Follow me for:

📰 News about AI, vibe coding and entrepreneurship.

💡 Tips on how to scale software companies. 

👀 A front-row seat to my building process.

Théo Canale

Jeff Bezos is buying companies that AI is about to kill.

His new venture, Project Prometheus, raised $6.2 billion at a $30 billion valuation. Now it is seeking tens of billions more from sovereign wealth funds and JPMorgan to acquire industrial companies that Bezos believes will be disrupted by AI.

The thesis: manufacturing, aerospace, and automotive companies will be gutted by AI-driven automation. Most will fail to adapt. Prometheus buys them, rebuilds them AI-native, and captures the upside.

Bezos is Co-CEO. His first operational role since leaving Amazon in 2021.

Think about what this means. The person who built the most efficient logistics operation in history looked at the industrial economy and decided: I can do this again, but with AI this time.

He is not investing in AI tools. He is buying the companies those tools will transform. That is a fundamentally different bet.

While most VCs chase the next model or the next wrapper app, Bezos is acquiring the atoms. The factories. The supply chains. The physical infrastructure that AI will eventually run.

If you are a CEO in manufacturing or industrial services, you just got a new competitor. And this one has unlimited capital and a 30-year track record of winning.

AI Advisor & Trainer | Buying AI tools is easy. Knowing what to do next isn’t.

$130 an hour.

No degree. No office. No laptop. Just a hard hat, the right certification, and a willingness to live in a temporary housing compound next to a data center construction site.

The AI boom has a labor shortage that nobody in tech talks about.

It is not engineers. It is electricians.

Bloomberg reported this week that "man camps" are popping up across the US to house data center construction workers. Free steaks. Golf simulators. Anything to keep skilled tradespeople from leaving for the next site.

Numbers are insane. The US needs 300,000 new electricians in the next three years. Training a qualified electrician takes four to five years. Math doesn't work.

Google published a policy report saying the electrician shortage "may constrain America's ability to build AI infrastructure." Microsoft's president named it the number one problem slowing their expansion.

Construction firms are pulling workers off oil pipelines at double their previous rate. The mobile housing market is now estimated at $700 billion.

You might want to reconsider your career goals, the most in-demand workers in the AI economy now hold a journeyman card and a pair of wire strippers.

undefined undefined

Andrick Mueller

In France, €10k/month as an entrepreneur gets you fewer flat offers than €2K /month on a payslip. Not a joke. 

What's harder than building a profitable company in France? Convincing a landlord you can pay your rent ! 😬 

I’ve literally seen tax returns treated like finger paintings compared to a basic employment letter. The French system doesn’t care about your wealth; it cares about your predictability.

Once you actually get the keys, you’re practically the new owner.

You could stop paying, start a sweat shop in the living room, and it would still take two years and a miracle to get you evicted. 🤣

Everywhere else, the deal is simple: Cash gets you the keys, and no cash gets you the sidewalk by Tuesday. It’s harsh, but it makes sense. 

Here, we’ve turned renting into a high-stakes marriage where the landlord is terrified of the divorce.

France is the only country where being broke but "stable" is a better flex than actually earning money.

Sam Altman held a talk last week. Headlines focused on the $110B round. That was not the take. 

Three signals deserve your attention instead. 

First: Altman described a Jevons Paradox for AI. As AI gets cheaper per unit, total usage explodes. Companies budgeting for 'AI savings' are about to discover their overall AI spend is growing. Costs per query fall. Total queries multiply. The bill goes up.

Second: OpenAI's biggest constraint is go-to-market. They build faster than enterprises can adopt. The gap between model capability and actual corporate usage widens every quarter. If you feel behind on AI, so does everyone else. Including OpenAI's sales team.

Third: memory is the real switching cost. The more context an AI accumulates about your business, the harder it becomes to leave. OpenAI is building lock-in through accumulated knowledge, one conversation at a time. That is stickier than any API pricing.

If you run AI strategy, pay attention to the Jevons point. Your board thinks AI will reduce costs. The data shows AI increases total spend while delivering more value per dollar.

Both things are true simultaneously. Most budgets only account for one.

I spend my days testing AI tools that replace the work of a three-person team. Yet, I remain optimistic!


If you don't feel a slight knot in your stomach watching AI evolve, you're simply not following the same news as I am.

Jobs will disappear, entire sectors of the economy are heading straight for a wall. I don't consider this "pessimism", but rather it's realism.

