The evidence-based research board
Confidence computed, not guessed.
A shared board where hypotheses, data and code live side by side. Every piece of evidence moves the probability — and the whole team sees what will settle the argument.
Email link sign-in · no password · no card
H1 · the effect is real and above 2 pp
A/B test, 14 days
LR 5 · w 0.9March seasonality
LR 2 · w 0.6Features
One board instead of five tabs
Hypotheses, exports, calculations, discussion and the decision — on one canvas, in real time, for the whole team.
- 01 —
Live evidence map
Hypotheses with a prior, evidence for and against. Confidence recalculates with every new arrow — you can see what will settle the question.
- 02 —
Python right on the board
A notebook in the browser: pandas, scipy, statsmodels, sklearn. Charts and numbers land on the board in one line and refresh on every run.
- 03 —
Low-code nodes
25 nodes — from filter and pivot to gradient boosting, CatBoost and XGBoost. Only the changed chain recomputes; its code goes to the notebook in one click.
- 04 —
AI assistant
A board summary for your manager, new testable hypotheses, a devil's advocate for the selected hypothesis, cell code from a description.
- 05 —
Calls on the board
Voice, video and screen sharing for everyone on the board. The presenter leads — everyone else follows their view.
- 06 —
Telegram bot
New research, milestones, data, calls and finished AI reviews — with buttons. Send the bot a link and it asks what to send, to whom and when.
- 07 —
Everything you need
Sticky notes, frames, LaTeX formulas, checklists, images, YouTube videos, search, templates, SVG/PNG export, iPad and Apple Pencil.
- 08 —
A window into the board
A public link shows only what is marked public: frames with conclusions and hypotheses with their current confidence. Evidence and code stay hidden.
Live map
Confidence is a calculation
Every piece of evidence has a strength (LR 2 / 5 / 20), a quality and a shelf life. Links between hypotheses carry confidence too.
logit(P) = logit(prior) + Σi ±wi·ln LRi
- LR
- evidence strength: 2, 5 or 20
- w
- weight: quality × freshness
- ±
- for or against the hypothesis
Hypothesis ledger
Decisive evidence
A/B test · +2.3 bits
To reach 90% on H3
3.9 bits missing
How it works
From question to decision
-
Ask the question
Create a research board — it opens right away with a template for your method: decision brief, experiment plan, evidence map, retro. Drop in CSV, XLSX, JSON or Parquet.
-
Compute and argue
In the notebook or with nodes — charts land on the board. Evidence for and against, a call right on the board, and AI hunting for holes in the argument.
-
Show the conclusion
A public link to the frames with results, an image export, or a Telegram broadcast — to the right people at the right time.
- 25 low-code nodes — from filter to CatBoost and XGBoost
- 3 boards free, no card required
- 0 passwords: sign in with an email link
- 1 click turns a chain of nodes into code
For people who need not just to compute, but to convincingly justify a decision.
- Product analysts
- Marketing
- Data science
- Research
- Consulting
Works in the browser on desktop and iPad: trackpad, gestures, Apple Pencil.
Security
Team data stays with the team
- RLS
Isolation in the database
PostgreSQL itself refuses to return other teams' rows — even if a filter is missing in code.
- 2FA
Passwordless sign-in
The email link lives for 15 minutes. A second factor and backup codes are optional.
- PY
Code runs with you
Python runs in your browser — data is not sent to third-party compute servers.
- URL
Public only on purpose
Links are off until you create one, and revocation is instant.
Pricing
The workspace pays, not the person
Card or crypto. Pay in advance, no automatic charges.
Your first board in a minute.
Free for up to 3 boards. Teams and public links come with paid plans.