Benzi – A Code Intillegence/Harness Beating Claude Code and CodeGraph
- Source
- benzi.fly.dev
- Author
- tweedler290
- Date
Context
Benzi gives a coding a searchable map of how a project's code fits together. Its repository describes parsing files into relationships between definitions, calls and data flow before the agent answers questions or edits code. The proposed benefit is finding relevant code through those relationships instead of repeatedly reading files to reconstruct them.
The project's CodeGraph comparison tests six repositories with one run per setup, all using Anthropic's Sonnet 5. Four AI reviewers prompted by the vendor ranked Benzi's standalone agent highest on average. The page also lists a higher cost for it than for bare Claude Code or either tool extension, without defining what the dollar figures are per. These vendor-run, model-judged results suggest a tradeoff between answer depth and cost on those tasks, rather than establishing a general advantage on coding work.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
- agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
- SWE-bench — The standard benchmark for AI coding agents: real GitHub issues from real repositories, scored by whether the agent's patch passes the project's own tests.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
Checking sign-in…
Loading comments…