A critical cryptographic domain-separation flaw, tracked as CVE-2026-114612 with a CVSS score of 9.9, affects LiteLLM proxy versions 1.87.0 through 1.104.0-rc.1, allowing an authenticated low-privileg…
The LFORLA RL Game Agents benchmark scores language models on their ability to write RL training scripts for classic Gymnasium/Atari games, using a fixed prompt, a deterministic judge, four explicit s…
A developer built Momentloom, a local-first open-source app that searches YouTube for candidate videos by topic or keyword, ranks them with an explainable "Hype" score, and automatically cuts source v…
A developer built Siesta, an open-source anti-todo list that runs Gemma 3 1B (q4, ~0.5 GB) entirely in the browser via Transformers.js v4 on WebGPU with a WASM fallback, requiring no account, server, …
A developer recommends mapping a dedicated folder to its own drive letter before running any AI coding agent, so that destructive commands like `rm -rf` stay confined to an isolated workspace rather t…
A developer published a guide explaining that Vite's VITE_-prefixed environment variables are embedded into client-side JavaScript bundles at build time and are therefore publicly readable, with the s…
A developer built LazyRepo, an open-source Python CLI that uses a locally running Qwen2.5-Coder 3B model via Ollama to investigate unfamiliar repositories before contributors start changing code. The …
A developer reverse-engineered Sony's closed-source PS4 Orbis southbridge DMA driver using AI to unlock the DDR3 memory chip on the PS4 Pro, which had been bottlenecked at roughly 10.6 MB/s because th…
A developer benchmarked Pillow, sharp, and sharp's mozjpeg option on three white-background product photos, finding that enabling mozjpeg: true produced files 23–28% smaller at quality 85 with 4:2:0 a…
A developer published a method for sandboxing AI coding agents on Windows by mapping the agent's workspace to its own drive letter using the built-in subst command, rather than relying on Docker or a …
A developer auditing Apple Podcasts' scraper Actor found that eight reviews-only and episodes-only input fields (keyword, minRating, maxRating, sort, minDurationSeconds, explicitFilter, and release-da…
A developer demonstrated a two-agent TDD workflow that enforces test/implementation separation through git rather than prompts: Codex writes tests from a spec alone in its own worktree and branch, the…
A developer built SlothScroll, an open-source tool that uses Google's open-weight Gemma model (gemma-4-26b-a4b-it) to read followed Instagram pages once a day and produce a topic-grouped email digest,…
A developer built SlothSort, an open-source tool that uses Gemma 3 (4B) running locally via Ollama to suggest folders and cleaned-up filenames for a messy directory, then moves files only after the us…
A developer tested DeepSeek's V4.1 Flash 1017 snapshot as the model behind two coding agents, finding it passed a simple file-creation task 3/3 in both Claude Code 2.1.310 (16–27 s per run) and OpenCo…
A developer outlined a lightweight method for sandboxing an AI coding agent on Windows by mapping the agent's working folder to its own drive letter, so the agent sees a clean filesystem root and dest…
Anthropic researchers ran a January 2026 study in which 52 mostly junior engineers learned the unfamiliar Python library Trio, with the AI-assisted group averaging 50% on a post-task assessment versus…
A developer tested the August 2026 revision of the llms.txt proposal, running this site's own /llms.txt through the proposal's roughly twenty-line reference parser (llms_txt/miniparse.py) and finding …
A developer built SlothDigest, an open-source local-first CLI agent that summarizes WhatsApp chat exports into a morning digest using open-weight models served through a local Ollama instance. The too…
A developer described switching to a per-drive-letter sandbox on Windows for Claude Code after the agent deleted half a node_modules folder during a refactor. Each agent session is now mapped to its o…