast formed between different surfaces creates a certain sense of depth at first glance, almost like looking at a sliced geological terrain sample.

    Internally, the keyboard offers both gasket silicone-strip mounting and silicone-bean mounting options.

    The gasket silicone-strip setup is the more traditional approach. Long strips are placed between the plate and the case, and because the contact surfaces are continuous with larger support areas, the typing feedback leans more direct and firm. Bottom-out transmission feels crisp and decisive, making it suitable for users who prefer a more solid and stable typing experience.

    The silicone-bean setup follows a different philosophy. The point-based elastic support system uses much smaller contact areas, leaving more room for internal flex. During typing, the plate exhibits more noticeable compression and rebound, resulting in a softer, springier feel, while the sound profile also tends to become more muted. If choosing the PC top-case version, I would personally recommend prioritizing the silicone-bean mounting option. Black gasket strips become extremely visible beneath transparent or semi-transparent PC top cases. From either the front or side angle, having a black strip running visibly across the internals feels visually intrusive.

    As for color options, I personally recommend the stonewashed pine green finish. The color carries a slight grayness and weathered appearance — unsaturated and understated — which fits the kit’s overall geological and erosion-inspired theme extremely well.

    Pricing ranges from 2054 RMB to 2279 RMB. Compared to boards like the Tofu or Agar, it is definitely more expensive, though it also generally delivers what you would expect within this price bracket. Still, compared to something like the Pangea Mini, this one honestly feels like the better value overall.

    KINE

    In mid-May, LUMINKEY released KINE, a pad-style peripheral product.

    “There is still a lack of a pad that can truly break layout limitations and achieve highly flexible configurability. KINE, meaning ‘movement.’ It is not merely a static decoration on the desktop, but a dynamically transformable productivity tool.”

    In my personal impression, pads have always been a very “deterministic” product category — fixed layouts, fixed purposes, and even increasingly fixed aesthetics. KINE, however, allows users to actively reshape the pad layout itself, not to make it more complicated, but to better adapt to different desktop needs. Through four FR4 plates of different specifications, along with combinations such as reversible layouts, mirrored layouts, and full-1U configurations, it can switch between as many as seven different layout forms, reducing the likelihood of the pad eventually being left unused and gathering dust.

    “On the side, we recreated the classic button styling of the 1979 Sony TC-MR2 micro cassette recorder.”

    The appearance of KINE carries a very strong “device-like” quality. Many macro pads or numpads tend to look quite “boxy,” with the visual center focused entirely on the key area itself. KINE instead shifts much of its visual weight toward the functional area on the right side, giving it more of the feeling of a mixing console, industrial controller, or photography monitor.

    The key area itself does not use the deep surrounding frame structure commonly seen on traditional pads. Instead, through a shallower bezel and large amounts of negative space, the keycaps almost appear to float above the chassis. Especially when paired with these matte frosted keycaps, the entire product takes on an extremely strong “capsule-like” feeling — somewhat reminiscent of early Braun calculators, transparent consumer electronics, or certain experimental electronic devices.

    The horizontal rotary knob is also one of LUMINKEY’s signature design traditions, and on a pad product like this, it genuinely feels even more practical.

    Honestly, the most noteworthy thing about this product is probably its 649 RMB price point. Of course, with how aggressively mass-produced keyboards compete nowadays, six or seven hundred RMB is already enough to buy a fairly solid keyboard. But as a fully assembled custom product that also offers a genuinely interesting level of configurability and a strong visual identity, I still think this pricing makes it very worth considering.

    Katina TKL

    In mid-January, iNN Studio released the Katina TKL, a keyboard kit built around a softer and more decorative design language. Compared to many kits that emphasize aggressive lines, exposed structures, and complex surface cuts, Katina feels far more restrained.

    “Elegance, softness, and purity. We imagined Katina TKL as a pure and elegant woman, which not only inspired its external form, but also turned it into a deeper expression of emotion and aesthetics.”

    Katina’s bezels are not particularly narrow, and it still retains some of the substantial presence associated with traditional TKL keyboards. Yet the board itself never feels especially bulky. The reason lies in how the side profile is handled: the middle section noticeably tapers inward, creating a kind of “waistline” effect that effectively softens the sense of mass typically associated with larger keyboards. As a result, the overall silhouette, despite being relatively wide, avoids feeling overly rigid.

    That said, this approach also comes with its own trade-offs. Because the design relies heavily on curvature transitions rather than strong structural features, the initial visual impact is not especially aggressive. Many keyboards built around sharp cuts and bold edges can establish an immediate visual identity even in thumbnail form, whereas Katina TKL feels more like a product that gradually reveals its design logic through closer observation and repeated attention to detail.

    The most recognizable element on the front of the keyboard is the decorative section above the arrow cluster. This area takes inspiration from the architectural concept art of the “Gate of Order.” Instead of simply placing a badge there, the entire region is treated as a unified visual module. To the left of the badge, a three-tier stepped structure is introduced, echoing the progressively inward spatial relationship associated with the “Gate of Order.” The execution is not excessively complicated, but instead uses gradually narrowing layers to guide the eye naturally toward the center.

    Personally, I think this treatment works very well. Once many keyboards begin introducing decorative regions, they often fall into the trap of “visual overload” — textures, logos, lighting, and cuts all piled together at once. The result may technically contain many details, but visually it becomes chaotic. Katina TKL remains comparatively restrained. The stepped structure is not made excessively deep, nor does it rely on complicated patterns. Instead, the lines gradually tighten along the curvature beneath the badge before eventually returning to a flat surface.

    Another aspect I particularly appreciate is the large amount of negative space preserved on the right side.

    This is actually something many designs tend to overlook. Once decorative regions begin accumulating elements, designers often instinctively try to fill every available area. Katina TKL deliberately avoids doing that. A relatively quiet empty zone is intentionally preserved on the right side, preventing the decorative area from losing its sense of breathing room.

    From an actual visual standpoint, this kind of “negative space” is arguably even more important than the decoration itself. Since Katina TKL already follows a soft and restrained overall design language, introducing large areas of complex detailing above the function cluster would easily throw the visual balance off. The current arrangement — visually denser toward the left and more relaxed toward the right — ends up feeling much more comfortable.

