Agent Context Research:The Evidence So Far
A guided map of five agent-context studies, including the 84-run replication that did not reproduce the pilot's cheaper-context headline, and the question each follow-up was built to answer.
8 beats · 4 min to read in full
This page is a reading mapfor the research program
This page is a reading map for the research program, not a new experiment. It shows how the question changed as the evidence accumulated: from whether a richer workspace helps, to whether agents consume it, to which information is actually…
Condensed from the note.
The evidencepath
The sequence below follows how the questions developed. Some later studies branch from the same result rather than forming a strict publication chronology.
1
Study How the Environment Affects Agent Performance and Token Cost · Question Does progressively richer project context change success or token cost? · What changed In a small pilot, every environment completed the task and richer conditions sometimes used fewer effective tokens. The full scaffold's input was heavily cached. The paper correctly labels this as a pilot.
2
Study The Incurious Agent · Question Does the cost result repeat, and what context do agents actually read? · What changed Across 84 runs, the cheaper-context headline did not repeat. Both agents largely skipped the deepest curated material; context that went unread could not improve the outcome. This revised the pilot's main cost claim and shifted attention from authoring to consumption.
3
Study The Self-Sufficient Agent · Question If the tasks and repository get harder, does a bare agent finally need the documentation? · What changed On this repository and task family, Codex and Claude solved all nine functional landmines without docs. They failed mainly on arbitrary project-specific formats. The useful surface for context narrowed from general capability to local convention.
4a
Study Relevance, Not Volume · Question Is the amount of context the lever, or whether it contains the needed rule? · What changed A generic contract and a same-length rule-bearing contract produced radically different conformance. On these tasks, useful context was information the agent needed and could not derive, not simply more documentation.
Condensed from the note, which carries 5 rows.
The synthesisevidence map
Across the context studies, four conditions now matter more than document count:
Headroom
the agent must have a gap that information can close. Context cannot improve a task the bare agent already solves at the ceiling.
Relevance
the information must supply something needed and underivable, such as a local wire-format convention.
Consumption
the agent must actually encounter the information. A perfect memory tree that stays unopened has no treatment effect.
Agent behavior
reading habits differ. Results from one agent cannot be assumed to describe every agent.
Condensed from the note.
Choosea reading path
Start with the correction
read the pilot, How the Environment Affects Agent Performance and Token Cost, then The Incurious Agent.
Understand what agents already know
continue to The Self-Sufficient Agent.
Turn the result into documentation practice
read Relevance, Not Volume.
Compare reading behavior across agents
read Curiosity Comparison Between Agents.
Condensed from the note, which carries 5 points.
Boundariesthat still matter
The studies cover a small number of repositories, tasks, agents, and runs. Several results are directional, and the experiments deliberately use different task families to answer different questions.
Condensed from the note.
That is the note,in one pass.
This is Agent Context Research: The Evidence So Far condensed to its beats. The full note carries every paragraph, the tables, the code, and the source it was adapted from.