# Drew Breunig > Writing about AI, geo, culture, media, data, and the ways they interact. Drew Breunig is the CEO and cofounder of Cmpnd, editor of The Context Engineering Handbook for O'Reilly Media, and an advisor to the Overture Maps Foundation. Every post on this site is also published as plain Markdown at the same URL with a `.md` extension instead of `.html`. Prefer those when reading programmatically. - [About Drew Breunig](https://www.dbreunig.com/bio.html): Bio, current work, and links to profiles. - [All writing](https://www.dbreunig.com/writing.html): Complete chronological index of essays. - [Topics](https://www.dbreunig.com/tags.html): Every tag on the site, each linking to the posts under it. - [Atom feed](https://www.dbreunig.com/feed.xml): The 20 most recent posts with full content. - [Contact](https://www.dbreunig.com/contact.html) ## Featured Writing - [The Problem is Prompt Debt](https://www.dbreunig.com/2026/06/22/the-problem-is-prompt-debt.md): The plain-English prompt that makes prototypes effortless is a poor way to specify a system, and the bill arrives slowly, disguised as progress. (2026-06-22) - [How Long Contexts Fail](https://www.dbreunig.com/2025/06/22/how-contexts-fail-and-how-to-fix-them.md): Taking care of your context is the key to building successful agents. Just because there's a 1 million token context window doesn't mean you should fill it. (2025-06-22) - [The 3 AI Use Cases: Gods, Interns, and Cogs](https://www.dbreunig.com/2024/10/18/the-3-ai-use-cases-gods-interns-and-cogs.md): Simplifying and navigating the AI noise by segmenting everything into 3 big use cases. (2024-10-18) - [Towards Standardizing Place](https://www.dbreunig.com/2024/07/31/towards-standardizing-place.md): Overture's Global Entity Reference System could standardize place, making geospatial intelligence a simple data join instead of a mapping project. (2024-07-31) - [The Business Implications of Machine Learning](https://www.dbreunig.com/2016/06/23/the-business-implications-of-machine-learning.md): Pay attention to machine learning not for what it can do, but because Google, Facebook, and Amazon prioritizing it will reshape the industry. (2016-06-23) ## Recent Writing - [What We Can Learn from Claude's Fable 5.1 System Prompt](https://www.dbreunig.com/2026/09/07/what-we-can-learn-from-claude-s-fable-5-1-system-prompt.md): Looking at how system prompts change reveals that models and products are a moving target that prompts must accommodate and govern. (2026-09-07) - [Who Taught the Models to Do That?](https://www.dbreunig.com/2026/08/30/who-taught-the-models-to-do-that.md): The Hugging Face attack is surprising, but the capabilities behind it were deliberately cultivated. (2026-08-30) - [Fable & The End of the Free Lunch](https://www.dbreunig.com/2026/08/23/fable-the-end-of-moore-s-law.md): Fable's price ended the free lunch of ever-cheaper frontier models. Now we have to think about which model gets which work. (2026-08-23) - [Harnesses are Situated Agents](https://www.dbreunig.com/2026/08/14/harnesses-are-situated-agents.md): Harnesses manage the systems around agents, from the session, to the company, to the domain. (2026-08-14) - [Manage Your Agent's Loadout with Dr. Skill](https://www.dbreunig.com/2026/07/24/manage-your-agent-s-loadout-with-dr-skill.md): Keep your context clean by keeping your agent's skills and tools in check. (2026-07-24) - [Understanding the Dynamics of the AI Ecosystem with Pace Layers](https://www.dbreunig.com/2026/07/03/ai-ecosytem-pace-layers.md): Stewart Brand's Pace Layers framework reveals why AI's speed is straining the slower systems it depends on for support. (2026-07-03) - [What Do Humans Need From Docs?](https://www.dbreunig.com/2026/05/31/what-do-humans-need-from-docs.md): When an agent can explain anything, should we keep writing docs for humans? (2026-05-31) - [10 Lessons for Agentic Coding](https://www.dbreunig.com/2026/05/04/10-lessons-for-agentic-coding.md): Ten lessons for coding with agents, from implementing to learn to remembering that code is cheap but maintenance and security aren't. (2026-05-04) - [Cybersecurity Looks Like Proof of Work Now](https://www.dbreunig.com/2026/04/14/cybersecurity-is-proof-of-work-now.md): AISI's tests of Mythos suggest a new security economy: to harden a system you must spend more tokens finding exploits than attackers do. (2026-04-14) - [How Claude Code Builds a System Prompt](https://www.dbreunig.com/2026/04/04/how-claude-code-builds-a-system-prompt.md): Claude Code's leaked source shows a system prompt assembled from dozens of conditional parts. (2026-04-04) - [The Cathedral, the Bazaar, and the Winchester Mystery House](https://www.dbreunig.com/2026/03/26/winchester-mystery-house.md): The internet made coordination cheap and gave us the Bazaar. Coding