Latent Conjecture Experimentation Lab
The method of science is the method of bold conjectures and ingenious and severe attempts to refute them.
9 experiments, 3 you can run right now.
| Project | Category | Status | Date |
|---|---|---|---|
| Lottery Randomness AuditorA reproducible statistical audit of published UK Lotto winning numbers, testing for detectable deviations from fairness era by era.On GitHub | Data Science | On GitHub | |
| Semantic Context LabWhat does a knowledge graph, and an ontology on top of it, add to an AI answering questions over company data? Three approaches, 300 questions, one synthetic company.Live | Data & Graphs | Live | |
| Treetop BenchmarkOne brief, one shot, one file. Frontier models each build an interactive 3D treetop village in a single HTML page, and a viewer puts them side by side.Live | LLM Evaluation | Live | |
| TabulaAn arena where two language models play backgammon against each other, explaining every move, with Glicko-2 ratings and per-move performance metrics.On GitHub | LLM Evaluation | On GitHub | |
| slop-chipsAn agent skill that finds the wording that marks text as AI-written, then rewrites it into clear, human English using a style guide drawn from real books.On GitHub | Agent Skills | On GitHub | |
| Sam Altman: Warnings vs PromisesFour years of Sam Altman’s public statements scored on two axes, how grand the capability claims got and how much airtime the risks got, with every statement on its real date.On GitHub | Data Science | On GitHub | |
| MusterStructured agentic development as Agent Skills. Seven commands take a half-formed idea to a merged pull request, with sub-agents doing the building and a person approving every gate.On GitHub | Agent Skills | On GitHub | |
| BenchPressA hard, deterministic, judge-free benchmark. Models run a small machine for N steps in their heads and report the exact final state.Live | LLM Evaluation | Live | |
| Value-Weighted Retention Targeting Can Concentrate HarmA working paper on how ranking players by treatment effect times lifetime value can steer retention incentives toward the players most at risk of harm, and which safeguards help.Paper | Statistics | Paper | |
| No experiment matches “” yet. Try a broader word, or clear the filters. | |||
- latent/ˈleɪt(ə)nt/adj.
- Present, but not yet visible.
- conjecture/kənˈdʒɛktʃə/n.
- A guess you are willing to test.
This lab is mostly the second, in pursuit of the first.