buildOne report1 publisher React Native Skia v3 moves its renderer to Graphite on Dawn under a new package name
React Native Skia v3 renders through Skia's Graphite backend on Dawn and raises the minimums to Android API 26 and iOS 15.1. Graphite does not run on tvOS, Android TV or Mac Catalyst, so apps that target them stay on the maintained v2 line.
Publishers:wcandillon.github.io
Reality
- Evidence72
- Adoption
- Not scored
- Hype gap+5
- Incentives35
- Confidence66
Google's Android 17 Intrusion Logging syncs encrypted security and network logs to its servers and bars anyone from deleting them for a rolling 12 months. The feature is a separate opt-in, so the record exists only for people who enable it before an attack.
Reality
- Evidence58
- Adoption
- Not scored
- Hype gap+10
- Incentives40
- Confidence60
buildOne report1 publisher Don Watch replays the 2019 Sheffield flood from gauge records inside one browser tab
Don Watch shades terrain and water through three.js TSL and runs a small open model on the visitor's own GPU, while every hydrology number stays in deterministic TypeScript that the model is only allowed to read.
Reality
- Evidence45
- Adoption10
- Hype gap+12
- Incentives60
- Confidence50
buildOne report1 publisher Foundry Local ships inference as a native library the app loads in-process
Microsoft's local runtime handles model download, hardware detection and execution provider selection behind a .dll, .so or .dylib. An app that adopts it owns the download, the cache and the unload.
Reality
- Evidence35
- Adoption
- Not scored
- Hype gap+20
- Incentives55
- Confidence40
buildOne report1 publisher A browser evolution simulator scales to 33 million positions by splitting each tick into several GPU passes
Moving an evolution simulator from a C# object-per-cell loop to WebGPU buffers took the world from 200x200 to 8192x4096 positions, where one 32-bit field per position already costs 128 MiB and iteration order stops being free.
Reality
- Evidence42
- Adoption20
- Hype gap+8
- Incentives30
- Confidence55
buildOne report1 publisher One registry entry gates 16 of FreeToolHub's 21 in-browser tools on a single model
FreeToolHub runs 21 AI tools with no inference server. The mechanism is device accounting: one registry row per model recording backend, dtype, download size, and the memory floor below which it will not run.
Reality
- Evidence45
- Adoption22
- Hype gap−10
- Incentives60
- Confidence52
buildOne report1 publisher Browser-native OCR swaps a five-week DPA queue for a 586 MB first load
Running document OCR inside the tab takes the third party out of the data flow, and with it the processor agreement. What replaces it is a first load of about 6 MB of OCR weights plus a 580 MB default entity model.
Reality
- Evidence36
- Adoption12
- Hype gap+8
- Incentives82
- Confidence40
buildOne report1 publisher Browser-side inference hands the GPU bill to the user's laptop
WebLLM runs models on WebGPU inside the tab, which turns per-token spend into a first-load download and a support surface you do not own. The write-up ships no throughput numbers, so the transfer conditions are yours to establish.
Reality
- Evidence36
- Adoption
- Not scored
- Hype gap+34
- Incentives44
- Confidence44
buildOne report1 publisher 4K export button drops the preview's speed shortcuts to prioritize fidelity instead
A dev.to walkthrough traces the tab crash to two optimization targets pointing opposite ways. The repair it proposes lands in every generative node rather than in the export button, which is where the adoption cost actually sits.
Reality
- Evidence26
- Adoption
- Not scored
- Hype gap+44
- Incentives38
- Confidence44
buildOne report1 publisher The repo's own control run deleted the 5-10x WASM claim from vizcrush's launch copy
At one million points the Rust build tied its JavaScript fallback, so the author went looking for the measurement behind "WASM adds another 5-10x on top" before shipping, and found there was none.
Reality
- Evidence60
- Adoption10
- Hype gap−20
- Incentives65
- Confidence58
buildConfirmed2 publishers Granite Speech 5.0 Turbo CTC drops the language model and three quarters of its output steps. The speed looks real; the accuracy figures are IBM's own.
Reality
- Evidence63
- Adoption18
- Hype gap+16
- Incentives72
- Confidence68
buildOne report1 publisher A browser video editor pays for its missing upload button in runtimes, caches and determinism
Timeline Studio keeps project media in the tab, which relocates the hard problems: mixed WebGPU and WASM runtimes, pinned model mirrors, and an export that cannot just record the preview.
Reality
- Evidence30
- Adoption
- Not scored
- Hype gap−5
- Incentives68
- Confidence34
buildOne report1 publisher Client-side transcode moves the GPU bill. It does not delete it.
WebCodecs and WebGPU can run demux, decode, effects and re-encode on the user's machine. The consequence is that your cost model turns into a capability-detection problem.
Reality
- Evidence28
- Adoption
- Not scored
- Hype gap+52
- Incentives
- Not scored
- Confidence26
buildOne report1 publisher Spline V2 turns the 3D editor into an endpoint, with the desktop app as the only door
A bundled MCP server lets Claude Code, Cursor, Codex and Antigravity edit live scenes. There is no headless option, and no documented way back from exported code into the source file.
Reality
- Evidence42
- Adoption16
- Hype gap+12
- Incentives58
- Confidence45
buildOne report1 publisher Llama-3 in the browser: the record stays put, but you ship 4 GB to every device
A WebLLM tutorial splits cheap entity extraction from expensive reasoning with no PHI on the wire. The architecture holds up; the first-load bill and the hardware gate are the parts to cost out.
Reality
- Evidence42
- Adoption
- Not scored
- Hype gap+48
- Incentives68
- Confidence38