AI coding tools feel faster—but a rigorous study found developers were 19% slower while believing they were 20% faster. Here's how to measure real AI ROI using DORA, SPACE, and metrics vendors won't show you.
Tools and small games built in the open. Source on GitHub.
One email per essay or tool — no drip, no upsell, unsubscribe in one click. Roughly twice a month.
Long-form on software, AI, and the practice of building things you actually use.
Single models are hitting diminishing returns. The bigger gains now come from making top-tier models critique and verify each other's work. Here's the pattern — debate, judge, synthesize — and runnable starter code to build it.
Anthropic shipped Fable 5, a new tier above Opus. Look at how it's priced and constrained and you can see a deliberate philosophy: the more capable the model, the more carefully it's gated. Here's what that means for builders.
Uber burned its entire 2026 AI budget by April. Microsoft started cancelling Claude Code licenses. Yet productivity gains are real. Is agentic coding a leap or an expensive trend? A cost-benefit verdict.
Enterprises will spend $665B on AI in 2026, yet 73% can't prove a return and one firm burned $500M in a month on runaway tokens. Here's a practical framework to stop AI experiments becoming budget leaks.
Open-source AI isn't free—it just moves the cost from an API invoice to your payroll. Here's the real total cost of ownership, the volume break-even point, and how to decide which is cheaper for your team.
The honest answer isn't 'jobs' vs 'no jobs'—it's about tasks. AI is automating the verifiable parts of coding and amplifying the judgment parts. Here's which developer tasks are changing, and what survives.