Independent audits of the open-source projects everyone is starring.
We clone the repo, rerun its benchmarks, read its licences and count its stars, then publish what holds up and what doesn't. Alongside the audits: practical guides to AI tools, software engineering and search.
348 articles and 328 trend notes, in English and Spanish.
Rust Accepted This Safe-Looking Code for 4,001 Days: Reproducing the Borrow-Checker Use-After-Free on Stable 1.99.0, Where the Fix Is 8 Lines and the Test-Suite Bill Is 38 Files
Issue #164250 landed on 2026-10-13: a struct's lifetime bounds can be bypassed by assigning through a covariant field, with no unsafe block anywhere. I bisected the compiler (Godbolt, 124 toolchains) and confirmed the code has been accepted by every release since 1.4.0 - 4,001 days - reproduced the use-after-free on …
Read the articleRecent articles
All articles- Developer Tools
Codex CLI Now Starts a Self-Updating Daemon by Default: 373 MB per Copy, 4.9 GB of Retained Releases, and Zero Mentions in the 226 KB Config Reference
Codex CLI 0.157.0 flipped daemon_auto_start to default-on in a 10-line pull request merged three days before release. Seventeen days and six stable releases later the …
- Open Source
IRCv3 in 2026: 48 Specs, Only 3 Required, 13 Still Drafts — and a Website That Out-Commits the Spec Repo 1,064 to 604
IRCv3 is the consensus standards effort behind modern IRC, and its specification repository contains 604 commits. Its website contains 1,064. Reading the project's own …
- Open Source
Bitwarden's Dual License Move: 823 Commercial Files, 52 Hidden Bundle Files, and the 95 Routes the Open-Source Build Never Ships
Bitwarden told users on October 9, 2026 that store builds would become commercially licensed. The repository licence has not changed since 2022 — a three-commit history …
- AI
Installing All 17 E-Commerce Skill Repos: 1,353 Skills, 2.5 Million Tokens, and the 26 Files That Never Load
A request to install every skill from 17 e-commerce agent repositories and report what each contains turns out to be a load test nobody runs. I cloned all 17 (all real, …
- AI
Fact-Checking the Viral 'Netflix Engineer' Video Repo List: 19 Owners, Zero Netflix, and 228,408 Stars That Mostly Predate Opus 5.5
A widely shared list credits one Netflix engineer with releasing 20 GitHub repositories that turn Claude Opus 5.5 into a video generation machine. I checked all 20 …
- Industry Trends
Cloudflare Acquires Deno: 365 Days of Bug Fixes, 180 for Deploy, and the 13-Commit Repo at the Centre of the Deal
Cloudflare announced on 9 October 2026 that it is acquiring Deno, and the two blog posts do not say the same thing: the Deno post gives the runtime 12 months of monthly …
- Developer Tools
Auditing REA: 15,095 Stars, 133 MCP Tools, and the 209,619-Token Bill for Analysing Eight Files
REA promises to let a coding agent reverse engineer anything, from app behaviour down to native binaries, and 15,095 stars say the pitch landed. I installed rea-agents …
- Developer Tools
Auditing text-to-cad: 17,978 Stars, a Two-Tool MCP Server, and the 822 MB Runtime Behind "CAD Superpowers"
text-to-cad promises CAD superpowers for coding agents, and at 17,978 stars it is the most visible project in the category. I installed cadgen 0.7.15 and measured what …
- Open Source
Auditing OpenReply: 2,544 Stars, Zero Releases, and the Meta Business Verification That Ends the Free Path
OpenReply open-sources the Instagram comment-to-DM workflow that ManyChat charges a subscription for, and its documentation is unusually honest about the cost. I cloned …
- AI
Auditing Strata: 125B on a 12 GB Card, 62 GB/s of RAM Bandwidth, and a 5× Contradiction in the Docs
Strata claims a 125B model on a 12 GB graphics card. I measured the repository, the physics and the documentation, and the number holds — because the constraint is not …
- Open Source
Auditing ECC: 68 Agents, 293 Skills, and the 43 Kiro Copies Frozen Five Releases Ago
ECC is the biggest agent-harness repository of 2026: 272,248 stars, 371 contributors and a 19.5 MB npm artifact. I cloned it at main and verified the headline numbers …
- AI
Auditing the Caveman Skill: 109,102 Stars, a 65% Headline, and the 3% Its Own Eval Publishes
Caveman promises to cut 65% of your agent's tokens. I cloned it at commit 8af1f1b (1,613 files, 19,130,901 bytes), reproduced its committed eval with tiktoken myself and …