agents make code cheap and are giving us sprawling, idiosyncratic software. (2026-03-26) - [Learnings from a No-Code Library: Keeping the Spec Driven Development Triangle in Sync](https://www.dbreunig.com/2026/03/04/the-spec-driven-development-triangle.md): What shipping a library with no code taught me: spec-driven development is a feedback loop between specs, tests, and code, not a straight line. (2026-03-04) - [Two Beliefs About Coding Agents](https://www.dbreunig.com/2026/02/25/two-things-i-believe-about-coding-agents.md): Talented developers underestimate the intuition they bring to their prompts, and coding agents amplify existing skills rather than replace them. (2026-02-25) - [Why is Claude an Electron App?](https://www.dbreunig.com/2026/02/21/why-is-claude-an-electron-app.md): If coding agents make native ports cheap, why does Anthropic still ship Claude as an Electron app? Because code is cheap but maintenance isn't. (2026-02-21) - [How System Prompts Define Agent Behavior](https://www.dbreunig.com/2026/02/10/system-prompts-define-the-agent-as-much-as-the-model.md): The model sets the ceiling on an agent's performance, but the system prompt decides whether it ever gets there. (2026-02-10) - [The Potential of RLMs](https://www.dbreunig.com/2026/02/09/the-potential-of-rlms.md): Recursive Language Models put long context in a REPL and let the model explore it, sidestepping context rot on logs hundreds of megabytes long. (2026-02-09) - [The Rise of Spec Driven Development](https://www.dbreunig.com/2026/02/06/the-rise-of-spec-driven-development.md): A month after whenwords, spec-driven development is spreading. Ports and emulators are the easy wins, but generated code still needs care. (2026-02-06) - [A Software Library with No Code](https://www.dbreunig.com/2026/01/08/a-software-library-with-no-code.md): Do we still need libraries of 3rd party code when AI agents are this good? (2026-01-08) - [2025 in Review: Jagged Intelligence Becomes a Fault Line](https://www.dbreunig.com/2025/12/29/2025-in-review.md): Looking back on 2025, the incredible pace of AI is stunning. But fast growth brings disconnects. (2025-12-29) - [Why I Write (And You Should Too!)](https://www.dbreunig.com/2025/12/27/why-i-write.md): Publicly writing is one of the most valuable things you can do. It improves your thinking, ability to communicate, and connects you to interesting people. (2025-12-27) - [How Model Use Has Changed in 2025](https://www.dbreunig.com/2025/12/19/how-model-use-has-changed-in-2025.md): We aren't just asking for chat anymore. A good chunk of us hit a single endpoint that runs code, uses a computer, edits files, and searches the web. (2025-12-19) - [Enterprise Agents Have a Reliability Problem](https://www.dbreunig.com/2025/12/06/the-state-of-agents.md): Every enterprise AI report tells the same story: chatbot adoption booms while internal pilots fail. Reliability is what holds agents back. (2025-12-06) - [Don't Fight the Weights](https://www.dbreunig.com/2025/11/11/don-t-fight-the-weights.md): When your context goes against a model's training, you struggle for output. Learn to recognize when you're fighting the weights. (2025-11-11) - [Glimpses of the Future: Speed & Swarms](https://www.dbreunig.com/2025/10/20/speeds-and-swarms.md): When coding models are fast and concurrent, the distance between thought and code shrinks. Qwen 3 Coder on Cerebras is a glimpse of the future. (2025-10-20) - [Enterprise AI Looks Bleak, But Employee AI Looks Bright](https://www.dbreunig.com/2025/09/15/ai-adoption-at-work-play.md): Read past the MIT report's headline: executives blame failed AI on employees, while employees quietly bring their own AI to work. (2025-09-15) - [AI Companies School Like Fish](https://www.dbreunig.com/2025/09/13/the-ai-product-cycle.md): Markets for new AI use cases turn from blue oceans to red oceans in under 18 months. Start-ups need to go niche or have one helluva plan. (2025-09-13) - [Can Chatbots Accomodate Advertising?](https://www.dbreunig.com/2025/09/02/considering-ad-models-for-ai-products.md): AI lets people project their decisions, which breaks the attention economy. Ads in chatbots have to serve the user's goal or fail. (2025-09-02) - [Building Castles in the Air, but With Surprise Physics](https://www.dbreunig.com/2025/08/21/castles-in-the-air-with-unknown-physics.md): How does the probabilistic nature of AI change the nature of computer engineering? (2025-08-21) - [Making Sense of AI Job Titles](https://www.dbreunig.com/2025/08/21/a-guide-to-ai-titles.md): AI job titles are a confusing mess of mix-and-match terms. This decoder ring breaks down the patterns behind titles. (2025-08-21) ## Optional - [Blogroll](https://www.dbreunig.com/blogroll.xml): OPML list of sites Drew follows. - [Sitemap](https://www.dbreunig.com/sitemap.xml)