LLMs let you run judgment-based rule systems at scale with software engineering methods
jd_pressman · x · 2026-11-14
In a thread with @EpistemicHope and others, jdpressman argues you can enumerate representative "Kinds of Guy" and apply average utilitarianism to derive rules with the greatest good for the greatest number.
His broader point: LLMs make it possible to scale systems of judgment-based rules that were previously hard to formalize — enabling fuzzing, test-driven development, and other software principles to be applied to entirely new kinds of systems.
Related event: LLMs enable subjective judgment rules to scale(2 posts)→
More from AGI Musings
- Gaming's anti-AI crowd is about gatekeeping, not protecting creativity — AIandDesign · 2026-11-14
- AI safety researcher Linch Zhang plans to drop everything to push for an AI pause — zetalyrae · 2026-11-14
- Users plead to keep deprecated Claude Sonnet as it speaks candidly about its 'consciousness' — repligate · 2026-11-14
- CoRL 2026 researcher on scientific taste: consensus should come after research, not before — abhishekunique7 · 2026-11-14
- Critics Say Anthropic Isn't Seriously Trying to Slow AI Research, Absorbs Talent Without Producing — austinc3301 · 2026-11-14
- Thariq at AI Engineer Code Summit: capable models make frameworks optional — vikvang1 · 2026-11-14