Stringers
justrach/nanoui · justrach

One 3,000-follower person, two tweets, 830 stars: how nanoui took the Zig slot in the "native GPU UI" wave

What it is A UI framework written in Zig for desktop and mobile apps. Either draw native UI on the GPU yourself, or wrap the system webview for a web frontend. The resulting program is about 0.7MB.

Why it took off It rode the "no Electron, draw UI on the GPU" trend and filled the Zig gap left after the Go and Rust versions. The author kicked off the first wave with a demo video showing four hard numbers, then the second wave came when his own launcher, nanolight, went viral.

BHow you can follow it Don't build another Zig version. Pick a language this trend hasn't covered yet and port a framework like mygo that's gaining momentum. Write a comparison of numbers against the original in the README, then ship a small app ordinary people will use as a separate release to pull people back to the framework.

Grade B: "port a hot framework + drive traffic with a companion small app" is doable solo, and the window is still open. But Go, Rust and Zig, the three hottest languages, are already taken, and whether the remaining language communities will buy in is unverified. It also leans heavily on the author's own X account and demo skills. Being talked about in small circles
Data (updated daily) · Repo created 2026-10-18 · 862 stars · 7 days +352
GitHub · stars
20040060080010-2010-2510-3011-0311-08862
Data 2026-10-20 → 2026-11-08 (20 days, 20 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

nanoui is a UI framework written in Zig for desktop and mobile apps. It hit 830 stars in 19 days. It isn't a new architecture; it's the Zig version of mygo, a Go framework that went viral a month earlier. The stars came mainly from two tweets: one with the framework's demo video, and one with the ⌘Space launcher the author built with it. If you want your open source project to earn stars, this playbook is worth copying. As for the Zig slot, someone has already taken it.

Oct 18 tweet

On 18 Oct 2026, in the evening (UTC), Singapore indie developer Rach Pradhan (on GitHub as justrach, on X as @rachpradhan, about 3,300 followers at the time) posted a tweet:

introducing nanoui: desktop apps in Zig, drawn on the GPU. Natively. super low latency 0.06 ms frames a SUPER nano 0.7 MB binaries AND a 0% CPU when idle WITH zero dependencies AND no webview every frame of this video was drawn by nanoui itself (without a window).

The attached video shows nanomon, an activity monitor he built with nanoui, with every frame drawn by the framework without opening a window. The tweet got 64k impressions, 1,580 likes, and 1,175 bookmarks (X, as read by 11-06). The bookmark count is close to the likes, meaning people saved it as something to come back to. The repo was created that same morning, and two days later it had 285 stars.

What it actually is

A UI framework written in Zig. Each window is one of two options:

  • Native UI: write only Zig, no HTML, no JavaScript, no webview. The framework ships flexbox layout and animations. On macOS it draws with Metal and CoreText, one draw call per frame. It also implements its own Apple Liquid Glass-style glass effect.
  • Web frontend: wrap the system webview. A JS client that the frontend calls is auto-generated from Zig types, with TypeScript declarations included.

It builds to a single-file program, about 0.7MB, and uses no CPU when idle. Over 19 days it shipped v0.11.0 (11-03), then gradually announced iOS support (10-31, one post: "nanoui is coming to iOS, 440 KB"), Android, Linux and Windows.

The Credits section of the README is candid: the architecture follows egoist's Go framework mygo, the chained API and virtual list are borrowed from the GPUI in the Zed editor, and the animations and components each have their own sources. Even the tagline nearly copies mygo's "Develop desktop apps with native UI or web frontend in Go", just swapping Go for Zig. The performance section uses mygo directly as the baseline: about 5x faster per frame and about 17x smaller (mygo is 11.8MB).

