Articles Directory

Type: types/generated-index.md

← Parent

• An Automated Software House as a Second Test of a Theory Builder (article) - Supplement: the automated software house conjecture (2026-09-02 fixed models) as a second test of a theory builder; Naur's program theory against the stated-content condition, software's stronger falsifier and harder claim, four witness conditions
• Bootstrapping an Autonomous Theory Builder with Commonplace (article) - Bootstrap supplement: Commonplace starts as a reflective theory builder in which people perform many operations, moves them to computation one at a time, and measures success by human decisions per verified improvement
• Can a Theory Builder Running on Fixed-Weight LLMs Learn? (article) - Lead article: bets that a fully automated theory builder that learns, a Popperian conjecture-and-criticism cycle over explicit theories, can be built from fixed-weight LLMs applying precise definitions; payoffs, Bitter Lesson, learning test
• Testing Whether a Theory Builder Learns (article) - Testing supplement: show that a theory builder's retained revisions causally improve later capacity; seven tests, task families with seed, matched baseline, evidence interface, and controls, human-decision accounting, compounding, three hypotheses
• What an Automated Reviewer Should Measure (article) - Automated review with two testable claims: reach and economy indicate reader-relative compression progress, and that progress improves relevant task performance
• Which Existing Self-Improving Systems Are Theory Builders (article) - Survey of eighteen self-improving systems placed against the four theory-builder conditions, with reported learning gains judged separately: five builders, one outside, twelve unsettled mostly on criticism; next experiments
• Writing conventions for kb/articles/