B
Offline
Step 1 of 10

Iterate or Publish

9of 10~5 min
  • The iterate vs publish decision framework
  • How to structure an iteration round
  • Publishing artifacts to your research program
What you'll do
  • Review your session scores
  • Decide: iterate or publish
  • Execute the chosen path

You've scored your session. Now comes the decision: is this artifact ready to publish, or does it need another iteration round?

The Decision Framework

Use your scores to guide the decision:

✓Publish When...

  • • Average score ≥ 4.0
  • • No dimension below 3.0
  • • Hypotheses are mechanistic
  • • Tests can discriminate
  • • You can execute on the plan

↻Iterate When...

  • • Any dimension ≤ 2.0
  • • Missing third alternative
  • • Tests don't discriminate
  • • Key assumptions untested
  • • Framing wasn't challenged

APath A: Iterate Another Round

Send a targeted intervention based on the lowest-scoring dimension:

Terminal
# Example: Low score on "Third Alternative"
brenner send \
  --thread-id "$SESSION_ID" \
  --from Orchestrator \
  --to BlueLake,PurpleMountain,GreenValley \
  --subject "ROUND 2: Third Alternative Focus" \
  --body "
The current H1 and H2 both assume X drives Y.
CHALLENGE: What if neither is correct?
- What mechanism would explain the data without X?
- Are we missing an obvious alternative?
- What would Brenner call our 'blind spot' here?
Please submit revised DELTAs focusing on this gap.
"
# Monitor the new round
brenner session status --thread-id "$SESSION_ID" --watch

Iteration tip: Target your intervention to the specific weakness. A broad "try again" produces worse results than a focused "challenge assumption X."

BPath B: Publish the Artifact

When the artifact is ready, publish it to your research program:

Terminal
# Validate the artifact one more time
brenner lint sessions/$SESSION_ID/artifact.json
# Publish to your research program
brenner publish \
  --artifact sessions/$SESSION_ID/artifact.json \
  --program "cell-fate-research" \
  --tags "development,neural,2026"
# View the published artifact
brenner program show cell-fate-research --latest

What publishing does:

  • • Adds artifact to your research program archive
  • • Creates a permanent record with session metadata
  • • Enables cross-session search and reference
  • • Tracks hypothesis evolution over time

Knowing When to Stop

Diminishing returns are real. Stop iterating when:

Score plateau:Scores aren't improving between rounds
Circular debate:Agents are repeating the same arguments
Good enough:The artifact is actionable even if imperfect
Pro Tip
Keep iteration history. When you finally publish, the evolution of hypotheses across rounds often reveals insights that the final artifact alone doesn't capture.

Decision Made? Whether you iterated or published, you're ready for the final step: teardown and review.