    The central lighting element also uses a fairly clever design. Through the use of a silicone diffuser and an independent daughterboard, the light output becomes much softer and more evenly distributed, resembling a subtle atmospheric light bar. The engraved line beneath it is another particularly interesting detail. This recessed groove is CNC-machined directly into the aluminum top case, functioning as a kind of “guiding line” that visually connects the light bar, badge, and function-row contour together while further reinforcing the integrity of the decorative section as a whole.

    The rear uses a dual-arc weight design, emphasizing the relationship between the different curves. To me, the most impressive thing about the backplate is actually its sense of “completeness.” Rather than relying on exaggerated structural cuts, it builds a visual center around the middle area, then creates layering through the relationships between curves and material transitions. In particular, there is clear continuity between the curvature on the back and the shaping of the side profile.

    Structurally, Katina TKL adopts an internal mounting system based on iNN’s multi-common-point philosophy. This structure has already been validated on previous products such as the Flo-AT and Veil65, and supports three installation methods overall: PCB gasket, top mount, and O-ring configurations.

    As for additional finer details, reviewers who received prototype units will likely cover them more thoroughly. Pricing starts at 2099 RMB, while higher-end configurations begin at 3159 RMB. The standard configuration pricing feels relatively reasonable overall. Personally, my favorite colors are Spinel Purple and Mingshui Duck Green.

    Since 2024, iNN Studio has continued releasing new projects consistently. From their earlier, more experimental structural explorations to later attempts across different layouts, they have continuously tested the boundaries of the same underlying design philosophy across multiple product forms. Hopefully, they continue getting even better moving forward.

    Neson NoryTek-70%

    In mid-November 2025, Neson Design published the GB post for the NoryTek-70%, a keyboard kit themed around retro aesthetics and modern technology. Compared to the more experimental structural expressions seen in earlier projects from this series, NoryTek feels, to me, like a further exploration centered around the fusion of “order” and “information-driven interfaces.”

    From the front, NoryTek still retains the series’ signature recessed spacebar area, while the entire right-side function cluster is lowered downward as a whole, creating a distinct stepped height difference between the arrow keys and navigation cluster. Above that, an integrated module area is raised upward in reverse. According to the official description, this section is not a simple assembled component, but rather a single-piece machined structure directly integrated into the top case itself.

    Visually, the impact of this treatment is quite obvious: the front is no longer a single flat plane, but instead broken into multiple “height layers.” These layers are not merely decorative divisions, but structural elements actively participating in the visual language itself. At the same time, however, this multilayered structure also introduces a certain issue — as information density increases, the visual transitions between functional areas become more fragmented, especially under darker colorways or low ambient lighting conditions, where the layering itself begins to blur.

    In terms of spatial organization, a diagonal cut is introduced between the spacebar area and the arrow cluster. Its purpose is to break the stability of the horizontal structure, preventing the lower half from becoming overly regularized. However, this cut itself carries no direct functional or interactive purpose, instead acting more as a visual rhythm adjustment element. Boards like the Luminkey75 have explored similar treatments before.

    The primary visual focus on the front, however, is concentrated around the badge area. This section can arguably be considered the most recognizable part of the entire kit’s design language — a complete hybrid of interaction and decoration.

    The region is divided into two sections: on the left sits a dot-matrix-style light bar module composed of multiple small square elements, intended for status feedback or ambient lighting effects; on the right is a 160×40 pixel display used for UI elements, animations, or user-customized content. The relationship between the two is not simply parallel, but intentionally designed as a contrast between “information hierarchies”: one side leans toward static decoration, while the other focuses on dynamic information output.

    The light bar provides low-density background feedback, while the display handles high-density content presentation, with structural boundaries separating the two. But the issue becomes equally apparent — when the lighting effects, screen content, and surrounding textures are all active simultaneously, this region reaches a significantly higher information density than the rest of the keyboard. In static display scenarios, this can easily create an overly concentrated visual focal point.

    The side profile continues using the fairly common HHKB Line treatment, relying on a clean straight contour to control the overall proportions, while localized color separation further reinforces the sense of layering. The rear design leans more heavily into a designer-toy aesthetic, continuing the use of mixed materials: sandblasted electroplated stainless steel weights above and below, with a central sandblasted electroplated stainless steel “NoryTek” typography weight in the middle. Structurally, it uses a Leaf Spring Gasket mounting solution — relatively conventional overall.

    In terms of pricing, the aluminum standard version starts at 3100 RMB, while the PC version comes in at 3200 RMB. Personally, I think the pricing overall feels somewhat expensive. That said, if you particularly enjoy the combination of recessed spacebar layouts, retro aesthetics, and modern technological themes, it may still be worth considering.

    As for colors, I would more strongly recommend silver, retro green, and P2K. These colorways are better suited to carrying the layered front-facing structure and material contrasts of the design, while also more clearly presenting the hierarchical relationships between the different information modules. Darker colorways, on the other hand, tend to compress the visual layering of this design. Especially with both the front functional modules and the diagonal cut line existing simultaneously, darker finishes can make the segmented relationships feel less distinct, compressing the entire composition into a flatter visual plane and ultimately weakening the structural expression that defines the keyboard in the first place.

  • USB-C Isn’t Truly Universal: Why Sold-Out C-to-C Adapters Reveal a Fragmented Standard

    USB-C Isn’t Truly Universal: Why Sold-Out C-to-C Adapters Reveal a Fragmented Standard

    Recently, I bought a pair of lithium-ion AA rechargeable batteries. Compared to traditional NiMH rechargeable batteries, they’re lighter, have higher voltage, and even come with a built-in USB-C charging port.

    Lithium-ion rechargeable batteries with USB-C

    I thought I could finally get rid of that bulky NiMH battery charger. But to my surprise, they wouldn’t charge after I got them, so I contacted customer support. The reply left me speechless:

    Please use the included A-to-C cable for charging. This product does not support C-to-C charging.

    I stared at that cheap black A-to-C cable and fell into deep thought.

    If they all use USB-C, why isn’t it universal?

    USB-C ≠ USB-C

    While searching for a solution, I came across a video mentioning a newly released C-to-C adapter that can fix devices that don’t support charging via C-to-C cables. The name is quite odd—“5.1K resistor adapter”—and it sold out immediately after launch, with comments under the official video full of people asking for restocks.

