Open Source Radar: August 10 — beyond the coding agent
Coding agents owned the trending board for two days straight. Today's signal is different: the hottest open-source repos are specialized — a legal-agent benchmark from Harvey, a 61.6k-star stock-analysis system, and ComfyUI still shipping weekly at the top of the image-gen stack.
ComfyUI (Python, ~126k stars) — The most-starred AI repo on GitHub, and it isn't coasting: v0.31.0 landed on August 8, the latest stop in a weekly release train whose newest commits add Qwen-Image 3.0 generation and editing nodes. It's the graph-based front end, API, and backend most of the diffusion-image ecosystem builds on — you wire nodes into a pipeline, run it in the GUI or headless, and reuse the workflow as a product backend. If you do serious image or video generation, this is the backbone to build on, not the competition.
daily_stock_analysis (Python, ~61.6k stars) — An LLM-powered watchlist analyst covering A-shares, Hong Kong, US, Japan, Korea, and Taiwan markets, created in January and already claiming Trendshift's #1 Python repo of the day. It pulls multi-source quotes and news, scores each stock, and pushes a decision dashboard — buy/sell levels, risk alerts, catalysts — to WeCom, Feishu, Telegram, Discord, Slack, or email on zero-cost scheduled runs. An agent-style Q&A mode runs your follow-ups through fifteen built-in strategies, from moving averages to trend and event analysis. It's a free research desk for retail traders; the repo itself reminds you it's analysis, not investment advice.
Harvey LAB (Python, ~950 stars) — Harvey, the legal AI company, open-sourced the benchmark it uses to judge whether agents can do real legal work: 1,671 tasks across 24 practice areas, each with instructions, real documents, and scoring rubrics, plus a harness that runs and grades agents in realistic settings like M&A due diligence. It's early — the star count is small — but it's the first credible open yardstick in a vertical where agents could move serious money. If you're building or buying a legal agent, this is the evaluation to run it against.
Worth watching this week: domain benchmarks are the pattern to track — whoever publishes the eval for your industry gets to set the bar for everyone else.
Which of these would you actually run this week — the trading desk, the legal grader, or the image pipeline? Tell us in the comments.
Sources: ComfyUI (GitHub) · daily_stock_analysis (GitHub) · Harvey LAB (GitHub) · Harvey's Legal Agent Benchmark post
