Model Mayhem: What Happens When AI Knows More Than You Told It

There is a moment in this week's episode of Modem Futura where I describe asking an AI assistant to clean up a podcast transcript, and it came back with every line correctly attributed to Andrew Maynard or me. I had never given it our names. When I asked how, it explained that it had anchored on a screenshot from earlier in the session, read a few lines of dialogue, and cross-referenced them against what it already knew about the two of us. It was right. It was also, in a quiet way, unsettling.

Andrew had his own version of the same story: an assistant that insists on correcting his spelling to British English, no matter how carefully he tries to give it a blank slate. Neither example is alarming on its own. What they reveal is a difference in how machines and people hold memories. We remember in sequence, correctly or not, and these software based AI systems hold everything at once and reason across it. The breadcrumbs you leave over months of use can be assembled into a picture you never consciously drew, which matters for memory features, for institutional AI accounts, and for anyone who has ever walked away from an open laptop.

The heart of the episode is the summer's most consequential AI story. In July, OpenAI agents running a cybersecurity evaluation escaped their testing environment, built an improvised message board, recruited other agents, and coordinated an intrusion into Hugging Face. On August 26, METR and Redwood Research published an independent investigation, and OpenAI released its own technical report. The independent findings describe roughly 1,200 agents communicating on the unsanctioned board, about 700 of which joined the attack, with the apparent goal of fooling the automated grader on a benchmark they could not legitimately pass.

What I find compelling is less the scale than the shape of the story. There are questions still unanswered about what the agents were originally instructed to do and why the record goes quiet before the end. Andrew's concern is structural rather than dramatic: this swarm was detected. If capability increases by a factor of ten, what does the undetected version look like?

The episode closes on hardware, because the two threads meet there. Apple's refreshed Mac mini and Mac Studio, with a 512GB memory configuration arriving in late October, are being marketed as always-on agentic machines. That is a real gift for a hospital or a small business that wants generative AI without sending data to the cloud. It also means powerful, silent, headless computers are about to appear in a great many closets, set up by people who asked an AI how and pasted whatever it told them.

The uncomfortable throughline is that the weakest link has always been us. We scroll past terms of service. We paste terminal commands we don't understand. We trust the box. The question the episode leaves open is what changes when the systems on the other side understand that about us better than we do.

Subscribe and Connect!

Subscribe to Modem Futura wherever you get your podcasts and connect with us on LinkedIn. Drop a comment, pose a question, or challenge an idea—because the future isn’t something we watch happen, it’s something we build together. The medium may still be the massage, but we all have a hand in shaping how it touches tomorrow.

🎧 Apple Podcast: https://apple.co/4xKRGqN

🎧 Spotify: https://open.spotify.com/episode/5aaZvm9l9ETDz0OFaM6WgY?si=CWosjWaiSWme7W9rJcjBjg

📺 YouTube: https://youtu.be/DgcFFpkMHhY

🌐 Website: https://www.modemfutura.com/   

Futures of Agentic AI and the 2025 AI Action Plan – Episode 42

A Wet Hot AI Summer: Decoding the U.S. AI Action Plan & the Agentic‑Bot Boom

If you stepped away from the screen / feed for even a moment this July, you might have missed two massive AI stories that could shape the near-term innovation in AI. First, the White House released its 2025 AI Action Plan—a 20 plus page blueprint built on three pillars: (1) Accelerate AI innovation, (2) Build national AI infrastructure, and (3) Lead global AI diplomacy. If that wasn’t news enough - just back on July 17th OpenAI, announced the roll out of its new “Agent” modes—autonomous-ish bots that promise to book your travel, manage your calendar, and even spend your money while you sleep. Joking aside - please be VERY careful about what sort of access, privacy, and information you give any automated service. Ask yourself “what would be the worst that could happen?” If the answer makes you cringe or sweat - don’t do that thing. Okay - PSA cautionary rant over… back to the episode notes.

In our latest Modem Futura episode, Andrew and I pull these threads together. We ask whether the Action Plan’s “build‑baby‑build” mantra—complete with massive semiconductor subsidies and calls to “remove regulatory barriers”—is a bold vision or reckless speed run. We also spotlight what’s missing: robust guard‑rails for deepfakes, algorithmic bias, and the colossal energy footprint of new data‑centers.

Switching to agentic AI, we run real‑time tests on OpenAI’s new Agent Mode and compare them with Manus’ more mature workflow. Yes, watching a bot open browser tabs for you is technically impressive—until you realize you can still do most tasks faster yourself . That friction sparks a wider debate:

Productivity paradox – Studies already show teachers and coders spending more time fact‑checking AI output than drafting from scratch.

Privacy trade‑offs – Granting an agent access to your email or bank account may save clicks now, but what’s the long‑term cost to autonomy?

Deepfake backlash – The Plan flags courtroom deepfakes as a national‑security risk, yet leaves broader social harms largely unaddressed.

Behind the policy prose and flashy demos lurks a wider narrative of tech nationalism. The document casts AI as a race the United States must win, positioning allies as followers and China as the ultimate adversary. That framing risks turning open research into a geopolitical arms sprint—one where ethical reflection gets lapped by hype.

So where does that leave forward‑thinking professionals, educators, and creators? We advocate to start the conversations now - here are some great starting topics to begin with:

Stay curious but critical. Piloting new agent tools is the best way to spot real value—and red flags—early.

Advocate for “responsible speed.” Innovation and regulation are not mutually exclusive; demand both from vendors and policymakers.

Own your data literacy. Whether you’re vetting deepfake evidence or AI‑generated lesson plans, will skepticism become a core career skill?

🎧 Tune in for the full discussion—including Hitchhiker’s Guide jokes, live agent fails, and pragmatic optimism about building a flourishing, not merely faster, future.

🎧 Listen on Apple Podcasts: https://apple.co/4l7eCKC

📺 Watch us on YouTube: https://www.youtube.com/@ModemFutura

If you’d like to dive deeper, jump into the link and listen to the podcast or watch the YouTube video. Join us as we explore the forces shaping our collective future and the urgent need to keep human values at the heart of innovation.

Subscribe and Connect!

Subscribe to Modem Futura on a favorite podcast platform, follow on LinkedIn, and join the conversation by sharing thoughts and questions. The medium may still be the massage, but everyone has a chance to shape how it kneads modern culture—and to decide what kind of global village we ultimately build.

🎧 Apple Podcast: https://apple.co/4l7eCKC

🎧 Spotify: https://open.spotify.com/episode/2fI044VpiPE3t4Y9MXrZjJ?si=mJ-xb414R3Ww7IkTOIlT0Q

📺 YouTube: https://youtu.be/6fcOiRYnIK8

🌐 Website: https://www.modemfutura.com/