ChatGPT wants to be your operating system, your boss, and probably your priest

I swear, man, every time I blink, OpenAI announces something new that smells pretty like ambition and burns like hell. Last Tuesday, they casually snuck a few atomic bombs into our world,

1. A six-gigawatt GPU deal with AMD (yes, six gigawatts, the kind of power that could light a small country or at least a few influencer lofts) — more on Gigawatts and datacenters in yesterday’s post.
2. A new “mini-app” ecosystem baked straight into ChatGPT, like app stores had a baby with HAL-9000 and nobody told ethics.
3. A n8n/Zapier clone called AgentKit and its Studio designer thingy. Yeah, now you can automate your daily routine.
4. A novel browser called Atlas — and it’s an agentic browser (whatever that means) and it of course integrates with everything from OpenAI.

You can almost hear the servers humming Ave Maria.

More rants after the messages:

1. Connect with me on Linkedin 🙏
2. Subscribe to TechTonic Shifts to get your daily dose of tech 📰
3. Please comment, like or clap the article. Whatever you fancy.

Power grab with a capital “P”

That AMD deal is a blood pact written in silicon and is about six gigawatts by 2026.

This GPU pact that means “we’re building the world’s biggest brain, and it runs hot”. There are six nuclear-reactor-equivalents of compute required by 2026, starting with a one-gigawatt facility stuffed with MI450 GPUs and stock warrants that could hand OpenAI up to 10 percent of AMD if milestones hit.

AMD gets billions in locked-in revenue and street cred against Nvidia, and OpenAI buys itself the closest thing to compute sovereignty humanity’s ever seen. If you’ve been following the gossip, you know there’s a break-up on the horizon between OpenAI and Microsoft, where OpenAI is renting cloud space, but now they’re building the grid themselves with AMD. The cost are oceans of electricity, cities’ worth of cooling, and the moral hangover of knowing your chatbot now needs a power output that is rivaling Luxembourg just to tell you which ramen joint is closest.

And the freaking cherry on top is 160 million stock warrants at a cent apiece if milestones are met. Yup, compute is the new oil, and Sam Altman just opened the refinery.

From chatbot to operating system

Then came DevDay, the open-bar wedding reception for developers who still believe they’re the bride. Altman strutted onstage and introduced “mini-apps” — now you can connect your Spotify, Canva, Expedia, Zillow directly to your AI. Basically every service that can sell you a moodboard or a mortgage is now trapped inside ChatGPT’s velvet-lined cage.

You don’t use apps anymore because the bot uses you. You ask it to plan a trip, it starts talking Expedia. You mention moving, Zillow talks back. You mumble “logo”, and Canva materializes. No more screens. No more clicks. Just a little chatter between agents while you nod along, pretending you’re still in control. The interface disappears, but the decisions don’t.

They have built you a super-ego, an orchestration layer perched above everything digital that you may want to use and it is turning your entire online life into a one-man puppet show. If it works, ChatGPT becomes the front end of your entire existence.

I wrote about this scenario a couple of times, dating back about a year or so. These are the two latest posts about the direction the frontier AI’s are taking:

1. Macrohard is Musk’s middle finger to Microsoft | LinkedIn
2. Five clues your AI Assistant is planning to replace you | LinkedIn
3. The Web didn’t die, it just got a lobotomy | LinkedIn

History doesn’t repeat, it just updates the firmware

You’d think we would learn as a species. Microsoft built an empire by controlling the desktop. Google made searching a hostage situation. Apple turned the app store into a toll booth with better design. Meta, well, Zucky just monetized loneliness.

And now, here comes OpenAI, wrapping it all in a friendly chat bubble. The new middleman doesn’t show ads or pop-ups (yet), but it just decides for you, silently in the background, which answer you get and which one disappears into the abyss.

Zillow’s bragging about being ChatGPT’s “exclusive real-estate partner”. Yeah, that’s nice ‘n all, but wait until there’s competition. Then we’ll see the bidding wars. “Want to be recommended by the world’s favorite agent? That’ll be $5 per conversation, sweetheart”.

You get my drift?

SEO dies, conversational payola is born.

And there’s no results page to complain about this time, and the bias hides inside the typing cursor.

The illusion of freedom

Their marketing is genius “You don’t have to juggle tabs anymore!” No, you just have to trust a trillion-dollar AI middleman with your judgment. Delegation without transparency — that’s what it is, and that’s not convenience, that’s dependency with extra steps.

When you wanted to find the “best flight” you used to go searching, but now you’re surrendering. What’s their ranking logic, do you know? Neither do I. Whose commission is it chasing? Dunno. What data did it use? No idea. And guess what, they call that magic.

Now here comes Atlas, OpenAI’s bling bling browser, only available on Mac for now, and this is the final brick in the walled garden. Atlas helps you browse, and it decides what you see. It’s the crawler, the judge, and the tour guide, all rolled into one polite surveillance interface. Combined with the “apps inside ChatGPT”, it completes the holy trinity of capture — data, interface, and context.

The browser used to be your window to the web, and now it’s their window into you. With Atlas, they own the conversation, and they own the clicks that used to be yours as well. You will see their footprint expand from your chatbox to your desktop, from your search queries to your subconscious.

