Actual internal beta run

Deep Analysis v0.5

A public-safe record of one run of a six-stage workflow, its recorded evaluations, and a human override.

Evidence status Capture-derived n=1 · Aug 7, 2026

Verified from captures

The authored method

WorkflowDeep Analysis
Version0.5
Stages6
Included files16
  1. Validate Inputs
  2. Research
  3. Build Thesis
  4. Challenge
  5. Risk Review
  6. Human Decision Human-owned

Recorded evidence

What the run exposed

Challenge was the highest point in the displayed token view. Provenance coverage and thesis completeness/falsifiability scored lower than cutoff compliance and challenge quality.

Average tokens270KPer observed stage sample
Average duration7mPer observed stage sample
Average score3.2Average workflow evaluation score
Stage success100%Resolved stage outcomes
Highest displayed token stageChallenge
Lower evaluationProvenance coverage
Lower evaluationThesis completeness & falsifiability
Human override1 of 1 review decisions

Stage success means all displayed resolved stage outcomes completed. It does not mean the analysis was correct.

True override means the human-selected outcome or next route differed from the captured recommendation or the workflow's authored route.

Evidence boundary

What this does not prove

  • Whether the analytical conclusion was correct.
  • Which outcome or route the human changed.
  • The decision outcome, rationale, or exact route taken.
  • A later workflow revision or improved subsequent run.
  • Savings, domain correctness, customer outcomes, or a benchmark.
Human control Approval and workflow revision stay human-owned.

Slipform does not approve its own work or revise a workflow autonomously.

Provenance

Actual captures. Editorial presentation.

The homepage presentation uses fields visible in five actual current-beta captures. Metric meanings were checked against the app implementation. Internal workspace, subject, run-title, local-path, and private prompt details were removed.