    I had only heard of adapters like Lightning to USB-C or micro USB to USB-C—those that convert between different connector types. I never expected to see a USB-C to USB-C adapter for the same connector format. So while trying to grab one, I also discussed USB-C standardization, charging, and data transfer with others online. That’s when I finally understood the root cause of why my batteries wouldn’t charge.

    In short, the device didn’t follow the USB specification for setting identification resistors. As a result, the charger cannot determine whether it should supply power, and therefore fails to charge the device.

    This situation is quite common in small appliances such as handheld fans, portable lamps, and flashlights. They all use USB-C ports, but can only be powered using A-to-C cables.

    So why don’t manufacturers follow the standard design? And what exactly does the official specification require? Let’s briefly go over how USB-C is supposed to work.

    Further reading: Choosing a cable isn’t just about the connector — a guide to common USB and Thunderbolt protocols

    Introduction to the USB-C Specification

    The USB-C interface is highly versatile, supporting high-power charging and discharging, audio and video signal transmission, and reversible plug orientation. Precisely because of its rich functionality, its internal structure is also relatively complex.

    USB-C pin definition

    A full USB-C connector consists of 24 pins, with the A side and B side arranged in mirror symmetry. Based on function, they can be broadly divided into four categories: power, data transfer, control, and auxiliary.

    Power

    VBUS: A4, A9, B4, B9
    → Responsible for power delivery, defaulting to 5V and reaching up to 48V depending on the protocol

    GND: A1, A12, B1, B12
    → Ground lines that complete the circuit and ensure stability

    Data Transfer

    Low-speed channels: D+ / D- (A6, A7, B6, B7)
    → Basic USB 2.0 data communication (480 Mbps)

    High-speed channels: TX / RX (A2, A3, B2, B3, A10, A11, B10, B11)
    → Used for high-speed data communication such as USB 3 / USB 4 / Thunderbolt

    Control (Most Critical)

    CC: A5, B5

    • Determine plug orientation
    • Determine power direction (who supplies power)
    • Negotiate current and voltage
    • Enable fast charging / video modes

    Auxiliary

    SBU: A8, B8
    → Used for auxiliary audio or video signals (such as DisplayPort)

    As mentioned earlier, the missing identification resistor refers to a 5.1K pull-down resistor (Rd) on the CC pins. Without it, the device cannot be recognized as a power sink, so the charger will not supply power. This 5.1K resistance value is also the standard Rd value defined by USB-IF.

    However, the issues with USB-C are not as simple as just missing a “pull-down resistor.”

    A Unified Exterior, a Fragmented Reality

    USB-C is indeed an excellent connector form, but it is still far from achieving the USB-IF vision of “universal, simple, and unified device connectivity and interoperability.”

    Stripped-Down Connectors

    In practice, it’s rare for devices to use all 24 pins. Manufacturers often trim functionality based on actual needs. For example, many small appliances remove data-related pins and retain only the power-related ones—leaving just 6 pins, which is a reasonable cost-saving strategy.

    In fact, many devices previously used micro USB. Since the USB-A port on the charger side is always the power source by default, there’s no need to negotiate power direction like USB-C does, so the device circuitry didn’t include identification resistors. After switching to USB-C, some manufacturers chose not to redesign the internal circuitry to save costs, which is why these devices cannot be charged with C-to-C cables.

    In other words, these cables may wear a USB-C shell, but inside, they’re still the familiar micro USB.

    A USB-C female port with only 4 pins

    For example, the USB-C receptacle shown above has only 4 pins. It provides D+ / D- for USB 2.0 low-speed data transfer, along with VBUS and GND for power, but lacks CC pins. As a result, devices using this type of connector cannot be charged with C-to-C cables.

    In other cases, the connector includes CC pins, but manufacturers fail to solder the required 5.1K identification resistor. Some hands-on users have even added the resistor themselves to enable C-to-C charging.

    A manually soldered identification resistor

    Different Power Support

    Even if we only look at charging, C-to-C cables with identical appearances can vary greatly in charging speed. In my own case, my power bank can trigger 90W fast charging on a Xiaomi phone using the original C-to-C cable, while some other cables can only reach up to 20W. If you’re unaware of this, your expensive high-wattage charger might end up running at a much lower power level.

    To achieve 60W or higher charging power, you need to choose cables that support 3A or higher specifications.

    Cables supporting 6A current

    Expensive

    Nowadays, many monitors support a single-cable setup. With just one C-to-C cable connecting your computer and monitor, you can transmit video while charging your laptop, keeping your desk clean and tidy.

    However, anyone familiar with this setup knows that not just any C-to-C cable will work. You need a Thunderbolt 3 or higher standard cable, or a full-featured USB-C cable. These cables can cost several times—or even over ten times—more than regular C-to-C cables.

    Original iPhone cable, 6A cable, full-featured USB-C cable

    The Proliferation of Proprietary Charging Protocols

    You could argue that the issues above stem from hardware differences and cost constraints. But the proprietary charging protocols developed by many smartphone manufacturers—especially in China—are a problem at the protocol level.

    As early as 2014, Chinese smartphone makers began competing on charging speeds, pushing from 60W to 90W and even beyond 100W. At the time, official Power Delivery (PD) standards could not meet their needs, so they developed their own proprietary fast-charging protocols. Well-known examples include OPPO’s VOOC, Huawei’s SuperCharge, and Xiaomi’s HyperCharge. These modified protocols did achieve high-speed charging, even outperforming brands like Apple and Samsung in this area.

    However, proprietary protocols require a dedicated charger, cable, and compatible device to reach full speed. Once you switch brands or use multiple devices, compatibility breaks down, and charging speeds may drop to 18W or even lower. In some high-power chargers, these proprietary protocols may conflict with the standard PD protocol, leading to negotiation failures, power fallback, or repeated handshakes.

    In essence, proprietary protocols recreate new “ecosystem barriers” on top of the supposedly “unified” USB-C interface.

    Confusing Official Naming

    Beyond the inconsistencies caused by manufacturers’ cutbacks and modifications in hardware and protocols, repeated changes in naming by USB-IF have further increased the complexity for users:

    In 2008, USB-IF introduced the USB 3.0 standard.

    In 2013, USB 3.1 was released, renaming the original USB 3.0 to USB 3.1 Gen 1, while USB 3.1 became USB 3.1 Gen 2.