Sometimes I close my laptop with a sudden urge to go raise goats in the French countryside. 😅

And yet…

The very next morning, I find myself excited about a new feature, or imagining how this technology could save precious time for an entrepreneur on the verge of burnout.

My brain is a conflict zone between a dark clarity about what's happening to us and an almost naïve enthusiasm.

It's exhausting being what I call an "enthusiastic pessimist."

But deep down, it might be the only healthy stance today: acknowledging that things are going to get bumpy, while keeping your hands on the wheel.

Even if I'm often concerned about the future, I've never wanted to shape it more than I do right now. 💪

Everyone yells “AI revolution.” Data says the opposite. 

The St. Louis Fed tracks real-time AI adoption. Their finding: daily use of AI for work has barely moved. The curve is flat. No inflection point. No hockey stick. 

37% of US workers use generative AI. But intensity of use? Stable. Not accelerating. 

Meanwhile: 
→ AI capex is 2% of GDP ($650 billion) 
→ 2,800 data centers planned in the US 
→ AI-adjacent commodities up 65% since January 2023. 

Massive investment. Modest adoption. 

This is the gap nobody wants to talk about. 

Companies are spending like it's a revolution. Workers are using it like it's a better search engine. 

It's an S-curve: early adoption is slow and expensive. Growth accelerates later. Saturation eventually sets in. 

We're still in the slow part. 

The real risk isn't that AI will replace your job tomorrow. It's that your company spends millions on AI tools nobody actually uses. 

I see this constantly with the executives I work with. 

The problem isn't technology. It's adoption, integration, and change management.

"When a measure becomes a target, it ceases to be a good measure."

Charles Goodhart wrote that in 1975. And it explains most corporate AI programs in 2026.

Look at how companies measure AI success. Tools deployed. Employees using copilots. Time saved per task. Hours of training completed. These numbers look clean on a board slide.

They also miss the only question that matters: did the business get better?

Companies optimize for the metric, then confuse the metric with the outcome.

They deploy 15 AI tools, report '85% adoption,' and wonder why margins stayed flat.

The problem runs deep. AI vendors sell on activity metrics because outcomes take years to measure. Executives buy on activity metrics because they need a number for the quarterly review. 

Both sides have incentives to track the wrong thing.

The companies getting real value from AI measure revenue per employee. Margin improvement. Win rates on deals. Time to ship product. They count business outcomes. AI tools... never. 

So if your AI dashboard looks impressive but your P&L doesn't, you're probably measuring the wrong thing. Goodhart figured this out 50 years ago.

We are making a historic mistake by restricting free speech.

I'm French, and I've always disagreed with the way we handle this. For years now, we have put the right not to be offended ahead of the right to speak.

Fake News law, Avia law, Digital Services Act… 

I just keeps coming. 

The debate has been muzzled to protect the ego of people who don't have the stomach for direct contradiction.

We build moral barriers everywhere, convinced that silence equals progress.

For me, that is a complete illusion.

Banning a racist or conspiracy theorist from speaking doesn't change his mind.

It just sends him into a basement with ten other people who think exactly like him. There, in the dark, with zero exposure to opposing ideas, his views don't disappear. They harden.

That is exactly the Obama vs Trump syndrome: during the Obama years, the people in power congratulated themselves for building a progressive paradise, because they couldn't hear any dissenting voices anymore.

In 2026, America wakes up with 30% of the population they no longer understand, because they spent 20 years cutting off their microphone instead of dismantling their arguments.

You don't fight an idea by making it illegal. You fight it by proving it is weak.

By sanitizing debate, we are not building a more tolerant society. We are just engineering an explosion that nobody will see coming.

It is long past time for France & EU to wake up.

No alternative text description for this image

What happens when your biggest investor is also your biggest competitor?

OpenAI is about to find out.

This week, it was reported that OpenAI is building a code-hosting platform to rival GitHub. 

The trigger: repeated GitHub outages that disrupted OpenAI's engineering workflow.

GitHub belongs to Microsoft. Microsoft has invested billions in OpenAI. Microsoft integrates OpenAI's models across its entire product line.

And now OpenAI is building a product to take developers away from Microsoft's platform.

This is not a small side project. Code hosting is infrastructure. Developers build habits around their repository. Switching costs are high. If OpenAI ships a serious GitHub alternative with native AI integration, enterprise teams will evaluate it.

