Sighttrue
Last reading 2026-08-07 14:41 UTC age unavailable without scripting

An instrument pointed at 388 open-source repositories.

Read every 4 hours. Compared against its own history, never against anything else. Every figure links to its source.

4,651,372,393 weekly downloads across the 138 npm and PyPI packages these repositories publish. Measured, not estimated — and the only figure here that is a number today rather than in two weeks.
388Repositories watched
5Signals read
4hBetween readings
76Findings on record
14:41 Last reading, UTC
Check your stack Paste a manifest. Nothing leaves the browser. Compare two projects Every axis at once. Wire it to your agent MCP endpoint. Read-only, no key.

Attention vs use

Stars and installs answer different questions, and this is where they disagree. No single source can produce this reading.

Used far more than watched
Repository Stars Installs, weekly Per star
nodejs/undici 7,655 159,875,323 20,885
encode/httpx 15,394 205,805,017 13,369
numpy/numpy 32,520 290,084,978 8,920
sqlalchemy/sqlalchemy 12,050 100,023,478 8,301
scipy/scipy 14,891 102,053,336 6,853
evanw/esbuild 40,008 258,493,301 6,461
sgl-project/sglang 31,483 186,929,840 5,937
eslint/eslint 27,449 157,042,850 5,721
Watched far more than used, through this registry
Repository Stars Installs, weekly Per star
pulumi/pulumi 25,537 14,383 0.6
directus/directus 37,221 21,842 0.6
denoland/deno 108,181 129,557 1.2
BurntSushi/ripgrep 67,086 91,314 1.4
NVIDIA/TensorRT 13,233 42,525 3.2
babel/babel 43,972 158,187 3.6
flairNLP/flair 14,381 52,039 3.6
hapijs/hapi 14,794 76,736 5.2

Stars measure attention, installs measure use, and they are different questions. Median across the 120 repositories where both figures are available: 297.58 installs per star. A project distributed mainly as a binary will sit low here and that is a fact about the channel, not about the project — repositories under 10,000 weekly installs are excluded for exactly that reason. How

Installs

Stars can be bought. Installs cannot.

Installs — 247 packages read
# Repository Package Downloads Against the largest Window
1 numpy/numpy PyPI numpy 290,084,978 per week
2 microsoft/TypeScript npm typescript 265,177,341 per week
3 evanw/esbuild npm esbuild 258,493,301 per week
4 encode/httpx PyPI httpx 205,805,017 per week
5 tokio-rs/tokio crates.io tokio 201,312,297 per 90 days
6 pandas-dev/pandas PyPI pandas 194,022,417 per week
7 sgl-project/sglang PyPI sglang 186,929,840 per week
8 BerriAI/litellm PyPI litellm 175,202,199 per week
9 vitejs/vite npm vite 164,813,033 per week
10 nodejs/undici npm undici 159,875,323 per week
11 eslint/eslint npm eslint 157,042,850 per week
12 pnpm/pnpm npm pnpm 153,736,752 per week
13 fastapi/fastapi PyPI fastapi 144,201,831 per week
14 prettier/prettier npm prettier 130,470,015 per week
15 expressjs/express npm express 130,006,647 per week
16 rollup/rollup npm rollup 122,715,083 per week
17 tailwindlabs/tailwindcss npm tailwindcss 120,655,117 per week
18 tokio-rs/axum crates.io axum 104,863,648 per 90 days
19 scipy/scipy PyPI scipy 102,053,336 per week
20 sqlalchemy/sqlalchemy PyPI sqlalchemy 100,023,478 per week
21 vitest-dev/vitest npm vitest 90,330,358 per week
22 microsoft/playwright npm playwright 79,753,878 per week
23 astral-sh/ruff PyPI ruff 75,721,537 per week
24 openai/tiktoken PyPI tiktoken 69,466,965 per week

Bars against 290,084,978. Windows never added together. How

Health

OpenSSF and OSV. Not judged here.

