Basis is deep research on steroids: a starting point for technical investigations that turns a question into evidence, branches, artifacts, and an exportable draft.
Use it at the beginning of a research thread, or when another LLM has run out of depth: start a run, inspect what it found, continue in Basis, or hand the output to your preferred LLM, code agent, editor, or team.
Basis is strongest as the starting point for serious technical work: the problem can still be messy, evidence matters, and you want a concrete artifact you can inspect, continue, or hand off.
Use Basis to compare hypotheses, research directions, engineering approaches, or source interpretations before committing to one answer.
Runs are designed to leave behind sources, drafts, logs, and generated artifacts instead of hiding the research process inside a chat transcript.
Continue in Basis or take the output to another LLM, Codex, Claude Code, LaTeX, or a human editor for final review and polish.
Ask the specific version of the problem and include the constraints that matter.
Use Basis to organize sources, claims, assumptions, and open questions before writing.
Generate a reviewable draft and artifact package instead of a one-shot answer.
Export the work for peer review, coding follow-up, manuscript polish, or continued research.
Basis is not a fake citation machine or a promise that every generated draft is ready to publish without review. You should still inspect the sources, conclusions, and artifacts.
It is not meant to replace your full workflow. It is meant to create a stronger research package for the next step, whether that is another Basis cycle or a handoff to tools you already use.
The product is best when the question is serious enough that a quick answer would be too shallow, but concrete enough that a research run can make measurable progress.
Review the examples, then try Basis as the first serious pass on a technical question that deserves more than a quick answer.