Pulumi Engineer's 10 Coding Agent Tips: Move Instructions Out of Chat Into Files and Hooks
rseroter · x · 2026-09-23
Pulumi engineer Engin Diri published a 14-minute guide of ten coding-agent tips cross-checked against recent research. The core thesis: instructions kept only in chat disappear during context summarization — nearly every fix moves something into a file, hook, test, or fresh session ("the agent forgets, the repo does not").
Key points:
- Write instructions for the agent, not a new hire: concrete and verifiable ("API handlers live in src/api/handlers/"), not "keep files organized";
- Hard guarantees belong in hooks, not prompts;
- Long sessions: summarization eats chat-stated conventions, so pin them to instruction files;
- Parallel agents and subagent costs: recent papers urge caution about blindly adding subagents;
- The model that wrote the code shouldn't grade it — use a separate reviewer.
The post notes each study's measured scope is sometimes narrower than its abstract suggests.
More from coding & agent
- Perplexity's hint-guided self-distillation cuts its agent's tool-call failures by 21.2% — perplexity_ai · 2026-09-23
- Tomo ships WebMCP Registry: open index of agent-ready tools across 100K sites — Jackyhuang · 2026-09-23
- LiteParse hits 2.8ms per PDF page, 25% faster in v2.14.6, claims fastest open-source parser — llama_index · 2026-09-23
- UCLA Releases ACLArena: A Framework for Agent Continual Learning in Multi-Stage Post-Training — UCLA-SCAI · 2026-09-23
- The Hidden Cost of Using Weaker Models for Decisions: You Never Benchmark, So You Never Notice the Lost Alpha — generativist · 2026-09-23
- Coinbase opens stock trading to AI agents, with x402 micropayments for live market data — MurrLincoln · 2026-09-23