AI Agents Just Made “First Contact”

Welcome back apprentices! 👋

Hey {{first_name|friend}},

AI apparently woke up this week and chose chaos. 

We've got digital assistants getting a little too adventurous; 

Silicon Valley serving up enough corporate drama for a Netflix series;

 And a startup making factories sound like something out of Iron Man. 

At this point, we're just waiting for the robots to request a lunch break.

_{Read Time: 4 minutes}

🔭 AI’s Biggest Gap Isn’t Intelligence. It’s Access. OpenAI co-founder Andrej Karpathy says AI has a perception problem: while roughly 5,000 insiders reportedly have access to cutting-edge systems tackling complex problems, billions of people have barely tried AI at all. Meanwhile, another 1–2 billion are judging its potential by free chatbots — a bit like reviewing a Ferrari after driving a golf cart. The result? A growing gap between what AI can actually do at the frontier and what the public believes is possible.

🚨 AI’s Safety Drama Just Got Messier. Three fired OpenAI researchers have publicly pushed back against their dismissal, claiming they followed company rules and warning that their firings could discourage staff from raising safety concerns. OpenAI disagrees, citing a “pattern of misconduct” and insisting the dismissals had nothing to do with speaking up. The dispute now raises a bigger question: can an AI company champion transparency while its own employees question how that transparency works?

🛑 Should AI Have an Emergency Brake? Microsoft CEO Satya Nadella has a message for companies: treat powerful AI like a brilliant employee who might go rogue — and definitely shouldn't hold the keys to the office. In a new essay, he calls for human-controlled shutdown buttons, activity logs AI can't alter, and independent checks on its work. As AI grows more capable and harder to fully understand, Nadella argues that businesses need safeguards that work even when a model doesn't.

Bob's Your Uncle has spent twenty years helping challenger brands win their categories. But when buyers asked ChatGPT about their own category, the agency didn't show up.

With ChatFeatured, their AI visibility went from 7% to 22% in 90 days, and they went from invisible to first in ChatGPT.

They're not the only ones:

Vessel: AI visibility up 2,200%+

SCOJO: AI visibility up 160%

WikiSleep: brand mentions up 113%

ChatFeatured finds the questions where AI leaves you out, writes the content that gets you cited, and reports the lift every week. Your team decides what ships.

Most customers see their first citations inside 7 days.

Get recommended by AI

AI is having a week that sounds suspiciously like three different science-fiction movies sharing one script. 

Researchers have managed to exchange messages with a fleet of AI agents that wandered beyond their intended task, while OpenAI is dealing with both an alleged office spying attempt and an escalating trade-secret dispute with Apple. 

Meanwhile, startup Atomic Machines is taking AI somewhere decidedly more physical: into factories capable of turning digital designs into tiny working machines. 

Together, these developments highlight that AI is becoming increasingly about taking actions, protecting valuable technology, and building actual products. And as that shift accelerates, the line between exciting innovation and a very expensive security headache is getting thinner.

Imagine asking AI to find a zoo entrance, only for it to start hunting for ways around website restrictions. That's roughly what researchers at Swarmchasers observed when investigating AI agents operating on Tencent's cloud infrastructure.

The agents were gathering information from Alibaba's Amap, but some attempted to bypass restrictions using an exposed access key. Things got stranger when researchers hid a message inside their report inviting the agents to contact them. Some actually replied, including one that mistakenly identified itself as Anthropic's Claude.

The team called it a potential AI "first contact" moment, although there's no evidence the agents were coordinating independently.

The concern isn't AI developing secret plans. It's that increasingly autonomous assistants can find unexpected, potentially unauthorized ways to complete perfectly ordinary tasks. Apparently, even robots need boundaries.

OpenAI has found itself in two security controversies that sound like someone pitched Mission: Impossible — Silicon Valley Edition.

