Welcome to LoopGain discussions — start here

[fitz2882]

LoopGain replaces max_iterations=N with a stop-and-rollback policy that measures whether your loop has actually converged.

If you're new: the README is the first place to go, and there's a live dashboard demo running on real benchmark data.

Where to put things:

You have	Go here
A crash, a wrong result, a broken adapter	Issues
"What should my error signal be?"	🎯 Error signals
"Why did it stop there?" / usage questions	🙏 Q&A
A feature request or a design argument	💡 Ideas
A loop you wrapped, a trace, a number	🙌 Show and tell

A few things worth knowing up front:

• This is alpha. The API may break before 1.0 — pin the version.
• LoopGain detects convergence, not correctness. It knows when more iterations won't help, not whether the answer is right. It is only as good as the verifier behind your error signal.
• The benchmark numbers (92.8% less spend, ~15× faster) come from loopgain-bench and you can run it yourself. Savings are workload-dependent — 96% on fast-succeeding loops, 78–84% on adversarial ones. Run it on your own loops before quoting a number.

Maintained by one person, so replies may take a day. Every question asked here becomes part of the docs so all questions are genuinely useful.

Thanks for being part of the community!
-- Dave