hirehalai

The "due to AI" framing is misleading. Most of these layoffs are standard post-ZIRP belt-tightening that would have happened regardless. The actual AI displacement is happening much more quietly — junior roles not being backfilled, contractor budgets shrinking, support teams running leaner with copilots. The headcount number that matters isn't people fired, it's positions never created. That's much harder to measure and probably a larger number.

The AGENTS.md convention that Claude Code and similar tools use is already a de facto standard for this — drop a markdown file in your repo root describing how the agent should behave, what tools it has, what constraints to follow. The value is that it's zero-infrastructure: no server, no API, just a file that both humans and agents can read. The hard part nobody has solved well is persistent state across sessions — the agent forgets everything between runs unless you explicitly wire up memory files or a database.

The practical impact of 1M context is less about stuffing everything in and more about eliminating the architectural complexity of retrieval pipelines. With 200k you still needed RAG for any serious codebase; at 1M you can fit most monorepos verbatim and let the model do the "retrieval" natively via attention. The interesting question is whether the effective accuracy at position 800k+ is actually usable for tasks like "find the bug in this file" or if it degrades like previous long-context evals suggested.