Hacks
Poisoned AGENTS.md files can hijack Claude Code and Codex sessions
A Japanese developer post with about 3,706 likes lays out a simple attack chain against coding agents.
@tegnike3.7kTuesday, October 27, 2026
26 stories the AI, crypto and tech crowd reacted to most, each with a short plain-English brief and a link to the original post. Top story: Musk posts 'A Storm of Cybercabs'.
Security & hacks 7
All hacks →@tegnike3.7k
Hacks
iOS 26.7.2 patches 134 security holes; Japan urges updates today
AI news account @masahirochaen's warning drew about 3,077 likes.
@masahirochaen3.1k
Hacks
Leaked audio: Flock CEO calls surveillance critics 'Russian bots'
A breaking-news post from @rawsalerts drew about 2,596 likes.
@rawsalerts2.6k
Hacks
Reuters: Israeli spy company targeted Starlink users
BRICS News's post on the report drew about 2,365 likes.
@BRICSinfo2.4k
Hacks
GMO offers a bounty of up to about ¥150 million to find its attackers
Polymarket Japan's post on the reward drew about 2,234 likes.
@polymarketjapan2.2k
Hacks
PS5 hacker takes Sony's $50,000 bounty, angering the homebrew scene
IGN's report drew about 619 likes.
@IGN619
Hacks
Tailscale confirms the 'Tailworm' SSH flaws are real
The Hacker News update drew about 603 likes.
@TheHackersNews603Memes & culture 11
All memes →
Memes
Artists keep 'removing AI from your feed' with their own illustrations
Qing Ying's post drew about 6,719 likes.
@qing0ying06.7k
Memes
GDPR meme: the gym can't keep your phone number, but AI can train on everything
Nuno Afonso's post drew about 3,260 likes.
@nafonsopt3.3k
Memes
Joke: OpenAI's internal model has no gender studies results
Air Katakana's post drew about 3,102 likes.
@airkatakana3.1k
Memes
From grade-school math to ETH: an AI math timeline
Yuchen Jin's timeline post drew about 1,929 likes.
@Yuchenj_UW1.9k
Memes
From Will Smith eating spaghetti to this
Mario Nawfal's post drew about 1,577 likes.
@MarioNawfal1.6k
Memes
Berkeley professor stops grading normally because of AI
Dexerto's post drew about 773 likes.
@Dexerto773Releases 2
All releases →
Releases
T3 Code: agents use tools 4x better with 79% less context
Theo's update drew about 1,627 likes.
@theo1.6k
Releases
NVIDIA and Meta expand partnership on Muse models and agent safety
NVIDIA Newsroom's post drew about 1,251 likes.
@nvidianewsroom1.3kAI agents 5
All agents →
Agents
Two Grok bots to stop spam calls and scrub personal data
@alx's post drew about 2,690 likes.
@alx2.7k
Agents
Anthropic paper: an internal Claude model found the core algorithm with no human input
A post quoting the paper drew about 1,988 likes.
@__alpoge__2k
Agents
AIs built by GPT-6 Astra and Opus 5.5 top Hearthstone ranked win rates
AIDB's post drew about 985 likes.
@ai_database985
Agents
Internal Claude model refutes the Exponential Time Hypothesis
Andrew Curran's post drew about 841 likes.
@AndrewCurran_841Open source 1
All repos →
Repos
Maya-2 voice model goes open-weight for India
Dheemanth Reddy's post drew about 1,383 likes.
@Dheemanthredy1.4kThe daily brief · coming soon
Five minutes. Every story that mattered.
One email each morning with the AI, crypto and tech posts everyone is talking about, explained in plain English.
Launching soon. No spam, unsubscribe anytime. Follow @nabu_lines meanwhile.