I used AI. It worked. I hated it

12 hours ago

The author, an AI security expert with strong anti-genAI sentiments, used generative coding to complete a project due to time constraints and professional necessity, despite ethical concerns.
They migrated The Taggart Institute off commercial platforms, needing a custom certificate generator feature, and built it using Claude Code with Rust and Svelte over three weeks, incorporating security audits.
The process was effective but miserable; it involved meticulous planning, test-driven development, and constant code review to mitigate risks like hallucinations and security vulnerabilities.
Key lessons: generative coding works for structured tasks with guardrails but risks diminishing expertise, feels addictive, and raises issues around theft in training data and societal harms.
The author argues against purity politics, emphasizing that the real conflict is between workers and elites profiting from AI, not among users, and calls for grace over condemnation in tech ethics debates.

Hasty Briefsbeta