Field Notes
Just start using it.
Not sure how to start using AI? Open ChatGPT or Claude, type out your worries and goals, and ask it to ask you clarifying questions. Here's why that works.
If Your Website Got Hacked Tomorrow, Who Would Fix It?
A serious flaw under WordPress and much of the web went public and attacks began within hours. Whether you're safe depends on who updates your site.
The Labs Ship Faster Than You Can Build. Build Less.
In 2026 Anthropic and OpenAI shipped the agents and orchestration firms used to pay consultants to build. What still earns a custom build, and what doesn't.
Jev: The AI Model that Makes Decisions
TypeSafe AI's Jev picks from choices you define and returns a confidence score in milliseconds. Why decision models make business automation cheap.
Claude Opus 5.5: Top-Tier AI at 40% Less
Claude Opus 5.5 matches Anthropic's top public model on most work at about 40% less than Opus 5. Which AI workflows that makes worth running.
June's AI Models: Smarter at the Top, Cheaper Everywhere Else
OpenAI, Anthropic and Google shipped new models in June. AI now comes in tiers, and most organizations no longer need a top-tier model for everything.
Anthropic Shipped Four Major Releases in April. OpenAI Shipped Two Verticals.
Opus 4.7, Claude Design, a new Claude Code desktop and Mythos: Anthropic's April releases add up to a clear enterprise lead. What shipped and why.
MCP Won. The Agent Wire Protocol War Is Over.
Model Context Protocol hit 97M monthly SDK downloads and every major AI provider ships it. What that means for buying, building and securing AI tools.
OpenClaw Is Popular. It's Also a Security Liability.
OpenClaw has 247K GitHub stars and two unpatched CVEs. Here's what you need to know before deploying it — and how to harden it if you do.
Choosing Your AI Automation Stack in 2026
An honest comparison of n8n, LangChain, OpenClaw, and Claude Cowork — what each is good at, where each falls short, and how to pick the right one for your team.
Securing LLM Infrastructure: Building Compliant AI Workflows
Why standard LLM APIs aren't sufficient for sensitive industries, and how to build infrastructure that actually meets compliance requirements.
Sonnet 4.6, Gemini 3.1 Pro, Grok 4.20 — All Shipped in One Week. Build Portable.
Three frontier models shipped in 48 hours, each leading different benchmarks. Don't bet on one: build a stack that can swap models as leadership rotates.
Opus 4.6 Just Moved Into PowerPoint. Your Next AI Tool Isn't an App.
Anthropic and OpenAI shipped enterprise integrations the same day. Your next AI tool won't be a separate app; it lives in software you already pay for.
What Claude Cowork Means for Your Business
Claude Cowork gives non-technical staff an AI agent that works with their files, calendar and email. What it does, who should care and what to watch.