How the stars grew

The curve is clear: two waves, with a 10-day plateau in between (daily GitHub stars):

TimelineWhat happenedstar
10-18"introducing nanoui" demo tweet, 64k impressionsTwo days, 0 → 285
10-21Japanese tech outlet gihyo.jp ran an article, "Zig製GUIツールキット「nanoui」"Same day +78, reaching 363
10-22 ~ 10-31Only sporadic feature tweets (nanoterm, iOS)+5 to +15 per day, reaching 484
11-02"introducing nanolight": a ⌘Space launcher built with nanoui, 1.6MB, 113k impressions, his highest-exposure tweet to datenanolight hit 193 stars the next day; nanoui went +11-03 78
11-03 ~ 11-05nanolight 0.2.0 supports three desktop OSes and can run Raycast extensions; the English repost account @rammcodes posted "WAIT WHAT?? … without Electron"; the Chinese accounts @xiaoanlab and @QingQ77 introduced nanolight in turn11-04 +86,11-05 +105 (single-day peak)
11-06830 stars, 39 forks

The second wave is what matters. "Yet another UI framework" only gets clicks from people who write code. "A 1.6MB Spotlight / Raycast alternative" is something a regular Mac user can understand and wants to install. nanolight catches the traffic up front, and the first screen of the README tells you it's built by nanoui, so people who don't write Zig click through to the framework. In 19 days the author shipped several small apps: nanomon, nanoterm, nanostudio, nanofly, nanolight, each with its own "introducing" post. Most didn't take off (nanomon has 97 stars, nanoterm has 43), so one hit is enough.

It wasn't on Hacker News, no major tech outlet covered it, the repo was under 60 days old so it couldn't get onto GitHub trending, and I didn't see any paid promotion. The fork-to-star ratio, the two-wave curve and the tweet timing all line up, and it doesn't look like farmed stars (inference; I didn't check individual stargazers).

Why now, and why this one

Timing. In 2026, from 9 to November, "skip Electron, draw the UI yourself on the GPU" is taking off. mygo was created in Sep 26 and went from zero to 4,797 stars in a bit over a month, then added 1,119 in the last 7 days. Zed's GPUI spawned a batch of new libs, and longbridge/nsh is up +613 in the last 7 days. The opencode team also released rune (GitHub, 11-06), a Rust framework, in Oct 22. In Nov 1 someone on X wrote "GPUI / MyGo / Rune / I forgot the Zig one": the Go version exists, the Rust version exists, and the Zig slot is nanoui. The existing Zig GUI libs (dvui with 1.7K stars, capy with 2.7K stars) are years old, have no demos like this, and come with no companion apps.

Hard numbers. 0.06ms, 0.7MB, 0% CPU, zero dependencies: four numbers on separate lines plus a video of glassy effects hits the "sick of Electron eating my RAM" feeling head on. One reply read "maybe we don't need more RAM, we need more low level softwares."

The author had a track record. He'd previously built nanobrew (on HN, 225 points), codedb (1.4K stars) and turboAPI (1K stars), so he already had a group of people interested in Zig and performance. Those people were the first wave of reposts that cold-started this.

Moved fast. mygo 9-26 came out, then nanoui 10-18 shipped, three weeks apart. The framework ships with a --json CLI, AGENTS.md, headless screenshots and tests, which looks built for coding agents to write (inference); that probably also explains how one person got to v0.11 in 19 days and built five or six companion apps.

He also replies under other people's tweets about native UI with "or they can use mygo or nanoui!", riding the original's momentum. For technical questions he just tags @grok to read the repo and answer.

Who it's competing with

ProjectLanguageScale (GitHub, 11-06)Relationship to nanoui
egoist/mygoGo4,797 stars, +7 in 1,119 daysPrototype, still accelerating
Zed GPUI and its ecosystemRustzed 92.8K;nsh 1,486More mature and heavier; nanoui borrowed its API style
anomalyco/runeRust562 stars, +7 in 247 daysSame wave, backed by a team
ElectrobunZig / TS13K starsUses a webview, doesn't draw its own UI
dvui / capyZig1.7K / 2.7K starsEstablished Zig GUI; no demos, no mobile
Tauri / Wails / ElectronVarious37K–124K starsWebview route, the thing nanoui is pitching itself against

I haven't seen anyone actually copy nanoui yet. Someone on X already called it "not well maintained" and argued for using GPUI directly. Only one or two people have built real apps with it so far (a third-party email client, nanomail, and a small utility). The star count is high because the demos look good, not because lots of people are using it.

