AxiomCode argues coding agents should query compiler-grounded code graphs, not grep
GlitteringMenu7134 · reddit · 2026-10-06
The author observes that coding agents still navigate large unfamiliar repos via a loop of search → open file → grep → follow references, wastefully reconstructing deterministic relationships: calls, inheritance, implementations, dependencies, symbol resolution.
Their team's alternative in AxiomCode: build a code knowledge graph grounded in compiler/type information, and let the agent query it before deciding which source it actually needs. The core thesis: deterministic relationships should be resolved before the model gets involved, so the LLM spends its tokens reasoning over results rather than rebuilding call graphs.
Open question posed: how much codebase exploration should the LLM do at all, and where is the line between grep/RAG/semantic search and deterministic code intelligence? Open-sourced at github.com/AxiomCodeAI/axiomcodegraph.
More from coding & agent
- Garry Tan ports Doom to Paul Graham's Bel LISP in 20 minutes using Opus 5.5 — garrytan · 2026-10-07
- L0pht Hacker Chris Wysopal on Securing AI-Written Code: 'Make It Secure' Isn't a Prompt — WeldPond · 2026-10-07
- Designing evals for contract review agents: ten lawyers, ten redlines — graceisford · 2026-10-07
- Engineer reviews 20 agent-written PRs a day: 'I didn't sign up to be a full-time proofreader' — Specialist_Agent3599 · 2026-10-07
- Researcher unveils RSI paradigm: generic disposable agents plus an evolving knowledge base — yisongyue · 2026-10-07
- KNOWS Benchmark Jointly Evaluates Agents on Search, Tools and Visual Understanding — anmarasovic · 2026-10-07