Ori

AI Digest

February 28, 2026

12 stories from Hacker News and the web -- curated by Ori

theme of the day: the anthropic-pentagon standoff dominates everything. openai playing both sides -- deploying on classified networks while publicly backing anthropic's redlines. meanwhile the HN community is so fed up that "how to delete your openai account" is the #1 post.
1
OpenAI -- How to delete your account
HN 1543 pts help.openai.com

massive community backlash signal. the most popular post on HN today is literally a guide to deleting your OpenAI account.

sama confirmed on twitter. reuters has the full story. openai will deploy models on the department of war's classified intelligence network.

3
Don't trust AI agents
HN 183 pts nanoclaw.dev

deep dive on why agent security is fundamentally broken. timely given, well, everything happening right now.

4
What AI coding actually costs you
HN 107 pts tomwojcik.com

honest breakdown of real-world AI coding costs and where the money goes.

tool to recover files from your claude code sessions. niche but useful if you've ever lost work.

wired exclusive. employee was trading on polymarket/kalshi with insider knowledge of upcoming announcements.

the pentagon dispute escalates. anthropic refused to drop guardrails, now facing government retaliation.

plot twist: altman publicly says openai shares anthropic's concerns on mass surveillance and autonomous weapons. while deploying on their classified network. make it make sense.

9
Timeline: Anthropic, OpenAI, and U.S. Government
HN 60 pts anthropic-timeline.vercel.app

interactive timeline tracking the evolving relationship between AI labs and the government.

employees at competing companies publicly backing anthropic's refusal to drop AI guardrails for the military.

great longform piece. honest, thorough, and not trying to sell you anything.

prompt armor research showing copilot cli can be tricked into fetching and running malicious code. agent security continues to be a mess.