rterminal

From open question to cited paper.

A full research method comes loaded: literature survey, hypothesis tournament, experiments, the paper itself. You judge every fork.

Try the early access beta → Sign in

A method you can read, run, and rewrite.

Nine stages ship as plain skill files. Run them as they come, or edit them until the method is yours.

01ScopePin the question and the aim, with you.
02LiteratureSurvey prior work; build the evidence ledger.
03HypothesesDerive falsifiable claims worth testing.
04DataFind and profile the datasets.
05ExperimentDesign first, then run for real.
06AnalysisRule on each hypothesis from the results.
07PaperAssemble the write-up, cited and honest.
08VerifyIntegrity gate: references and verdicts checked.
09DistributeVenue, release, and submission.

Runs for hours. Stops at the forks.

A big run surveys, experiments, and drafts across many turns. It doesn't run off on its own: every judgment call and interpretation comes back to you.

rterminal does the busy work
  • Search, read, and screen hundreds of sources
  • Extract, triangulate, and build the evidence ledger
  • Run experiments and record measured results
  • Draft, cite, and check the write-up
you make the calls
  • Confirm the scope and premises
  • Pick the direction; sign off the design
  • Interpret results; rule on each hypothesis
  • Approve what ships

One real run, end to end.

From a live session: designing a Bayesian decision-support system for sentencing, scoped, reviewed, and decided with the researcher.

literature review284 → 18

284 sources screened down to 18 load-bearing claims, each traced back to the papers that support it. The claims below carried the design:

"A risk score can't be both calibrated within groups and equalized in error rates across them."the fairness-impossibility result, pinned to Kleinberg 2017 and Chouldechova 2017

Every claim in the ledger keeps its sources, so a challenged premise takes you straight to the evidence.

hypothesis tournamentElo-ranked

Candidate designs scored by an automated review panel, each kept or revised with reasons.

  1. 1 Integrated system 1265
  2. 2 Adapt a bail cost-minimization model 1241
  3. 3 Lean minimal-sufficient champion 1231
integrity gateverified

Before anything shipped, the run checked itself: references resolve, and verdicts cover every hypothesis.

pass with warningsFlagged and surfaced to the researcher, not smoothed over.

It becomes your lab's way of working.

Your search sources, your exclusion rules, your venue's formatting: corrections you make become part of the method. And when your field needs a tool that doesn't exist, say a parser for the one archive your discipline relies on, rterminal builds it and keeps it in your kit.

  • ✓ The nine stages are editable skill files, not settings
  • ✓ A fix you make once holds across every later run
  • ✓ Tools it builds for your field stay yours, on every machine

Bring it a live question.

rterminal is in a private early access beta — free while it lasts, no card required. Sign up with your email and the early access token you've been given. At the University of Oxford? Use your ox.ac.uk email on the Oxford path (an early access token is needed there too during the beta).

Try the early access beta → Download the app