Hypotheses Get Evidence Summaries
The Hypotheses page now has a real evidence layer. Founders can save evidence against experiments, see generated reports and raw results in one place, and refresh an AI summary for each hypothesis.
This is where hypotheses start to become more than a list of assumptions. A founder can now attach actual evidence to experiments and see it roll up under the hypothesis the experiment is trying to validate.
We also bring generated research reports and experiment results into the same evidence view. That gives the decision area a better memory: not just what we decided, but what signals led us there.
The AI summary is intentionally narrow. It reads the saved evidence and experiment outputs, then gives a concise synthesis without inventing new facts.
- Evidence records: users can add text evidence to an experiment with a name, type, source, date, and evidence body.
- Evidence table: each hypothesis now shows saved evidence, generated research reports, and experiment results as one timeline of validation material.
- AI evidence summary: hypotheses can generate a concise Markdown summary of the available evidence, including pending, ready, and failed states.
- Evidence backend: new Convex tables store user evidence and hypothesis-level evidence summaries so the evidence layer can persist across sessions.
- Hypothesis detail view: evidence now sits next to related experiments and decisions, so validation calls are grounded in visible proof instead of only status labels.
- Experiment detail drawer: experiment details now include generated reports, saved evidence, and result context for the selected experiment.
- Evidence refresh flow: users can manually request a new evidence summary after adding or changing evidence.