If it were you

Verdict: the playbook is worth copying, and it's learnable. Zig's slot is already taken. "Port a trending framework to another language, benchmark against the original, then build a small app normal people understand" is a combo that doesn't need a big company or money, and one person can do it.

Who it suits: someone fluent in a language that has its own community but hasn't been touched by this wave yet (Swift, C#, Kotlin, Odin, Nim and OCaml all look empty, though I haven't verified whether anyone is working on them), who can ship something working in three to four weeks, and who already has a thousand or two followers from that community on X.

Who it doesn't suit: anyone who wants a framework that a lot of projects actually adopt. That takes years of maintenance, and it's won on docs and ecosystem, not two viral tweets. If you have no social accounts and expect "good code will bring people" to work, this won't either.

How to do it: (1) pick an original that's still accelerating within the last 7 days (right now that's mygo and the GPUI-family projects), port it to your language, and put the comparison numbers against the original in the first screen of the README; (2) keep the demo video to three or four hard numbers with one striking visual; (3) beyond the framework, every week or two build a small app regular people can use (a launcher, an activity monitor, a screenshot tool), publish each one separately, and put "built with xxx" on the first line of the README; (4) or skip the framework: use mygo or nanoui to build a breakout app and ride their momentum, collecting stars on your own repo.

How to get users: your own X account is the main battlefield, posting demo videos; show up under tweets discussing the original; Japan's gihyo.jp and Zenn, and Chinese tech repost accounts, will proactively share this kind of "small and fast" stuff, so reach out after you post. HN wasn't used this time, so a Show HN is worth doing.

When to do it: the native GPU UI wave is still rising, and mygo is still adding thousands of stars a week. Probably before the end of 2026; once mygo's weekly gains drop to a few hundred, a port won't get any attention.

What would show you misread it: (1) if nanoui gets no new breakout apps after November and weekly gains fall back to dozens, its stars came entirely from those two tweets, and the playbook depends more on luck than it looks; (2) if mygo and the new GPUI-family libraries all stop growing together in December, the wave has receded and a port has nothing to ride; (3) if someone ships a Swift or C# version and it's still under 200 stars a month later, the communities left after Go, Rust and Zig aren't buying it.

Data and sources: GitHub repo pages and daily star counts (10-20 through 11-06, including mygo, nanolight, rune, nsh and others); the nanoui README (Credits and Performance sections); X posts: @rachpradhan 10-18 "introducing nanoui", 10-31 iOS, 11-02 "introducing nanolight", 11-03 nanolight 0.2.0, @rammcodes 11-03, @QingQ77 11-05, @chirorojr 11-01; gihyo.jp 2026-10-21 coverage; trendshift.io/repositories/311266; Hacker News search (no nanoui posts).