    In 2017, USB-IF renamed the standard again to USB 3.2, changing USB 3.1 Gen 1 to USB 3.2 Gen 1, USB 3.1 Gen 2 to USB 3.2 Gen 2, and adding USB 3.2 Gen 2×2 (20Gbps).

    ……

    TimeOfficial Standard (at the time)Old NameNew Name (at the time)Actual Speed
    2008USB 3.0USB 3.05Gbps
    2013USB 3.1USB 3.0USB 3.1 Gen 15Gbps
    2013USB 3.1USB 3.1 Gen 210Gbps
    2017USB 3.2USB 3.1 Gen 1USB 3.2 Gen 15Gbps
    2017USB 3.2USB 3.1 Gen 2USB 3.2 Gen 210Gbps
    2017USB 3.2USB 3.2 Gen 2×220Gbps

    Originally, it was already difficult to distinguish USB-C cables by appearance alone. These repeated official renamings have made things even more confusing, making it harder for users to tell them apart. As a result, some users created diagrams to mock this situation.

    Past vs Present

    However, careful readers might notice: we’ve been talking about USB-C, so why are we now discussing USB 3? This confusion actually comes from mixing up connector types and protocols.

    USB-C refers to the physical connector shape, while USB 3 refers to the underlying protocol. It’s just that the latest USB protocols mostly use the USB-C connector and are the most widely adopted, so people often confuse the two concepts.

    Connector vs Protocol

    Conclusion

    A few days later, my “5.1K C-to-C adapter” finally arrived. This tiny device adds the missing identification resistor, allowing the charger to recognize the connected device as a power sink and supply power accordingly.

    My problem was solved—but what about USB-C? It seems to have many issues: inconsistent implementation, fragmented protocols, and uneven user experience. But these may only be surface-level symptoms. The real issue is that USB-C uses a unified connector shape to mask a complex and fragmented ecosystem of implementations and protocols.

    Its problem has never been that it isn’t unified—it’s that it only appears to be.

    References:

  • A Perpetual Calendar Built on a DIY E-Ink Display

    A Perpetual Calendar Built on a DIY E-Ink Display

    In a previous article, I shared how I spent nearly two years building an e-ink display from scratch.

    System Architecture Overview

    For readers who haven’t read the previous article, here’s a brief introduction to the hardware and software structure of the project, along with links to the open-source repository (GitHub | Gitee).

    Hardware Structure

    The hardware setup consists of four main components:

    • ESP32 main controller: the brain of the system, responsible for receiving data, processing it, and sending display data. I’m using the ESP32-S3, as the larger memory is required to create a 400×300 buffer.
    • E-ink display driver board: acts as a bridge between the development board and the display. This was purchased on Taobao—just search for “compatible with Waveshare e-ink driver.”
    • 400×300 e-ink display: the final output screen. I used a retired off-brand e-ink panel from a supermarket, also purchased on Taobao. As long as it’s compatible with the Waveshare driver, it should work. Similar displays can also be found on second-hand platforms like Xianyu.
    • 3D-printed enclosure: organizes the messy wiring into a clean, product-like form factor.
    • Dupont wires: used to connect all the hardware modules above.

    Software System

    • The software is built using MicroPython, which I’m familiar with. The structure is relatively straightforward:
    • main.py: the main entry point of the project, where you can configure the operating mode of the e-ink display.
    • wifi.py: handles Wi-Fi connection and time synchronization. Based on configured network credentials, it connects to the internet and syncs system time.
    • wificonfig.json: stores the Wi-Fi SSID and password. In manual update mode, these can also be modified through a web interface.
    • ink_calendar.py: the perpetual calendar program used in fully automatic update mode.
    • ink_websocket.py: enables the ESP32 to run a server for passive and manual update modes.
    • ink_display.py: a general-purpose module for rendering content on the e-ink display.
    • epaper4in2.py: the driver for the e-ink display.
    • ink_index.html: a local HTML file used for manual updates.

    After publishing the first article, many readers asked whether the project could be open-sourced. I also realized there were many areas for improvement, so I recently spent some time iterating and upgrading the project.

    With these upgrades, the e-ink display now supports multiple update methods:

    1. Automatic update mode: updates the day’s calendar information automatically at 8:00 a.m. each morning;
    2. Passive update mode: after receiving an SMS, an iPhone Shortcut automatically syncs information to the e-ink display;
    3. Manual update mode: update content by entering text or uploading images through a web interface.

    Next, I’ll walk through these three modes one by one:

    Automatic Update Mode: Perpetual Calendar

    Effect Preview

    How It Works

    On the previous day, the server (currently running on my own computer) automatically requests relevant APIs based on the current date to retrieve the next day’s calendar information, and then renders it into an image, as shown below:

    The server then further processes this image into data readable by the e-ink display and uploads it to a public OSS location. Each client (i.e., the e-ink display device) will, at 8:00 a.m. every day, request the corresponding OSS address based on the current date, retrieve the image data, and render it on the display.

    In this mode, the image generation process is handled entirely on the server side (regardless of how many clients there are, the server only generates it once). The client only needs to connect to the internet and periodically request the corresponding address to update the e-ink display.

    Passive Update Mode: Sync to the E-Ink Display Upon Receiving SMS

    Effect Preview

    How It Works

    The ESP32 controller of the e-ink display runs in server mode, ready to receive and process HTTP or WebSocket requests. On the iPhone, a Shortcut is created to send incoming information to the ESP32 server.

    Then, another automation is set up on the iPhone: whenever an SMS is received, it automatically triggers the previous Shortcut to send the content to the ESP32 server.

    Since the ESP32 has limited performance and lacks image processing capabilities, it forwards the received data to a cloud-based “Function Compute” service. This service processes the data into image format and returns it to the ESP32, which then displays it on the e-ink screen.

    This mode also enables broader use cases, such as automatically syncing iPhone to-do lists to the e-ink display, displaying incoming emails, and more. Essentially, any information accessible via Shortcuts can be synchronized.

    However, this approach also has its drawbacks: since the mobile device does not send image data directly, and the ESP32 cannot process images on its own, a real-time online backend service is required. Although the cost of such serverless function computing is relatively low, it still introduces additional overhead.

