Gergely Orosz · 8 min read · Gen AI Tools
Context engineering with Dex Horthy
Read the original on newsletter.pragmaticengineer.com →
This Pragmatic Engineer episode with Dex Horthy dives into context engineering and related methods (harness engineering, loop engineering, software factories) for AI-assisted software development, focusing on how to manage LLM context limits without sacrificing code quality. Dex shares lessons from conversations with ~100 AI engineers, including why shipping unread code can cause production disasters, how current coding models can degrade codebases, and how to identify the 'dumb zone' where context limits bite.
Key Takeaways
- Dex's 12-Factor Agents principles for building reliable LLM applications, based on conversations with ~100 AI engineers; he moved from frameworks like LangChain and CrewAI to building his own pipelines.
- Shipping unread code leads to disaster; a July 2025 experiment where the model wrote code without human review caused production issues and complex re-onboarding, underscoring the need for human oversight.
- Today's coding models can degrade codebases because they’re optimized for SWE benchmarks rather than long-term architecture quality; unit tests may not capture architectural weaknesses, necessitating better evaluation approaches.
- Context engineering 101: identify where the 'dumb zone' begins to effectively manage LLM context limits and guide engineering decisions.
Excerpt
Stream the latest episode Listen and watch now on YouTube , Spotify , and Apple . See the episode transcript at the top of this page, and timestamps for the episode at the bottom. Brought to You by • Antithesis — with Antithesis, you can use AI agents to work on critical systems without worrying about correctness. Going far beyond code review, you can run your complete system in a hostile environment, analyze its behavior, and reproduce every issue perfectly. Teams like Jane Street, Fly.io , and the etcd community use Antithesis to ship better code, faster. Learn more. •…