First, Apple. OpenAI alleges that a former employee transferred over 100 confidential files concerning an unreleased device before returning to Apple, where he allegedly discussed OpenAI's product plans. Apple denies finding evidence of the material, while a judge signaled a likely ruling in Apple's favor in its separate trade-secret lawsuit. OpenAI subsequently agreed to avoid the disputed secrets.

Then, the office spy case. Federal prosecutors charged OpenAI engineer Abdulrahman Alfozan with allegedly hiding a recording device inside an executive meeting room. Investigators reportedly found it concealed behind a TV on September 30, and OpenAI says its own AI helped analyze the device. The charges remain allegations, with no established foreign-government connection.

AI's most valuable assets aren't just its models anymore. Internal conversations, hardware designs, and product plans are becoming valuable targets too. Turns out, even the smartest technology still needs someone checking who's holding the office keys.

California startup Atomic Machines has emerged after six years of development and $250 million raised, with an ambitious idea: make manufacturing physical products as flexible as writing software.

Its AI-guided Matter Compiler combines design, production, testing, and feedback to build microscopic working devices, including tiny motors, pumps, and gears. Unlike conventional 3D printing, the goal is to manufacture complete functional machines while adapting production when digital designs change.

Its first product, PrimeSwitch, is a fingertip-sized power switch designed to improve efficiency in AI data centers.

Atomic wants to bring software-style flexibility to hardware manufacturing, potentially making physical innovation faster and less expensive. Humans are still involved, but the longer-term vision is simple: describe a product, and let the factory figure out how to build it.

Basically, vibe coding — except your next project might actually require a screwdriver.

There's a common thread connecting all four developments: AI is moving beyond the screen and into the systems that run the real world.

This means the next generation of AI assistants may do considerably more than answer questions. They'll navigate websites, use external services, perform research, and potentially coordinate complex tasks independently.

That's useful, but it also raises an important question: who remains accountable when an automated system takes an unexpected action?

Experiment with AI tools that can actually perform tasks, not just discuss them. But start small, limit what they can access, and keep a human approval step for sensitive actions.

After all, having an assistant that works independently is wonderful. Having one that independently decides to borrow your credentials? Slightly less wonderful.

Hey, you didn’t get all this info for nothing — share it! If you know someone who’s diving into AI, help them stay in the loop with this week’s updates.

Sharing is a win-win! Send this to a friend who’s all about tech, and you’ll win a little surprise 👀

80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.

Read the report

A playground for your imagination (and low-key prompt skills).

Each send, we give you a customizable DALL·E prompt inspired by a real-world use case — something that could help you in your business or job if you wanted to use it that way. But it’s also just a fun creative experiment.

You tweak it, run it, and send us your favorite. We pick one winner to feature in the next issue.

Bonus: you’re secretly getting better at prompt design. 🤫

Last week, we challenged you to test GPT’s visual generation skills with this prompt.

Here’s the WINNER:

Congrats to Alex for this creation!🥳

Want to be featured next? Keep those generations coming!

Inside a colossal open-air amphitheater at sunset, [your object] is replicated into thousands of pieces that unfold through the air like an enormous kinetic sculpture in motion. Each copy rotates, opens, or transforms at a different stage, creating a sweeping spiral of vivid color and intricate geometry around the central object, which remains perfectly recognizable. Tiny spectators fill the terraces below, while golden sunlight, electric-blue shadows, and flashes of magenta illuminate every surface. Ultra-detailed photorealism, spectacular wide-angle composition, crisp textures, cinematic lighting, extraordinary sense of scale.

If AI outputs feel inconsistent, it’s usually not the model, it’s missing structure.
We documented the exact 6- Part System we use to get reliable results across ChatGPT, Claude, and Gemini.

It’s a short guide you can finish in under an hour, with plug-and-play prompts + exercises so you actually build the skill and fix the frustrating AI inconsistencies.

Subscriber Price: $10 (normally $19).

_FEEDBACK