The author explains how AI coding agents have revived their belief in Test-Driven Development (TDD) after abandoning it years ago. They describe how AI tools can automate the TDD workflow, making it more practical and effective for modern development.
Claude Code has 2-3 million users, representing about 1% of knowledge workers. The compute requirements become concerning as adoption scales from this point.
DRAM shortages, not capital constraints, will be the primary limiting factor for AI infrastructure growth through 2027. The article analyzes how memory bottlenecks will shape compute capacity expansion in the coming years.
Anthropic's red team discovered over 500 critical vulnerabilities using Claude AI, focusing on maintained software. The greater concern lies in the long tail of vulnerabilities in abandoned software that will likely never be patched.
AI compute demand is growing exponentially while supply constraints are severe. The next 18-24 months will be characterized by shortages, rationing, and price discovery.
Current approaches to securing coding agents, including permission systems, Docker sandboxing, and log file protections, have significant limitations. These security measures often fall short of providing adequate protection for coding agents, suggesting that alternative solutions may be necessary.