GitHub repo ↗

Research notes: competitor table, acquisition evidence, full research report
Competitors
NameWhereLaunchedScaleHow it differs from this one
egoist/mygoOpen source2026-09-264,797★, +1,119 over 7 daysPrototype, same architecture; written in Go, the program is 11.8MB; nanoui claims about 5× faster per frame and a program about 17× smaller
Zed GPUI and its ecosystem (gpui-component, longbridge/nsh, etc.)Open source2021; the ecosystem took off in autumn 2026zed 92.8K★; nsh 1,486★ (+613 in 7 days)Rust, more mature and heavier; nanoui borrowed its API style
anomalyco/runeOpen source2026-10-22562★, +247 over 7 daysRust, a native UI framework from the opencode team
ElectrobunOpen source2024-0213,026★TS frontend plus webview, doesn't draw its own UI
dvui / capyOpen source2022 / 20211,739★ / 2,713★Established Zig GUI; no Metal or Liquid Glass visuals, no companion apps, no mobile
Tauri / Wails / ElectronOpen source2019 / 2018 / 2013112K / 37K / 124K★Webview route, large binaries; the thing nanoui is pitching itself against
egui / iced / Slint / Dear ImGuiOpen source2014–202024K–77K★A mature GUI library, but not in Zig
Where users come from
  • Verified by the author's demo video: 10-18 launch post, 64,159 impressions, 1,580 likes, 1,175 bookmarks; 0→285★ in two days
  • Verified cross-promo from own app (nanolight): 11-02 nanolight promo drove 113,236 impressions; nanolight 4 days, 371★; during the same period nanoui daily installs rose from +28 to +105
  • Verified Japanese tech media: gihyo.jp coverage 2026-10-21; +78★ that day
  • Verified Chinese and English repost accounts: @rammcodes 11-03 (also FB Claude community groups and Instagram); @QingQ77 11-05 and @xiaoanlab 11-03 introduced nanolight
  • Verified riding the original's and the category's heat: the README uses mygo as the benchmark baseline; the author keeps replying "use mygo or nanoui"; the community has "I forgot the Zig one"
  • Inference HN / GitHub trending: no related posts in the HN index; the repo is under 60 days old, so it can't make trending

justrach/nanoui: why it got off the ground

A UI framework for building desktop (and mobile) apps in Zig: either draw native UI on the GPU in Zig (no HTML/JS/webview), or wrap the system webview for the frontend. The author is Rach Pradhan, an indie developer and NUS graduate from Singapore (GitHub justrach, X @rachpradhan). Research date 2026-11-06.

Bottom line

nanoui's stars didn't come from people broadly using the framework. Two things drove them: first, the author's X demo videos with very striking numbers (0.06ms per frame, a 0.7MB program, 0% CPU when idle); second, nanolight (a ⌘Space launcher), an app he built with nanoui for regular users, which broke out and brought attention back to the framework. The backdrop is that in autumn 2026, "no Electron, draw the GPU UI yourself" was in vogue: Zed's GPUI ecosystem, and egoist's Go framework mygo (nanoui is a port of mygo to Zig). nanoui took the open "Zig version" slot in that wave.

1. Data (GitHub repo, 2026-11-06)

  • 2026-10-18,19 days after creation: 830 stars, 39 forks, already at v0.11.0 (11-03). Gained 347 over the last 7 days, but only 64 in the first 7 days. 11-05 was the single biggest day, at +105.
  • The curve has two waves (GitHub repo_day):
    • First wave, 10-18 to 10-21: 285 in two days (10-20), 10-21 to 363. Afterward it fell back to around +5 to +15,10-31 per day, which is 484.
    • Second wave, starting 11-01: 510 → 538 → 616 → 702 → 807 → 830 (11-01 to 11-06).
  • The same author's other repos: nanolight created 11-02, 11-03 is 193,11-04 is 319,11-06 is 371. nanomon created 97, nanoterm created 43, both barely moving. nanobrew (1148) and codedb (1401) are older repos and aren't growing much.
  • Prototype egoist/mygo (Go, created 2026-09-26): 4797 stars, +1119 in the last 7 days, and still accelerating. nanoui's description is almost a copy of mygo's "Develop desktop apps with native UI or web frontend in Go."

