Field notes on building AI software
Shipping Trustworthy AI: A Deep Dive into Avoiding Hallucinations
Hallucinations are not a model quirk — they are a systems problem. Here is the production playbook our members use to keep AI software honest: grounding, constrained outputs, verification passes, and eval harnesses that catch drift before users do.
Read article
Why Humanizing Your Blog Content Matters More in the Age of AI
When every competitor can generate a thousand competent articles a week, competence stops being a differentiator. The blogs that win now are the ones that sound like a specific person with specific experience. Here's how to keep yours human.
Read article
This Week in AI: Gemini 4 Argon Goes Wide, Meta Ships Muse Wonder, and the Enterprise Gets Agentic
A packed seven days: Google's Gemini 4 Argon hits general availability with a million-token context, Meta enters the closed-weights frontier race with Muse Wonder, Claude Haiku 5.5 drops the price of subagent work, and agents quietly move into HR, finance, and the Copilot stack.
Read article