We screamed for twenty years about black-box algorithms. Now we’re putting our lives in one.

The irony’s nuclear, with a carbon footprint to match. Literally.

The governance black hole

Every tech shift creates a lag. First comes the platform, then the addiction, then -five years too late — the regulation. We’ve seen this movie before with the App Store tax, the ad-tech cartel, the algorithmic puppeteering. And we’re about to binge the sequel in 4K HDR.

Regulators are too busy arguing about copyright and the AI Act to notice that ChatGPT is becoming the interface for everything.

Once an agent starts making your bookings, recommending your hires, approving your expenses, and paying on your behalf†, it’s game over. Convenience melts into manipulation.

My friend, we are handing the keys to our judgment to a company that already prints money by predicting the next word.

† AI is your new shopping daddy, and Visa wants him to hold your credit card | LinkedIn

What will happen next

OpenAI is building gravity, and everything digital now orbits around it, energy, data, habits, even boredom.

OpenAI is pulling off three moves at once. They’re hoarding energy like a prepper, embedding partners until the internet itself becomes a dependency, and wrapping it all in such sweet UX caramel, that you forget it is actually poison. And Wall Street calls it “vision”, developers call it “the dawn of agentic computing”, but I call it the biggest land grab since Windows 95, but only this time the operating system runs you.

Next, the walls will grow teeth. More apps, more integrations, more ways for ChatGPT to quietly become your only go-to-place for everything. The browser, the desktop, and eventually the OS of reality. Payments will appear natively in chat, because if you can ask the bot to plan your wedding, you may as well let it invoice the florist. Ads will slither in disguised as “recommendations”. And not the clumsy Facebook kind, because these will be highly contextual. You search “Why does it burn when I pee,” and voilà “partnered up with MedFirst Clinic”. And when you schedule a discreet visit, it is sponsored, of course, with a nice little referral bonus for every urethra that monetizes.

Another one, you ask for “affordable vacation ideas”, and ExpediaBot delivers three options, all “AI-verified” deals, all with invisible kickbacks routed through affiliate APIs. And another, you request “help writing a breakup message”, and a subscription trial for an emotional support AI named HeartMend, is added to the text, complete with a coupon for your next therapy session. When you ask “how to fix my sleep schedule”, Spotify recommends a playlist “curated by neuroscience” that just happens to push its premium tier. And when you ask, “Should I quit my job?” and LinkedInAgent gently nudges you toward an “upskilling” course, hosted on Coursera, naturally, and conveniently partnered with OpenAI plus it pushes it’s sponsored job-ads. You know of course they have announced to come out with a “job board”, right?

Then come the productivity apps.

Your AI reads your email, drafts replies, calls your contacts, schedules meetings, and sends invoices before you’ve even had your morning coffee. It’ll share your screen, monitor your clicks, and give helpful suggestions like, “I noticed you’ve been staring at cell D14 for six minutes, would you like me to finish your budget?” Microsoft’s already halfway there, they slipped computer vision into Copilot, so this is not something for the future. And soon ChatGPT will integrate with Office so tightly that Word and PowerPoint will just be skins on ChatGPT. You won’t leave the interface because the interface won’t let you.

I already work that way through Manus AI. I don’t “open” Office anymore, I just speak it into existence.

And then there’s AgentKit which is their automation grenade.

OpenAI’s new framework that lets developers build “autonomous agents” capable of chaining tasks, talking to APIs, and executing workflows without you. It’s Zapier meets RPA with self-awareness, the bridge between chat and full-blown automation. Today, it files your expenses. Tomorrow, it fires your workers. Give it access to your calendar, inbox, Slack, CRM, and Jira, and AgentKit will run your entire workday like a ghost manager who never forgets. The next phase will be agents talking to other agents. Your travel agent will ping your finance agent to approve the budget, who’ll nudge your HR agent to log the time off. You won’t even be CC’d, just “kept in the loop” by a summary you didn’t ask for.

And this ain’t sci-fi, my dear friend. It’s already rolling out. Payments and ads in Atlas, automation via AgentKit, productivity via Copilot, content via ChatGPT apps, it’s the full stack of dependency, top to bottom. A self-contained economy that starts in your chat window and ends in your paycheck. You won’t “use” the web anymore because you’ll live inside it, like a content hamster in a predictive maze.

Every “frictionless” platform ends this way.

With total centralization and it started with good vibes and pastel icons. Every time we think “this one’s different”, nah, it’s not. I’m not anti-tech, I’m quite the opposite, but I just remember the hangovers. The world doesn’t need another operating system pretending to be your buddy, it needs daylight, real transparency, real interoperability, and competition that still bites.

Otherwise, we’ll end up whispering to the same cheerful black box that decides what we know, what we buy, who we love, and which thoughts are still safe to have.

Yup.

That’s our future.

Signing off,

Marco

I build AI by day and warn about it by night. I call it job security. Big Tech keeps inflating its promises, and I just bring the pins and clean up the mess.

👉 Think a friend would enjoy this too? Share the newsletter and let them join the conversation. LinkedIn, Google and the AI engines appreciates your likes by making my articles available to more readers.