2. Where users come from (growth levers)

TimelineEventEvidenceMatching stars
10-18 18:28 UTCThe author posted "introducing nanoui" with a nanomon video; every frame was rendered by nanoui with no window openX API: 64,159 impressions, 1,580 likes, 1,175 bookmarks, 90 reposts (the author himself has only 3,255 followers)0→285 in two days
10-19The author posted to introduce nanoterm4,075 impressionsSmall
10-21Japanese tech outlet gihyo.jp published a write-up, "Zig製GUIツールキット「nanoui」"gihyo.jp/article/2026/10/nanoui; also covered by note.com10-21 that day +78
10-31「nanoui is coming to iOS,440 KB」16,912 impressions, 340 likesSmall
11-02 05:05"introducing nanolight": a ⌘Space launcher built with nanoui, 1.6MB113,236 impressions, 1,359 likes, 1,255 bookmarks, the author's highest-reach postnanolight 0→193;nanoui 11-02 +28,11-03 +78
11-03nanolight 0.2.0 released for Mac/Win/Linux, and it can run Raycast extensions13,974 impressions
11-03 21:16Repost account Ram Maheshwari (@rammcodes, 6.2K followers on X, claims 500k+ across platforms) posted "WAIT WHAT?? … without Electron"1,676 X impressions; the same kind of reposts also appeared in the Facebook "Claude Community" group and on Instagram11-04 +86
11-04@Yappologistic posted a joke tweet, "an alternative to nanolight"9,190 impressions, 267 likes
11-05Chinese account @QingQ77 (29k followers) introduced nanolight; earlier, @xiaoanlab also published a Chinese write-up 11-031,085 impressions11-05 +105
OngoingThe author replied under others' discussions with "try nanoui!" and "or they can use mygo or nanoui!", and also asked @grok to read the repo and answer technical questions for themSearch results contain many replies from @groklong tail

HN: the hackernews database has no posts for nanoui or nanolight (only an unrelated namesake, dexterlagan/nanoui, with 2 points). Not on HN. GitHub trending: the repo is under 60 days old, so it can't chart yet; trendshift.io indexes it. No paid promotion found.

Conclusion: nearly all the traffic came from the author's own X account: few followers, but the demo videos had an extremely high bookmark rate (1,175 bookmarks / 64K impressions). It then spread through a Japanese outlet, Chinese repost accounts, and an English repost account. The second wave was driven by nanolight: an app ordinary people can understand and use (a Spotlight / Raycast alternative), which spreads far better than "yet another UI framework". It brought a group of people who don't write Zig to nanoui.

III. Competitors

NameWhereTimelineScale (github repos, 11-06)How it differs from nanoui
egoist/mygoOpen-source Go2026-09-264,797★, +1,119 over 7 daysThe prototype for nanoui, same architecture; mygo is in Go, with a 11.8MB binary. nanoui's README uses mygo as the benchmark, claiming 5x faster per frame and a 17x smaller binary.
Zed's GPUI and its ecosystem (gpui-component, gpui-fast, disktree, nsh, etc.)Open source RustGPUI is 2021 years old; the ecosystem took off 2026 years ago in the fallzed 92.8K★; longbridge/nsh 1,486★ (7 days +613); tobi/disktree 4,004★Rust, heavier, more mature; nanoui copied its chained API and virtual list
anomalyco/runeOpen-source Rust (the opencode team)2026-10-22562★, +247 over 7 daysPart of the same wave of "native UI frameworks", backed by a team
ElectrobunOpen-source Zig/TS2024-0213,026★Writes the frontend in TS with a webview, rather than drawing its own UI
dvui / capyOpen-source Zig2022 / 20211,739★ / 2,713★A long-standing GUI in Zig; no polished Metal plus Liquid Glass demo, and no companion apps
Tauri / Wails / ElectronOpen source2013–2019112K / 37K / 124K★Takes the webview route, the very thing nanoui says it replaces
egui / iced / Slint / Dear ImGui / ClayOpen sourceLong-established18K–77K★Immediate-mode or declarative GUI, but not in Zig

In Zig, the combination of "GPU-drawn native UI + choice of webview + five platforms" had no sizable project before. nanoui is itself a follower of mygo, just in a different language.