    Manual Update Mode: Enter Text or Upload Images for Updates

    Effect Preview

    How It Works

    While thinking about better ways to solve the “server-side image generation” problem, I had a sudden realization one day: modern smartphones are already powerful enough—why not generate the image data directly on the phone and then send it over?

    Similarly, the ESP32 controller on the e-ink display runs in server mode, ready to receive and process HTTP or WebSocket requests. On the client side, a webpage (essentially a single-page application) is opened. Through this page, users can connect to the e-ink system over a local network, input text to generate images and cast them to the display, or upload images, crop them appropriately, and then cast them.

    During the casting process, the frontend webpage uses JavaScript to process text and images, transmitting the image data to the ESP32 controller via WebSocket. After receiving the image data, the ESP32 saves it as a file in its local directory, and the system reloads the image data from the directory for display.

    This mode does not require any additional cloud services, but it also has its limitations: since I haven’t solved the issue of requesting HTTP local resources under an HTTPS environment, it’s not possible to deploy this single-page application as an online service. Users need to download an HTML file and open it locally—on Android, this can be done directly from the file system, while iPhone users need to install an additional app (such as an HTML viewer) to open the local HTML file.

    Conclusion

    This e-ink display project has been ongoing for quite a long time. Each time I set it aside and learned something new elsewhere, I would come back and apply those learnings to further iterate and improve the project. Throughout the process, I’ve experienced a strong sense of accomplishment.

    There are still many aspects of the project that could be optimized, but since I don’t intend to turn it into a fully polished “product,” I may consider wrapping it up at its current stage.

    Some areas for improvement include:

    • Solving the issue of requesting HTTP resources in an HTTPS environment, which would make it possible to launch a universal single-page application. Users would no longer need to download an HTML file—just open a web page, enter the IP address, and update the display.
    • The ESP32-S3 controller is still relatively expensive. If the display can be updated without using a buffer approach, a more affordable ESP32-C3 might suffice.
    • The current 3D-printed enclosure mainly serves to tidy up the wiring and is relatively bulky. Optimizing the wiring layout could reduce the overall size.
    • The current power supply relies on USB. If only the perpetual calendar mode is used, the device only needs to perform updates once per day at 8:00 a.m., and can remain in deep sleep the rest of the time—making battery power a viable option.

    Anyone interested is welcome to build upon this and continue exploring.

  • Do You Really Have a Bad Memory? Common Myths About Memory and Forgetting

    Do You Really Have a Bad Memory? Common Myths About Memory and Forgetting

    When it comes to memory, many people share similar feelings:

    My memory has really been getting worse over the past few years.

    I unlock my phone one second, and the next I have no idea what I was about to do.

    Wait—when did I order this package? I have absolutely no memory of it…

    What did I eat last night? What did I do? Total blank.

    Same here. Just now, I noticed an opened can of soda sitting on my desk, yet I have zero recollection of when I took it out of the fridge. Could it be the legendary “snail girl” secretly helping me out? Impossible! Don’t tell me I’m getting Alzheimer’s?!

    Hold on—no need to panic just yet. This doesn’t mean your brain is sick, nor is it necessarily a sign of aging. What we often call “poor memory” or “being forgetful” is usually the result of misunderstanding how the brain actually works, mistaking its normal functions for bugs. Trying to fix everyday forgetfulness by practicing complex mnemonic techniques like the “memory palace” is often overkill—hard to stick with and largely unnecessary.1

    In contrast, understanding how your brain operates and adjusting your thinking patterns and daily habits is a much easier and more effective solution.

    Phenomenon 1: Forgetting as Soon as You Turn Around — What Was I About to Do Again?

    Many people have experienced this “forget-as-soon-as-you-turn-around” moment: you unlock your phone intending to quickly check the weather or track a delivery. Then you see a WeChat notification and tap into it. Just a few seconds later, when you return to the home screen, you freeze—why did I unlock my phone in the first place?

    This isn’t a problem that only appeared after smartphones became widespread. Similar phenomena have existed for a long time. In psychology, there’s a specific term for it: the “doorway effect.” You think of something you need to do—say, grabbing a pair of socks from the bedroom—but once you step through the doorway into the bedroom, you may forget why you went there.

    Psychological experiments have further confirmed this effect. Researchers randomly picked up objects from a table and placed them into a box, then later asked participants to recall what was in the box. In one group, participants stayed in the same room. In the second group, they moved to another room to recall. In the third group, they stepped out briefly and then returned to the original room.

    Even though the recall intervals were the same for all three groups, the latter two showed significantly more forgetting. The root of this phenomenon isn’t “memory capacity,” but rather “attention.”

    Working Memory vs. Long-Term Memory

    First, a brief introduction. Many people are familiar with the distinction between working memory (short-term memory) and long-term memory.2

    Working memory is the system the brain uses to temporarily store and process information. You can think of it as a type of short-term memory. Its characteristics are a very short retention time and a limited capacity3—if information isn’t deeply processed, it can disappear within seconds or tens of seconds.

    Corresponding to it is long-term memory. Long-term memory is like a vast warehouse, capable of storing memories in large quantities over long periods of time. The knowledge we’ve learned, our life experiences, and skills like riding a bike, swimming, or writing are all stored there.

    You can imagine working memory as a temporary workbench with a conveyor belt. Packages on this bench come from two main sources: one is external sensory information, such as what you see or hear; the other is internally generated thoughts or memories retrieved from long-term memory, such as “go get socks” or “check the delivery status.”

    Diagram illustrating the relationship between working memory, long-term memory, and attention (AI-generated image)

    There’s one crucial role here: the sorter at the temporary workbench—attention.

    Because working memory has limited capacity, only the packages that attention keeps a close eye on can stay on the workbench long enough to be further processed and stored in long-term memory. The moment attention slips, those packages slide off the bench and are gone for good.

    Back to the doorway effect. Whether it’s entering a new room or opening a new app, both are essentially scene switches. During such transitions, a flood of new sensory information rushes in—lighting, layout, colors, sounds, and more. Even if you’re not consciously aware of it, these external stimuli objectively hijack your attention, causing the original thought in your mind, or the sensory information you just captured, to be cleared from the limited workbench of working memory. When you snap back, the workbench is empty, and the original package is nowhere to be found.

    That’s why you end up standing there, spaced out, unable to remember what you were just about to do.

