Exploring Maximally Deterministic AI Usage: Return Offsets, Not Substrings
rickasaurus · x · 2026-08-28
The author proposes principles for maximizing determinism in AI applications: return offsets instead of substrings, pass references around instead of values, and have the AI describe necessary actions rather than executing them directly. These aim to improve controllability and reproducibility.
Related event: Engineering Practices for Using AI with Maximum Determinism(2 posts)→
More from coding & agent
- Open-source Mandate: A unified economic layer for AI agents to manage finances — RichardsonDx · 2026-08-28
- Thesis: Investing in AI Agent token consumption—long inference and low-voltage stacks — RichardsonDx · 2026-08-28
- Google Releases WikiSkill for Persistent Agent Knowledge — google · 2026-08-28
- Paper: What Makes Good Agentic Data? An ACE Lens — Xingshan Zeng · 2026-08-28
- Vibe coding procedural trees in Three.js: 30+ tree types, open sourcing soon — jamestagg · 2026-08-28
- How AI coding tools reshape developer workflows: A one-week recap — vista8 · 2026-08-28