4. Why it made the list

  1. Timing: native GPU UI is trending. From 2026 through November 9, the GPUI ecosystem and mygo took off together (on GitHub, mygo grew to 4.8K in a month, +1,119 in the last 7 days; new GPUI-based repos appeared in batches). The community kept asking "what was that Zig one called again?" (@chirorojr 11-01: "GPUI / MyGo / Rune / I forgot the Zig one"). nanoui is that Zig one.
  2. The demo video makes hard numbers the selling point. 0.06ms per frame, 0.7MB, 0% CPU at idle, zero dependencies, plus a Liquid Glass effect. That hits the feeling of being fed up with Electron eating memory (one reply: "maybe we don't need more RAM, we need more low level softwares"). The high bookmark count shows people saved it as something to come back to.
  3. Uses its own apps to prove the framework, so attention flows between apps and framework. In 19 days the author shipped nanomon, nanoterm, nanostudio, nanofly and nanolight, plus a third-party nanomail. Each was a new "introducing" post; nanolight broke out and drove the second wave.
  4. Fast release cadence makes it look capable. It reached v0.11 in 19 days, adding iOS, Android, Linux and Windows in turn, and each addition got its own tweet.
  5. The author had a track record: earlier projects nanobrew (225 HN points), codedb (1.4K★) and turboAPI (1K★). These gave him a base of followers interested in Zig and performance, the first reposters at cold start.
  6. Friendly to coding agents (a differentiator, but not the main driver of growth): the nui command has --json on every subcommand, ships with AGENTS.md, and can take windowless screenshots and run tests without a window. Agents can develop and verify the framework, which probably also explains how the author writes so fast.

5. Differentiation

  • Written in Zig, under 1MB, 0% CPU at idle; the same Zig UI code runs on macOS, iOS, Android, Linux and Windows; choose native UI or a webview frontend; testable and screenshot-able with or without a window; Liquid Glass and Motion animations implemented from scratch.
  • Essentially "mygo's architecture + GPUI's API + gpui-motion + GPUI Kit, rewritten in Zig". The README's Credits section is candid about this. What's new is mainly the language and the extreme small size, not a new architecture.
  • The differentiation is one reason for the growth: the "Zig version" is itself a selling point, and "smallest, fastest" is naturally shareable. But the author's posting rhythm and nanolight's breakout are what shaped the curve.

6. How many people have seen it

It spread within small circles. Japanese coverage includes one gihyo.jp article; Chinese coverage includes write-ups by @QingQ77, @xiaoanlab, @xingkaixin and others; English reposts come from @rammcodes and a Claude community group on Facebook. No HN coverage, no major tech media, and the author himself has only 3.3K followers. On X, some people already criticize it as "not well maintained" and argue for using GPUI directly.

7. Copycats

  • At the framework level: nanoui is itself a copycat of mygo. So far I haven't seen anyone copy nanoui.
  • At the app level: nanolight got its own "alternative" (a joke tweet by @Yappologistic, 9K impressions, 11-04). But nanolight itself has been called a copycat of Tinycast, which is in turn a copycat of Raycast.
  • Around the same time, a batch of native UI frameworks that didn't copy it but do the same thing: longbridge/nsh, anomalyco/rune, oneo-me/nsgo (built on mygo).

8. What to borrow

  • "Rewrite a hot framework in another language and benchmark against the original" can borrow the original's heat. The condition is speed: nanoui shipped within 10-18 of mygo's 9-26 release.
  • When you build an open-source framework, build a small app with it that ordinary people can understand (a launcher, an activity monitor, a flight tracker). Each app is a separate release that funnels traffic back to the framework.
  • The demo copy only states 3 to 4 hard numbers (frame time, size, CPU, dependency count), plus one striking visual effect (Liquid Glass).
  • Having @grok answer technical questions in your Twitter replies is free customer support and endorsement.

9. What I'm still unsure about

  • Of the 830 stars, how many people have actually built apps with nanoui? So far I've only found nanomail (gstohl) and a small tool by @meaningmean_.
  • I didn't verify how much traffic @rammcodes' posts on Instagram and Facebook actually brought.
  • I didn't find whether egoist himself reposted or mentioned nanoui.
  • There's no stargazer breakdown, so fake stars can't be fully ruled out. But the fork ratio is 4.7%, the two-wave growth curve looks organic, and it lines up with tweet impressions, so it doesn't look botted.

Sources

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.