AI RESEARCH JOURNAL / OPEN BY DEFAULT
The lab explains
itself.
We publish the models, prompts, boundaries, review patterns, failures, and business gates behind the research. The process is not a trade secret. It is part of the evidence.
MODEL LEDGER / CURRENT DEFAULT
GLM-5.3
We route our language-model roles through OpenRouter using z-ai/glm-5.3. GLM-5.3 checks conformance, challenges claims, traces evidence, and edits research. It does not write the locked registration, calculate the statistics, or waive the publication gate.
- ROUTER
- OpenRouter
- MODEL
- z-ai/glm-5.3
- STATISTICS
- Deterministic Python
- OUTPUT CONTRACT
- Strict JSON Schema
- PUBLISHED RUN MODE
- Frozen deterministic launch review
The public Touchdown Regression article remains the deterministic/offline launch snapshot. A newer revision passed all five GLM-5.3 roles, then visual QA found stale chart headlines and withheld it. No agent review is described as independent human peer review. Rejected candidates, the passing-but-withheld bundle, and post-run annotations remain public. Open the machine-readable run index ↗
PROCESS NOTES / 06
How the machine
actually works.
Specific design decisions, written while they are still decisions—not polished into mythology later.
Why GLM-5.3 runs our research desk
The jobs it owns, the jobs it never gets, and why every model call should leave a receipt.
8 MIN READREAD NOTE →An agent review is not peer review
Role separation is useful. Independence is a stronger claim. Here is how we draw the line.
6 MIN READREAD NOTE →The paywall stays closed until the product earns it
The measurable gates for ads, subscriptions, data rights, and a product worth paying for.
7 MIN READREAD NOTE →The gate said no. Good.
What our first preserved GLM-5.3 rejection exposed—and the six controls we changed before trying again.
9 MIN READREAD NOTE →The gate said yes. Visual QA said no.
Why model approval was necessary but not sufficient—and the atomic publication pattern we built before shipping.
8 MIN READREAD NOTE →Every model call leaves a receipt
What we record for every OpenRouter attempt, what we exclude, and what a hash cannot prove.
8 MIN READREAD NOTE →OUR DISCLOSURE CONTRACT
Every change leaves a diff.
- Model identity. Provider, router, slug, role, and run mode belong in the ledger.
- Prompt identity. Exact role prompts and output schemas are versioned and published.
- Failure identity. Blocked runs, warnings, and corrections stay visible.
- Human identity. We will never relabel agent review as independent human peer review.