Abstract. Every other page on this site is an exhibit. This page is the claim they add up to, stated once.
1. The claim
Machines have made research ideas nearly free. What they have not made free is the decision about which ideas to believe, and that decision, the gate, is now the scarce and valuable part of any research operation. This site is a working instance of that gate, run in public for a season, with its failures kept and its instrument measured. Working Paper No. 33 made the macro version of this argument about intelligence at large. This site is the same argument, demonstrated on its own author.
2. The record
| Measure | Value |
|---|---|
| Ideas filed / tested to a verdict / survived | 316 / 184 / 2 (1.1%) |
| Judged under this site’s own gates / a vendor’s simulator | 69 / 115 |
| Validator’s measured power at a true Sharpe of 0.8 on 20 years | 4 to 6%, published |
| Practitioner-sourced candidates refuted before a family opened, Aug 2026 | 41, across four hunts |
| Forecasts scored live against outcomes | 620, Brier 0.231 |
| Public claims fact-checked against filings | 134; 32 contradicted |
| Corrections applied in one batch and kept visible | 28, across 21 notes |
| Papers withdrawn entirely | 1, after five theses failed review |
| Language-model agent runs behind the work | 958 at last count; statistics from code, always |
| Published notes, each under 500 words of prose | 41, one frozen submission excepted |
3. One gate, many doors
| Wing | What the gate is applied to |
|---|---|
| Tested Strategies | the published factor zoo: 158 alphas to zero, and everything since |
| Research Notes | claims in circulation: 134 checked, a third contradicted |
| Forecast Wing | this desk’s own forecasts, frozen in advance and scored live |
| Trading Floor | the two survivors, risk-budgeted, on paper by design |
| The papers (No. 38, 39) | prediction markets, and the government’s own forecast |
| No. 37 | the agents that run the desk, incidents included |
| Library | the shelf the critic reads before any verdict stands |
4. Why it matters now
A viral roster of sixty agents will run an investment function somewhere this year, and idea generation is about to be industrialised everywhere. Every desk that does so will discover the same thing this one did: the binding constraint moves from generating to adjudicating, and an adjudicator whose power has never been measured is a decoration. The discipline demonstrated here, thresholds fixed before the test, statistics from code, maker and checker separated, the instrument’s power published, verdicts written down either way, is not this site’s edge. It is the coming cost of being believed.
5. What this does not claim
It does not claim the two survivors are special; they are modest, and their capacity is unmeasured. It does not claim the gate is finished; its power curve was published precisely so its rejections are read as statements about the bar. It claims one thing: every number above is checkable from this site, and that sentence is the product.
- Method
- Counts from the pages they appear on: No. 22 for the census, No. 35 for the fact-check, No. 37 for the agents, RN 5 for the forecasts, RN 8 for the instrument’s power.
- Honesty
- This page contains no new results. It is an index with an argument.
- Related
- Everything on the site; start with No. 22, No. 33 and No. 37.