Basis is built for the hard middle of research: getting unstuck, exploring branches, assembling evidence, and producing an artifact you can keep refining or hand off to your preferred LLM, code agent, editor, or team workflow.
You can continue in Basis to build toward a final report, or use the output as a strong starting point for ChatGPT, Claude, Codex, Claude Code, LaTeX, or your own review process.
Basis is not trying to replace every LLM in your stack. It gives you a structured research loop and a concrete artifact you can inspect, continue, export, or take somewhere else.
Start a targeted run when a technical question deserves evidence, branches, and a reviewable artifact instead of a quick answer.
Use Basis to turn a difficult topic into findings, assumptions, open questions, citations, draft structure, and a next-step plan.
Explore competing hypotheses or implementation paths before committing days of work to one direction.
Give Basis the real research question, the constraints, and what would make an answer useful.
Basis iterates through evidence, assumptions, branches, draft structure, and artifacts.
Review the report, PDF, TeX, logs, and any generated outputs before treating it as settled.
Run another Basis cycle, export the package, or polish the work in your preferred tool.
Basis is strongest when you give it a concrete research goal and let it do the slow middle: branch exploration, evidence gathering, technical drafting, and artifact assembly.
It is still research software. You should review conclusions, check citations, and use your own judgment before publishing, shipping, or making high-stakes decisions.
The goal is not to trap your work inside Basis. The goal is to produce something strong enough that the next step is easier, whether that next step is another Basis cycle or a handoff to the rest of your workflow.
Start with a targeted problem, review the artifact, then decide whether to continue in Basis or carry the output into the tools you already use.