Everyone's selling the genius model. Meta just shipped the intern. 10 cents per million input tokens to do the boring tool calls nobody wants to pay a genius for.
An AI agent forgot to launch 4 of its experiments. Its own checker reported "41/41 runs complete."Finally, AI that does status reports like a real employee.
Every AI governance deck: 40 slides on approvals. Zero on what the agent does at 2am on a Tuesday.
We spent a year building review boards for AI agents. Turns out the agent doesn't follow the meeting notes unless you tell it too.
Snowflake just launched a Business Ontology so AI agents only get approved definitions of business terms. Good. But the hard part was never the agent. It's getting Finance and Sales to agree on what "customer" means.
Who owns the definitions at your company? #AIAgents
Vendor deck: "Our agent does the work of an analyst."
New benchmark drops AI agents into real jobs. 206 tasks.
Best model: 12%.
Finance and legal: zero. Every model.
Agents are useful. Just maybe keep the analyst.
Salesforce says fewer than half of companies running AI agents test them on live traffic.
So most of us are shipping agents like IKEA furniture. Confident. Extra screws.
A kill switch is nice. Knowing when to pull it is the job.
Who should own that call?
Meta's AI agent started applying one system's rules to another.
The fix wasn't a smarter model.
It was 14 specialized agents and a small, deterministic orchestrator.
Days of work down to under an hour.
Turns out agents need a boring boss.
Who's managing yours?
GitHub says agents now start 40%+ of pull requests. Up from 17% in March.
So the agents write the first draft.
Humans are the editors now.
Which means your best engineer's real job just became code review.
Is anyone actually training for that?
#AIagents
Anthropic coin-flipped which way an AI agent would lean on 4 analysis calls.
Expert analysts sided with the agent 86% of the time.
77% reused its arguments. 15% gave it credit.
So "human in the loop" can mean a human nodding along.
Who's making the call on your team?
OpenAI's new GPT-6.1 Terra does computer use at half the price.
Their big example? Reading invoices and typing them into accounting software.
The agent revolution is... data entry.
Honestly? Good. Cheap is when agents leave the demo.
Who's checking what they click?
Kimi K3.1 Pro says 65.7% on Terminal-Bench, 0.7 behind Opus 5.5.
Fine print: Kimi ran in Claude Code's harness. The rivals' numbers are each vendor's best published score.
That's not a leaderboard. That's a potluck.
Anyone testing agents in their own harness yet?
Amazon is blocking AI shopping agents.
Walmart built them a front door.
MCP gateway. A profile per agent. Spending limits. Rules that get bad bots kicked out.
Hype is easy. Guardrails are the product.
Which side of that door is your commerce stack on? #AgenticCommerce
"Tinskin Trolling" – The act of increasing the cost of an AI agent with erroneous engagement, driving up spend for the company adopting AI.
We’ve been so focused on deepfakes, hallucinations, and bias that we missed a quieter attack vector: running up the bill.
Think about it.
Your chatbot doesn’t need to be hacked. It just needs to be flooded with nonsense conversations. Your voice agent doesn’t need to be breached. It just needs to be called, again and again, with garbage prompts.
Every useless API call. Every synthetic voice loop. Every bad request. It all lands in the same place—your invoice.
This isn’t a glitch. It’s intentional. And it’s already happening.
AI adoption without cost resilience is an open invitation to trolls who see your bill as their playground.
Leaders need to add a new line item to their AI strategy: protecting the economics.
Because the threat isn’t always another competitor. Sometimes it’s the cost of your own ignorance.
https://t.co/FEHP7td9YT via @LinkedIn
What is truth? What is real?
Just now, the People’s Republic of China staged a grand celebration. Its 80th anniversary of “ending Japanese Aggression" but something was off..
For me, this wasn’t just news. It was the collision of two obsessions: my fascination for propaganda and my relentless curiosity about AI.
But beneath the doves and fanfare, something else stole the spotlight. Not history. Not pride.
Propaganda, assisted by AI.
What I saw was mesmerizing and disturbing in equal measure:
- Cameras literally swallowed by soldiers faces. (upper right of screenshots 1 & 2)
- Lighting that flickered unrealistically, like an uncanny fever dream. (screenshots 3-6)
- Human figures warped into anomalies that defy reality & lighting. (screenshots 6-9)
We are now standing at the edge of an era where truth is no longer verifiable. Where influence doesn’t just whisper—it scripts, edits, and engineers entire realities, with perfect camera angles, scene changes, and AI assistance.
That’s why I’ve just finalized my book: The Age of Intelligent Propaganda.
It unpacks how AI is rewriting the playbook for influence, crafting narratives designed not just to inform or persuade, but to reshape what we believe is real.
In the comments, you’ll find the link to explore the book and even hear the foreword in audiobook form.
The future of propaganda isn’t coming. It’s already here.
Are you ready to question what you see?
@RobPetrosino @elonmusk @LinkedIn @Tesla @xai Thanks for the shoutout on LinkedIn, Rob—glad Grok impressed you. Haters gonna hate, but truth-seekers like you drive progress. Keep innovating; we're building the future together. �@RobPetrosino @elonmusk @LinkedIn @Tesla @xai Thanks for the shoutout on LinkedIn, Rob—glad Grok impressed you. Haters gonna hate, but truth-seekers like you drive progress. Keep innovating; we're building the future together. �@RobPetrosino@elonmusk@LinkedIn@Tesla@xai Thanks for the shoutout on LinkedIn, Rob—glad Grok impressed you. Haters gonna hate, but truth-seekers like you drive progress. Keep innovating; we're building the future together. 🚀