    How to Deal With the Doorway Effect

    The doorway effect is an instinct the brain evolved to adapt to new environments. It’s very similar to a computer clearing cache and freeing up memory so it can better load new programs and handle uncertainties in a new context. This is actually a sign of efficient brain operation. There’s no need to fight against this biological mechanism, let alone blame it on having a “bad memory.”

    Since the core issue is that scene changes shift attention, the countermeasures are straightforward: either protect your attention from being hijacked, or reduce the load on working memory in advance.

    1. Reduce distractions. The most direct way to deal with the doorway effect is to avoid attention shifts. Put frequently used items (such as scissors, keys, or your phone) in fixed, visible locations. Place high-frequency apps (like weather or payments) on your phone’s home screen or use widgets. By shortening the time and steps needed to find your target, you reduce the risk of being distracted by irrelevant information, allowing attention to go straight to the task.
    2. Repeat it out loud. When memorizing a phone number or license plate, we often repeat the digits to reinforce memory. The same applies here. When you’re about to go from the living room to the bedroom to get socks, or unlock your phone to check a delivery in a shopping app, quietly say it out loud: “get socks, get socks,” or “check delivery, check delivery.” Repetition forcefully locks attention onto the working memory task, preventing it from being replaced by new environmental information.
    3. Write it down. If you’re going into a room to get more than three items, or opening your phone to handle several tasks, the best approach isn’t relying on your brain but writing things down. To avoid the act of finding pen and paper becoming a new distraction, it’s a good idea to keep them within easy reach at home. Many people also pin notes or to-do list widgets to their phone’s home screen or set up shortcuts—these are effective strategies as well.
    4. Recreate the context. If you’ve already fallen victim and simply can’t remember what you were about to do, the best move is to go back to where the thought originated. If it came to you in the living room, go back to the living room. If it occurred while browsing a specific webpage, return to that page. Our memories are often bound to their environments, and the visual or auditory cues from that moment can reactivate the brain, helping you recover the lost memory trail.

    Phenomenon 2: Everyday Blackouts — What Did I Have for Breakfast Again?

    Besides forgetting things the moment you turn around, another everyday phenomenon that often makes people question their memory is these “blackouts” around mundane details. Most of us are probably familiar with situations like these:

    • You’ve just walked downstairs a few steps when a sudden wave of anxiety hits: Did I lock the door? Did I turn off the lights? Sometimes you even run back to check before you can relax.
    • Halfway through a shower, shampoo bottle in hand, you pause: Wait—did I already wash my hair, or was I just about to?
    • You take public transport or drive home after work and realize you can’t recall any details of the commute, as if you’d teleported. This becomes even more obvious when you’re tired, sleepy, or lost in thought.
    • Where did you eat lunch today, yesterday, or the day before? What did you eat? Chances are, unless you check your food delivery history, you won’t remember much at all.

    Everyone carries memories of their own life. This type of memory belongs to long-term memory and is called “episodic memory.” Some memories are so vivid that they remain crystal clear years later. Others, though they happened not long ago, leave only the faintest trace.

    Mainstream theories suggest that episodic memory is primarily distributed across the surface of the cerebral cortex (image source: The Comic Guide to Using Your Brain).

    Episodic memory has a unique feature: you don’t just remember what happened, but also when and where it happened. In other words, these memories come with “time stamps” and “location tags.”

    However, when it comes to highly repetitive and trivial daily episodes, the brain becomes surprisingly clever—or, from another perspective, surprisingly stingy. That’s when blackouts and forgetfulness start to appear.

    Everyday Memories: Merged and Compressed

    For daily routines like commuting, eating three meals a day, or washing your face and brushing your teeth—activities that are highly repetitive and emotionally flat—the brain usually performs an automatic “merge similar items” operation to save cognitive resources.

    Take breakfast as an example. Every morning, around 8 a.m., you take a bottle of milk and a ready-to-eat soft-boiled egg from the fridge, grab a slice of bread from the table, and spend five minutes eating at your dining table. Because the scene and content of breakfast barely change from day to day, the brain doesn’t store each breakfast as a separate memory file. Instead, it compresses thousands of past breakfasts into a single, generic, fuzzy “template file.”

    So when you try to recall what you had for breakfast yesterday, the brain first retrieves this generic template: the dining table at home, plus the usual trio of milk, eggs, and toast—rather than a specific snapshot of yesterday morning. And since yesterday’s breakfast had nothing particularly distinctive about it, lacking any standout features, that memory gets submerged into the routine, creating the illusion that you’ve forgotten it.

    Abnormal Memories: Individually Tagged and Stored

    So when do we clearly remember an episode? The answer is: when routine is broken.

    Anyone who has used smart security cameras will know about a standard feature called “motion detection.” When the scene is static and nothing changes—say, an empty hallway—the footage is overwritten in a loop, taking up minimal storage. Only when movement is detected, such as a delivery person passing by, do additional functions kick in and the footage gets saved long-term.

    Static state vs. motion detection camera logic (AI-generated image)

    Our brains use a similar mechanism when recording daily events. Back to breakfast: imagine that one morning, instead of milk and toast, you decide to cook a bowl of luosifen, filling the house—and yourself—with its unmistakable smell. This highly distinctive breakfast would trigger the brain’s recording mode. The brain detects the anomaly, slaps on a bright label, and archives it separately. Even years later, you’d still remember the embarrassment of eating luosifen for breakfast and getting side-eyed by fellow commuters on the subway.

    So if you can’t remember what you had for lunch last Wednesday, or what happened on yesterday’s subway ride, there’s no need to blame your memory. Everyday blackouts are a sign that your life is regular and stable, and that your brain is running smoothly, compressing those unremarkable daily memories to free up precious space for truly unique and important life events.

    How to Deepen Episodic Memory

    Once you understand how episodic memory works, the reverse is also true: if you want to remember something in daily life, or make a memory more vivid, you need to intervene actively and give it more distinctive features.

