Drew Breunig's State of Context Engineering: 6 tactics, one key tradeoff
rseroter · x · 2026-10-29
Drew Breunig released "The State of Context Engineering," his AGNTCon talk with full slides and transcript; his O'Reilly book The Context Engineering Handbook is forthcoming.
Core thesis: the job is knowing when to be deterministic vs probabilistic. Conveying clear, consistent, noise-free instructions stays hard no matter how good models get—but models aren't people, and their unique construction creates unique failure modes.
Six tactics from the slides:
- Context Augmentation: add relevant info to assist the LLM
- Context Quarantine: isolate subtask contexts in dedicated threads
- Context Summarization: boil down accumulated context into a summary
- Tool Loadout: select only relevant tool/skill definitions
- Context Pruning: remove irrelevant or unneeded information
- Context Offloading: store information outside the LLM's context
The talk traces how context engineering emerged, where it stands, and why each new model capability creates fresh challenges for harness builders.
More from coding & agent
- GitHub open-sources Replica, an AI prototyping tool built on your design system — round · 2026-10-29
- Link to the essay: agentic debugging is a constraint satisfaction problem — goyal__pramod · 2026-10-29
- Agentic debugging framed as a constraint satisfaction problem in viral essay — goyal__pramod · 2026-10-29
- Pydantic webinar: trace coding agents in Logfire to compare cost and quality — samuelcolvin · 2026-10-29
- ChatGPT's dot can now text you via iMessage, with per-channel routing tips — nickbaumann_ · 2026-10-29
- Prevention beats repair: proactive AI saves ~$225K/year vs $120K for fix-it agents — louisdorard · 2026-10-29