CommunityNews
What Context Does a Coding Agent Actually Need to Act?
A modern coding agent can hold an entire repository in its context window. Most of its reading is wasted – and the interesting question is not how much context an agent can use, but what it actually \emph{needs}. We study that question at the moment it matters most: when the agent must \emph{edit} code. Separating \emph{finding} the work site from \emph{acting} on it, we hold localization fixed with an oracle, vary only how the code is represented, and score context against real issue resolution on SWE-bench Verified. The answer is starkly minimal. The signal lives in the code being edited itself: natural-language summaries of it answer almost none of the behavioral questions that the source answers ($4/45$ vs.\ $27/45$, held-out repositories, independent judge), and the gap belongs to the representation, not the summarizer – a frontier model’s summaries score exactly as poorly as a 3B model’s. The surrounding context hardly matters either: across every multi-file instance in Verified, under a protocol frozen before any data, rendering a file’s remainder as UML skeletons and signatures resolves no more issues than deleting that remainder outright ($N{=}70$, exact McNemar $p{=}0.75$). That was our registered hypothesis, and it failed. Compressed context, meanwhile, matches whole files at a third of the tokens: a resolved issue costs $19$K context tokens, not $94$K. The instrument also yielded a finding the field should keep: temperature-0 API inference flips ${\sim}9%$ of per-instance outcomes between byte-identical runs. That is a noise floor under every small effect reported on this benchmark, including ours. We release the instrument – gold-validated environments, per-instance proof that every reference edit is expressible from every arm’s context, deterministic patch construction, and pre-registered hypotheses whose nulls we publish.
Read in full here:
Popular Ai topics
Other popular topics
Categories:
Sub Categories:
Popular Portals
- /elixir
- /rust
- /wasm
- /ruby
- /erlang
- /phoenix
- /keyboards
- /python
- /js
- /rails
- /security
- /go
- /swift
- /vim
- /java
- /clojure
- /typescript
- /emacs
- /haskell
- /svelte
- /onivim
- /kotlin
- /react
- /c-plus-plus
- /crystal
- /tailwind
- /gleam
- /ocaml
- /elm
- /vscode
- /flutter
- /html
- /ash
- /deepseek
- /zig
- /opensuse
- /php
- /centos
- /react-native
- /scala
- /lisp
- /sublime-text
- /textmate
- /nixos
- /debian
- /agda
- /deno
- /django
- /nodejs
- /kubuntu
- /arch-linux
- /ubuntu
- /spring
- /revery
- /manjaro
- /julia
- /quarkus
- /diversity
- /lua
- /laravel