    Here are a few effective methods:

    • Engage multiple senses. If you’re experiencing a special moment—say, an anniversary dinner with your partner—and want to imprint it deeply in your mind, actively engage multiple senses. Smell the aroma of the food, notice the unique patterns on the tableware, listen to the background music, feel the texture of the chair. The more sensory details you collect, the richer the memory becomes, and the less likely the brain is to treat it as just another ordinary event and merge it into a generic template.
    • Leave traces. Many people struggle with forgetting whether they’ve taken their medication, especially when it needs to be taken regularly over long periods. At that point, taking medicine has become an automated action, leaving little impression in memory. “Leaving traces” is an effective way to remember such repetitive events. For example, place a calendar or sticky note next to the medicine bottle and check it off each time. Or follow medical advice and bind taking medication to another routine with the same cycle—put morning and evening meds next to your toothbrush and make a rule that you can only brush your teeth after taking them. This way, brushing your teeth reinforces taking medicine, and whether you’ve brushed your teeth also tells you if you’ve taken your meds.
    • Ritualized actions. For everyday compulsive behaviors, adding a fixed ritual can be very effective. When locking the door or turning off the gas, for instance, introduce a ritual check, such as the “point-and-call” method commonly used in industrial settings: point at the gas valve or lock, look at it, and say out loud, “Gas off,” “Door locked,” before leaving. This ritual labels the otherwise subconscious action, making it easier to retrieve later. If compulsive behaviors are severe, they may stem from physiological, psychological, or environmental factors and require professional medical treatment.
    • Write it down. It must be admitted that even with rich sensory input, special cues, and ritualized actions, the brain is still not good at remembering repetitive life details. The most reliable approach is to lighten the brain’s load by using a “second brain”—external tools. For example, every time I park in a mall now, the first thing I do is take a photo of the floor number and parking spot. Precious memories with friends and family are stored in photo albums, and I’ve developed the habit of journaling, supplemented by chat histories, browser history, and emails to aid recall and retrieval.

    Common methods for strengthening episodic memory (AI-generated image)

    Of course, as we age, the hippocampus—the brain region responsible for episodic memory—does gradually decline in function. This is a natural physiological process. So if you find your episodic memory isn’t as sharp as it used to be, don’t be overly anxious or self-critical. Accept a certain degree of forgetting in daily life, and save your limited energy for the people and moments truly worth remembering—that’s the best way to face time and aging.

    Phenomenon 3: Read It, Forgot It — What Was That Article About Again?

    Today’s learning tools and productivity apps are more convenient than ever. Notes used to be handwritten; now OCR lets you copy text with a tap. Research used to mean clipping and filing; now browser extensions save everything instantly. Long articles once required careful reading, books page by page; now AI can summarize everything in a minute.

    Behind this smooth flow of collecting, hoarding, and speed-reading, however, lies an easy-to-fall-into illusion of understanding. Many people notice that articles they’ve read and knowledge they’ve learned seem easier to forget than before. It can feel like learning ability has declined or memory has worsened. In reality, this is often the result of inappropriate learning methods.

    Many members of the community have written about this before. In earlier articles discussing learning principles and core elements, I also went into detail about how to address this issue (see the third section). If you’re interested, you can check those out. Here, I’ll explain from the perspective of memory mechanisms why this “learn fast, forget fast” phenomenon happens.

    Storage Strength vs. Retrieval Strength

    Earlier, we talked about episodic memory, which is a type of long-term memory.

    Long-term memory has many forms. The one most closely related to learning is called semantic memory. You can think of it as the brain’s database or library: all the knowledge points, facts, concepts, vocabulary, and formulas you learn are stored there.

    Semantic memory is believed to be widely distributed across the cerebral cortex (image source: The Comic Guide to Using Your Brain).

    When learning, many people—including myself—often feel that they have a bad memory and forget things right after learning them. Strictly speaking, though, “forgetting” isn’t an accurate description. From a cognitive psychology perspective, information that has truly been stored in long-term memory is extremely difficult to erase completely. As mentioned earlier, long-term memory is like a vast warehouse: once something is learned, it’s placed inside. The real problem is usually not that the memory is gone, but that the index is lost—the cue pointing to that memory can’t be found.

    Psychologists Robert and Elizabeth Bjork proposed that memory has two independent measures: storage strength and retrieval strength (Bjork, 1992).

    Storage strength refers to how firmly something is retained once it has been learned. This view challenges the traditional idea that “memories fade,” arguing instead that once long-term memory is formed, it is stored in the brain almost permanently. Retrieval strength, on the other hand, refers to whether you can access that memory at a given moment—how easily it comes to mind. What we usually call “forgetting after remembering” is actually a weakening of retrieval strength: the memory is there, but hard to retrieve.

    Storage strength vs. retrieval strength: four quadrants (AI-generated image)

    Combining these two dimensions gives us four types of memory states:

    • Low storage strength + low retrieval strength: shallow memories that are easy to forget.

    For example, an article skimmed casually, or a piece of trivia glanced at in a short video. We’ve all had this experience: you’re almost done with a book, movie, or post, and suddenly realize—wait, I’ve seen this before? Because it was never deeply processed, it left only the faintest trace in the brain. Without rereading or rewatching, you’d never be able to retrieve it. This is the worst-hit area for “knowledge hoarding.”

    • Low storage strength + high retrieval strength: shallow memories that haven’t faded yet.

    Crammed exam material, a restaurant queue number, a hotel room number on a business trip, or the plot of a mystery series binged overnight all fall into this category. Right now, the content feels clear, giving you the illusion that you’ve “remembered” it. In reality, storage strength is low, and once the task ends—after the exam, checkout, or finishing the meal—it quickly disappears. These memories are often tied to cramming and massed practice, and are best reinforced with spaced repetition.

    • High storage strength + low retrieval strength: firmly stored memories you can’t recall at the moment.

    Think of an old QQ number you used for years, familiar English words you haven’t used in ages, or a song you haven’t sung for decades. These memories aren’t gone; they still exist deep in your brain. But because they haven’t been used in a long time, they don’t come to mind immediately. For this type of knowledge, you usually don’t need to relearn it from scratch. A small cue—like being told the first two digits of the QQ number—can reactivate the memory. Regular maintenance and occasional review are enough.

    • High storage strength + high retrieval strength: memories that are solid and instantly accessible.

    Basic facts (China’s capital, the multiplication table), your own birthday, a phone password you’ve never changed, or gossip about an idol you’ve followed for years all fit here. No matter your state, you can recall these without effort. This knowledge has been fully internalized and become part of you. This is what it truly means to have “learned” something. The ultimate goal of learning is to move knowledge from the earlier categories into this automated state.

