Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-07 · high
Caveman handles the output side, and benchmarks posted alongside both tools show that side collapsing to single digits on coding runs.
Why now: This compression split was covered on August 7, 2026.
Headroom compresses what enters an AI agent's context, tool results, logs, RAG chunks, using AST- and schema-aware compressors, per the project's GitHub repo. A separate tool called Caveman compresses the other direction: the narration an agent writes back out, trimmed with prompt-level style constraints.
The split matters because tool-heavy agent loops burn most tokens on what tools return, not on what the model narrates back. Benchmarks posted with the two tools put input-side reductions at 60-95%, per Headroom's repo.
Output-side compression collapses to single digits on agentic coding runs by the same measure. That's the gap between an agent that runs out of context mid-task and one that doesn't. The repo doesn't break out results by compressor type or say which coding runs were tested.
The project's guidance follows that gap. Spend compression effort on the input side for tool-heavy agent loops. Save output-side trims for chat-shaped work, where the model's own text makes up most of what moves back and forth.
Each link below shares sources, entities, or timing with this story.
headroom benchmarked against GSM8K / Shared entities / Same source / Shared topic / Earlier coverage
Linked by a graph relationship (headroom benchmarked against GSM8K); both cover AST, Headroom, RAG; cite the same source (Headroom).
output uses Claude Code / Shared entities / Same source / Shared topic / Earlier coverage
Linked by a graph relationship (output uses Claude Code); both cover Caveman, Headroom, RAG; cite the same source (Headroom).
output uses Claude Code / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (output uses Claude Code); both cover Headroom, RAG, Token; reported by the same outlet (github.com).
Linked by a graph relationship (output uses Claude Code); both cover RAG, Token; reported by the same outlet (github.com).
Linked by a graph relationship (output uses Claude Code); both cover Caveman, Output; reported by the same outlet (github.com).
output uses Claude Code / Shared entity: Caveman / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (output uses Claude Code); both cover Caveman; reported by the same outlet (github.com).
headroom uses LLM / Shared entities / Same source / Shared topic / Earlier coverage
Linked by a graph relationship (headroom uses LLM); both cover AST, Headroom, RAG; cite the same source (Headroom).
output uses Claude Code / Shared entity: Token / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (output uses Claude Code); both cover Token; reported by the same outlet (github.com).