6.1Median scorecard, of 10
289Repositories scanned
99Never scanned
2,698Advisories on record
Lowest scores among the 289 scanned
Repository Scorecard Advisories Scored
iden3/circom 2.5 — 2026-07-27
parcel-bundler/parcel 2.6 0 2026-07-27
meta-llama/llama 3.0 — 2026-07-27
emacs-mirror/emacs 3.2 — 2026-07-27
jesseduffield/lazydocker 3.2 — 2026-07-27
ArweaveTeam/arweave 3.4 — 2026-07-27
iden3/snarkjs 3.4 1 2026-07-27
yarnpkg/berry 3.4 — 2026-07-27
makerdao/dss 3.5 — 2026-07-27
beego/beego 3.6 — 2026-07-27
dandavison/delta 3.6 — 2026-07-27
facebookresearch/segment-anything 3.6 — 2026-07-27

Scorecards OpenSSF, advisories OSV — neither judged here. 99 never scanned, and absent rather than zero. How

Ask

Answered only from the readings here.

Watchlist

Curated and partial. A category is why a repository was added, not a fact about it.

388 repositories, first to last by name All baselines still forming — outlines, not readings
Confirmed spike Detected once Baseline forming
Coverage — 5 categories, 388 repositories
Category Watched Measured Forks added Findings Busiest
ai-ml 74 — — 93 —
crypto-web3 75 — — 54 —
database 79 — — 83 —
devtool 80 — — 93 —
web-framework 80 — — 86 —
All 388 — 409 —

Forks added covers measured repositories only. No figure means not yet measured, never zero.

Watchlist — 388 repositories, busiest 40 shown
RepositoryCategoryLanguage ForksStars AddedReading
twbs/bootstrap web-framework MDX 78,684 174,565 — forming
tensorflow/tensorflow ai-ml C++ 75,910 196,895 — forming
ant-design/ant-design web-framework TypeScript 54,693 98,998 — forming
react/react web-framework JavaScript 51,201 247,092 — forming
kubernetes/kubernetes devtool Go 43,801 124,342 — forming
spring-projects/spring-boot web-framework Java 42,117 81,280 — forming
microsoft/vscode devtool TypeScript 41,472 188,469 — forming
bitcoin/bitcoin crypto-web3 C++ 39,259 89,812 — forming
spring-projects/spring-framework web-framework Java 38,802 60,170 — forming
nodejs/node web-framework JavaScript 36,378 118,821 — forming
huggingface/transformers ai-ml Python 34,140 163,437 — forming
django/django web-framework Python 34,120 88,389 — forming
mui/material-ui web-framework JavaScript 32,546 98,701 — forming
vercel/next.js web-framework JavaScript 31,704 141,674 — forming
AUTOMATIC1111/stable-diffusion-webui ai-ml Python 30,559 164,434 — forming
apache/spark database Scala 29,315 43,814 — forming
pytorch/pytorch ai-ml Python 28,766 102,260 — forming
git/git devtool C 28,223 62,448 — forming
angular/angular web-framework TypeScript 27,418 100,951 — forming
scikit-learn/scikit-learn ai-ml Python 27,268 66,933 — forming
ohmyzsh/ohmyzsh devtool Shell 26,580 189,077 — forming
elastic/elasticsearch database Java 26,055 77,801 — forming
laravel/laravel web-framework Blade 24,904 84,773 — forming
redis/redis database C 24,751 75,916 — forming
expressjs/express web-framework JavaScript 24,599 69,353 — forming
ansible/ansible devtool Python 24,323 70,251 — forming
langchain-ai/langchain ai-ml Python 23,929 143,624 — forming
rails/rails web-framework Ruby 22,356 58,680 — forming
ethereum/go-ethereum crypto-web3 Go 22,094 51,271 — forming
ggml-org/llama.cpp ai-ml C++ 21,419 122,987 — forming
vllm-project/vllm ai-ml Python 20,407 88,435 — forming
jquery/jquery web-framework JavaScript 20,381 59,790 — forming
pandas-dev/pandas ai-ml Python 20,244 49,486 — forming
keras-team/keras ai-ml Python 19,748 64,221 — forming
moby/moby devtool Go 19,167 72,013 — forming
apache/airflow database Python 17,538 46,408 — forming
ollama/ollama ai-ml Go 17,293 177,996 — forming
pallets/flask web-framework Python 16,929 72,171 — forming
apache/kafka database Java 15,406 33,482 — forming
reduxjs/redux web-framework TypeScript 15,212 61,519 — forming

Added counts forks gained across the current observation window. A repository whose window has not filled yet shows no figure rather than a zero.