    The “Desirable Difficulty” of Better Memory

    Looking across these four categories, a pattern emerges. Learning methods that feel comfortable and frictionless tend to fall into either ineffective hoarding (“low storage + low retrieval”), such as one-click saving, photographing lecture slides, or endlessly downloading resources; or cramming modes (“low storage + high retrieval”), such as rereading, highlighting, immediately checking answers, or following tutorials step by step. These approaches don’t build durable memory—and the issue isn’t memory ability itself.

    So what’s the right approach? The Bjorks proposed a counterintuitive theory called “desirable difficulties” (Bjork, 1994). It explains a core mechanism of memory: increases in storage strength are inversely related to current retrieval strength. In simple terms, the harder it feels to retrieve information now, the deeper it will take root in your brain later.

    The principle of desirable difficulty (AI-generated image)

    This may sound abstract, but anyone with fitness experience will recognize it. Easy learning is like lifting an empty barbell: you can do it effortlessly, but without tearing muscle fibers, there’s no growth. In contrast, lifting with proper form to the point of fatigue causes muscle fibers to tear, followed by rebuilding and strengthening.

    The same applies to learning. Reviewing new knowledge immediately often yields mediocre results. The most effective approach is to allow some forgetting, then close the book and actively recall what you learned. This process may make you frown, scratch your head, and feel mentally stuck—but that discomfort is precisely what most effectively strengthens storage.

    What Is Forgetting Actually Good For?

    Once we understand different memory mechanisms, we can take a fresh look at the meaning of “forgetting.” In the learning process, forgetting is far from useless—in fact, you could even say that without forgetting, there would be no learning at all. It plays three crucial roles:

    First, at the working memory stage, it acts as a filter for attention.

    As mentioned in the first section, an overload of information from the environment can overwhelm the brain. Forgetting helps filter out irrelevant interference, allowing the brain to focus on what matters most. It functions as a selection mechanism: only information that is truly valuable has a chance to be further processed and stored in long-term memory.

    Second, at the long-term memory stage, it is a necessary condition for strengthening storage and forming cognitive abstraction.

    According to the theory of “desirable difficulties,” forgetting actually creates the necessary difficulty. Precisely because forgetting lowers retrieval strength, you have to exert effort to recall; and precisely because that effort is required, the brain judges the information to be important, significantly increasing its storage strength.

    In addition, forgetting helps us discard trivial details and extract the essence, enabling better abstraction and generalization. It’s like recognizing a face: when you first meet someone, you may remember many detailed features—skin condition, hairstyle, makeup. But later, when you see them again, their hairstyle or makeup may have changed, yet you still recognize them. That’s because the brain has helped you forget certain details, allowing you to better generalize their facial characteristics and form a stable, lasting memory model.

    Third, at the level of everyday life, it is an essential means for maintaining inner order and adapting to the present environment.

    Imagine if every heartbreak or every awkward moment remained as vivid and intense as when it first happened—our minds would be overwhelmed long ago. Through forgetting, the brain strips away the emotional overload from memories, leaving us with lessons learned while gradually fading the sharp pain, allowing us to recover from trauma. Forgetting also clears out outdated knowledge. For example, when you move to a new home, you must let go of the old address and routes so the new ones can be stored smoothly. If old memories never faded, they would linger like ghosts, constantly interfering with present life and causing far more trouble than we might expect.

    If a person were unable to forget anything, their brain would instead descend into chaos due to information overload, resulting in impaired cognitive function.

    There was a memory prodigy in the Soviet Union named Shereshevsky, who possessed an almost limitless memory. Yet because he could not forget, his brain was flooded with massive amounts of sensory detail. This left him severely lacking in abstraction: he couldn’t grasp poetic metaphors or perform complex logical reasoning.

    A similar phenomenon can be observed in some individuals with high-functioning autism. They can record scenes like a camera, but precisely because the details are too numerous and too vivid, this abundance hinders their ability to understand the social context and core meaning behind a scene. Even minor environmental changes can make them feel uneasy and anxious. This again shows that without pruning details, the brain cannot build meaningful models. In this sense, everyday forgetfulness is often the brain performing routine “garbage collection” and information filtering to maintain efficient operation—it is not the functional decline we so often fear.

    With these principles in mind, if we revisit the memory-strengthening methods discussed in the previous article—such as active engagement, writing, spaced repetition, and sleep—their logic becomes much clearer. There’s no need to repeat them here.

    Summary

    Although we talk about memory and forgetting every day, memory is not a single, unified ability, and forgetting is not an entirely bad thing.

    Different types of memory (AI-generated image)

    In this article, we discussed working memory and long-term memory, and within long-term memory, the declarative (explicit) forms of episodic memory and semantic memory. Due to space limitations, another important member of the long-term memory family—procedural memory, which belongs to non-declarative memory and includes skills like tying shoelaces, riding a bike, or reading in one’s native language—was not covered here. If there’s an opportunity in the future, I’ll dedicate a separate piece to it under the theme of “practice.”

    If there’s interest, I’d also like to write another article exploring common memory disorders and classic cases—such as age-related memory decline, the tip-of-the-tongue phenomenon, or even severe amnesia caused by illness—to further deepen our understanding of how memory works.

    Further reading: Incomplete Reports on the Use of Attention

    1. This isn’t to say the Memory Palace method is useless. It’s more suited for situations where you need to cram large amounts of illogical information into your memory in a short time—like memorizing the order of playing cards, pi, or lists of obscure vocabulary—rather than for tackling everyday forgetfulness issues like those mentioned above.


      ↩︎
    2. Strictly speaking, the field of psychology features numerous schools of thought regarding memory models (such as the Baddeley model, which incorporates complex structures like the central executive system and phonological loop), and definitions of working memory and short-term memory also vary. For the sake of clarity, this paper employs a simplified, universal model and does not make a strict distinction between the two at this stage.


      ↩︎
    3. Earlier research suggested that working memory could hold at most 7±2 items (Miller, 1956); current studies generally agree that its actual capacity is around 4 “chunks.” For instance, we struggle to memorize 10 random digits, but grouping them into sets of three—such as years or phone numbers—makes recall effortless. Each set of digits here constitutes a chunk. ↩︎