#AI

Two days after I wrote a love letter to Claude Fable 5, the US government forced Anthropic to shut it off. Here's what actually happened, why it happened, what it means for developers outside the US, and what Anthropic does next.

Building a spending prediction engine that tells you how much you'll spend next month — two model versions, confidence intervals, and a lot of humility.

A junior dev committed a .env file to staging. We thought we deleted it. Git never forgets — and in 2026, AI agents are making secret leaks dramatically worse.

How networking actually works in your app, why your AI chatbot is choking on its own history, and why OpenAI just flipped the script with Realtime WebSocket

Why does Claude rewrite your auth layer at 2am while Codex stares at a CSS bug and refuses to touch it without a screenshot?

One engineer, $1,100 in AI tokens, and seven days. Cloudflare's vinext is a full reimplementation of the Next.js API built on Vite — and it's already running a government website in production.

After 6 days of testing Claude Opus 4.6 against Codex 5.3 on a production codebase, the benchmarks lie. One eats 9GB of RAM and architects chaos; the other surgically refactors but needs screenshots to fix CSS. Here's what the SWE-bench scores won't tell you about agentic coding in 2026.

New Anthropic research reveals AI coding assistants might be stunting developer growth. Here's why your Copilot addiction is creating a skills crisis—and what to do about it.

Every once in a while you inherit a greenfield problem that feels equal parts thrilling and terrifying. Ours was simple to describe and brutal to execute: build a vendor onboarding experience where no field is hardcoded, every service type defines its own rules, and UI + validation + submission logic are all generated from server-delivered JSON that we lovingly called DNA. This is the story of how that system came to life inside `webapp_frontend`, what it looks like today, and the lessons learned while making it production-ready.

Spent 50 hours on GPT-5 for Flutter development: fixing bugs, adding features to MS Bridge. My honest take on OpenAI's 'game-changer' claims for software engineers.