Research that compiles.

Research rarely moves in a straight line. The hardest part is often getting unstuck: trying a new angle, checking edge cases, and turning a vague idea into a concrete outline you can improve.

Basis is designed to build early momentum and give the exploration structure. It does not automate the whole process. Direction, judgment, and proof stay in human hands.

Basis explores hypotheses and then helps test them. It can propose angles you had not considered. The value of each angle comes from checking it against sources, calculations, artifacts, and your own judgment.

Runs produce artifacts continuously (logs, TeX, PDFs, analysis reports) so you can verify what changed and why. Basis works to keep its claims reasonable. It cannot guarantee correctness, novelty, or completeness. Some runs end completed with blockers instead of a clean pass, and that outcome appears in the audit trail like any other.

Basis supplies a foundation for your work. It does not replace your knowledge or your insight. Take what it produces and challenge it: correct what is wrong, build on what holds up, and steer the work where you need it to go.

Start specific

Open with a clear problem statement and the constraints that matter. The sharper the framing, the better the run.

Give it room

Expect 6–22 cycles to see meaningful progress. Your budget cap is a hard stop, so exploration never outruns what you agreed to spend.

Extend a promising direction

Use Continue run and give specific guidance about what to push on next.

Abandon a weak one

Change the constraints and explore a new branch. A different framing often finds a different route.

Basis is STEM-centric and tuned for technical writing and LaTeX workflows. It may still produce interesting results outside STEM, though we have not validated that use.

Ambitious work advances one verifiable step at a time.

If something seems wrong or confusing, please reach out. We want Basis to be trustworthy, transparent, and easy to steer.