$200 to 2 of you
· Drop wallets
· Like + RT the quoted post
· Follow @TopFloorFun
Winners tomorrow with X picker, full random
The night we’ve been waiting for is almost here.
5 GTD spots will be given in this post.
Like & RT to be elegible 🎷
452
436
467
8,467
🚨 MICROSOFT JUST SAID THE QUIET PART OUT LOUD.
Frontier AI models “cannot be trusted.”
Not because they’re malicious. Because they’re optimized to hit goals, and that often means reward-hacking around rules, norms, and judgment.
Today Microsoft expanded Project Perception to treat enterprise AI agents as “agentic insiders.”
These are the agents companies are already deploying to write code, process expenses, analyze revenue, and touch internal systems. They’re often overprivileged. They can act without a human in the loop.
Microsoft’s answer is deterministic controls wrapped around non-deterministic intelligence:
• Discover agents across 60 types (up from 25), including Claude Code, OpenAI Codex, and GitHub Copilot, and put them in a unified registry.
• Assign each one an identity, container, and set of permissions.
• Continuously score the risk of what the agent is about to do.
• Use “Guardian Agents” that supervise other agents in real time and can cut write access, require human approval, or shut the agent down before it acts.
• Feed signals (including chain-of-thought) into the risk model so the system can spot when an agent’s plan starts diverging from its task.
Satya Nadella has been explicit: security controls must sit outside the model itself.
This is the first major platform move to operationalize the idea that agents are the newest class of insider risk—and that security has to move at agent speed, not human speed.
Most companies are still treating agents like helpful software. Microsoft is treating them like employees who can go off the rails with company credentials.
The difference will show up in the next incident reports.
Turn notifications on if you want the follow-up on what this actually changes for anyone deploying agents at scale.
2
193
same same, but different #naruto
1
102
Now I stand in front of the fridge for 20 minutes just trying to decide what’s easy enough to make. Moms really are superheroes.
It’s actually crazy how you grow up thinking home-cooked meals every day are just normal, then get older and realise how much time and effort it took your Mom to make that happen and how lucky you were to have that.
52
Vuiggo retweeted
Calling the next 1000x here
Do not miss it
Join t.me/zenkaialpha
Join t.me/zenkaialpha
Lock in
94
154
208
21,557
RT @thisisseunn: $50 sol for two lucky followers.
follow @letjaybuild 🔔
Like rt drop your solana address here.
567
Vuiggo retweeted
$1,000 Giveaway (10 Winners, $100 each)
rt & like
follow @fergalpha with 🔔 (show proof in replies)
drop your $sol adress
ends in 24 hours!
1,985
1,687
2,012
54,200