Microsoft will notice.

The broader pattern matters more than this single product. OpenAI has been quietly expanding from an AI model provider into a full platform company. Agents, enterprise tools, consumer apps. Each new product puts OpenAI closer to competing with one of its partners.

The same dynamic played out between Google and Samsung. Between Apple and Intel. Between Amazon and its marketplace sellers. When a partner becomes a platform, the relationship changes.

The question for every startup relying on a strategic investor: how long before your roadmap collides with theirs?

For OpenAI and Microsoft, the collision just started.

Everyone's calling it the SaaSpocalypse.

$285 billion wiped in four weeks. Thomson Reuters worst trading day on record. HubSpot down 39%. Figma down 40%. Atlassian down 35%.

All because Claude Cowork launched at $20/month on January 12.

The math is brutal: Cowork does roughly 40% of what enterprise seats at $150/month do. At $20. 

And the gap narrows every week.

This isn't just a pricing problem. This is an architecture problem.

Cowork doesn't replace a tool. It reads your files, spins up parallel workers, connects to Slack, Notion, Jira, HubSpot, 50+ tools, and executes entire workflows on its own.

The people extracting real value from it have stopped thinking of it as software. 

They think on an outcome. And walk away. 

The SaaS companies that expect to survive by bolting AI onto their existing features and UI, will have a hard awakening. 

The ones who will thrive, will be the ones who realize their entire value proposition has to be rebuilt around AI. Not just updated. Rebuilt.

Thomson Reuters sees this and that's why they're sitting on an $11 billion acquisition war chest.

When I sit down with C-suite executives to discuss AI strategy, I usually lay out three scenarios based on probability:

➡️ Low: AI is just another tech cycle and and a gradual evolution.
➡️ Medium: AI isn't "superhuman," but it’s a game-changer that shifts market dynamics fast enough to wipe out established players.
➡️ High: AI achieves super-intelligence. I tell my clients bluntly: "If this happens, my consulting is useless and your company likely won't survive.

This is typically where I lose them. The more "seasoned" the leader, the more they refuse to look at scenarios 2 and 3. 

I’ve heard it a hundred times: "I’ve been in business for 30 years, I’ve seen it all, and my business is healthy."

But there are two things these leaders choose to ignore:

1. AI is not human-led. Every previous industrial shift moved at a human pace. AI can generate infinite technological revolutions autonomously, at a speed completely decoupled from our ability to upskill.

2. "Healthy" is relative. We look at a giant like L’Oréal and see $450k in revenue per employee. Then you look at the new world: Airbnb is at $1.5M, and YouTube is at $8M. Legacy companies are often highly cash-inefficient and entirely dependent on US-built infrastructure they don't control.

The "I've seen it all" mindset is a survival mechanism, but in an era of exponential velocity, it’s a terminal liability.

Past success is the loudest whisper telling you that you don't need to change, right until the moment you realize you're obsolete.

The most dangerous thing you can bring to a revolution is a track record of surviving the last one.

A year ago, Suno was a party trick.

"Oh cool, I made a song." People played with it for 10 minutes, then went back to Spotify.

Today: $300 million ARR. 2 million paid subscribers. Growing faster than Spotify did at the same stage.
​
The music industry sued them in 2024. Labels assumed that would slow things down. It didn't. Revenue kept climbing while the courtrooms filled up.
​
Here's what nobody in the industry wanted to say out loud:

People don't just want to listen to music. A lot of them want to make it. They just couldn't. The barrier was a decade of lessons, a band, a studio, $10,000 in equipment.

Suno removed the barrier. The demand was already there.

This is the pattern most incumbents miss. They look at a new tool and ask: "Is it as good as what we have?" Wrong.

The right question: what does this unlock that was never possible before?

Spotify unlocked music for listeners. Suno unlocks music for everyone who ever hummed a melody in the shower and thought, "I wish I could do something with that."

That's not a niche. That's most people.

The labels are fighting the wrong war. They're defending a distribution model in a world where creation just became free.
​
The real threat isn't that Suno replaces Drake. It's that Drake's audience become creators too.

taplio

The Battle

Théo Canale

Iker Tenes

Followers

Posts

Engagements

Likes

Performance ScoreThéo Canale's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Théo Canale

Iker Tenes

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Théo Canale's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreThéo Canale's LinkedIn Performance