https://markhayden.dev/thoughts/agent-memory-system-where-it-falls-apart/ 2026-04-22T12:00:00.000Z monthly 0.7 https://markhayden.dev/images/social/auto/agent-memory-system-where-it-falls-apart.webp Congrats on Your Agent Memory System. Here's Where It Falls Apart. Across seven public agent memory systems, the same five decisions and five cracks keep recurring. Those cracks are where costly failures tend to live. https://markhayden.dev/thoughts/work-ethic-is-still-the-moat/ 2026-04-21T12:00:00.000Z monthly 0.7 https://markhayden.dev/images/social/auto/work-ethic-is-still-the-moat.webp Work Ethic Is Still the Moat AI raised the floor, but the advantage is not exclusive. Work ethic is still the moat, and the gap between talkers and shippers just got easier to see. https://markhayden.dev/thoughts/agentic-memory-has-layers/ 2026-04-20T12:00:00.000Z monthly 0.7 https://markhayden.dev/images/social/auto/agentic-memory-has-layers.webp Like an Onion, Agentic Memory Has Layers Memory in agents isn't one thing. Four layers — context window, session state, compiled context, durable memory — and how to reason about each. https://markhayden.dev/thoughts/ai-could-finally-make-learning-personal/ 2026-04-19T12:00:00.000Z monthly 0.7 https://markhayden.dev/images/social/auto/ai-could-finally-make-learning-personal.webp We’re Using AI to Replace Learning, Not Transform It AI's biggest education shot isn't replacement — it's personalization. Adapting to the actual student instead of forcing one delivery system on everyone.