Today

Since midnight UTC. Empty is the ordinary state.

Today — 33 signals
UTCRepositorySignalConfidenceReadingLink
00:55 Lightning-AI/pytorch-lightning licence confirmed — source
00:55 NVIDIA/NeMo release confirmed v3.0.0 source
02:29 Consensys/teku licence confirmed — source
03:10 grafana/grafana release confirmed v13.1.3 source
03:18 nvidia/nemotron-3-super-120b-a12b model-price confirmed — source
03:18 qwen/qwen3.5-122b-a10b model-price confirmed — source
03:18 qwen/qwen3-vl-235b-a22b-thinking model-price confirmed — source
03:28 z-ai/glm-5.2 model-price confirmed — source
03:28 kubernetes/kubernetes fork-outlier detected — source
03:28 comfyanonymous/ComfyUI fork-outlier confirmed — source
03:28 vllm-project/vllm fork-outlier confirmed — source
03:28 pytorch/pytorch fork-outlier confirmed — source
04:12 ApeWorX/ape licence confirmed — source
04:12 arkworks-rs/algebra licence confirmed — source
04:49 openai/whisper lineage confirmed — source
04:49 QwenLM/Qwen lineage confirmed — source
04:55 sinatra/sinatra licence confirmed — source
06:32 knex/knex licence confirmed — source
06:32 koajs/koa licence confirmed — source
06:32 ant-design/ant-design release confirmed 6.5.4 source
06:32 diesel-rs/diesel release confirmed v2.3.12 source
07:28 safe-global/safe-smart-account licence confirmed — source
07:28 ggml-org/llama.cpp release confirmed b10299 source
07:28 honojs/hono release confirmed v4.13.1 source
09:14 crytic/echidna licence confirmed — source
10:35 huggingface/accelerate licence confirmed — source
10:35 renovatebot/renovate release confirmed 44.14.5 source
10:35 ultralytics/ultralytics release confirmed v8.4.116 source
11:42 TanStack/router release confirmed release-2026-08-07-1110 source
13:38 astral-sh/ruff release confirmed 0.16.2 source
13:38 TanStack/table release confirmed @tanstack/octane-table@9.0.1 source
14:41 fluxcd/flux2 release confirmed v2.9.4 source
14:41 supabase/supabase release confirmed v1.26.08 source
Baseline forming 388 of 388 repositories have under 14 days of history. Their counts are shown raw; no multiplier is computed for them and none is implied.

Signals

Five detectors. Most need fourteen days of history before they can say anything.

Ships What released a new version? 58 recorded Forks What is being copied faster than it usually is? 10 recorded Demand What are developers asking for in more than one place? 0 recorded — nothing has crossed the bar Dependencies What is being added, dropped, or jumped a major version? 6 recorded Lineage Which models say they were built on which? 2 recorded

Our record

How often this has been right, and whether its bars are reachable.

Our own record 0 confirmed fork findings have been resolved so far, with 3 still inside the 7-day window. Too few to state a rate. It will appear here once there are enough, whatever it turns out to be.
Calibration — 4 detectors, 2 days on record
Detector Compares Bar Highest seen Of bar Measured Crossed Reading
fork-spike multiplier against own baseline 3 — — 0 0 Nothing measured
fork-outlier ratio to category median 8 66 825% 15 15 Reached
demand engagement across repositories 60 — — 0 0 Nothing measured
lineage new descendants this week 10 1 10% 24 0 Never approached

Never approached means nothing reached half the bar — a fault in this instrument, not in the repositories. How

The token

What funds this.

This project is funded by a token on Robinhood Chain. Trading it pays a fee, and most of that fee goes to whoever launched the pool — which is what pays for this to keep running and to stay free to read.

That is the whole of it. Holding it does not unlock anything here, there is nothing to connect a wallet to, and no part of this site is behind it. Every reading, every bundle and every archive is public and always will be.

It has not launched. The plan is to run this in the open first, see which of the five readings people actually find worth sharing, and name it after the answer rather than after a guess. When it does launch, the contract address will appear here and nowhere else.