Observed arrival · 2026-10-31
Rig tests whether coding-agent changes actually help
A free, open-source local app compares coding tasks with and without a change to an agent, then reports the difference with a 95% interval.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Developers testing coding-agent workflow changes
- Worth noticing
- In its token-saving example, 40 pairs show a 4% increase with an interval from −17% to +27%, so Rig credits no saving.
Field notes
The examples make the scoring logic inspectable rather than presenting a single uplift number: one comparison scores pass rate, while another measures tokens per solved task with pass rate as a guard. The page says the examples come from two repository files, and lists Claude Code, Codex, OpenCode, and Gemini among supported agents. Local model runners named include Ollama, LM Studio, llama.cpp, vLLM, and Jan.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue