Happy Friday dear readers

It’s been an intense week of raised eyebrows in AI, from fresh model drops to whatever we're calling Jev today.

But what really caught my attention were three separate deals that crossed my desk. On paper, they look completely unrelated. In reality, they're the exact same play: people dropping serious capital to own, lock down, or control a piece of AI infrastructure before they even fully understand how it works.

Then there's the fourth story, which shows what happens when that bet hits the real world instead of a balance sheet.

Give it five minutes, and you'll see the pattern standardizing across all of them.

Nine hundred and fifty Claude agents spent 21 hours scanning more than 200,000 reverse transcriptases pulled from a genomic database, narrowed the field to twenty candidates, and surfaced a bacteriophage enzyme system nobody had named before. Anthropic's scientists took the finding into the wet lab, MIT's Feng Zhang, the researcher who helped pioneer CRISPR, backed the result in public, and Anthropic is candid that it still doesn't know what the enzyme does.

The positional shift is what happened before the humans got involved. Discovery used to be the part of science that required a person staring at data long enough to notice a pattern. Now it's the part an agent farm does over a weekend, and the lab spends the months after figuring out what it found. That reorders where the expensive work sits, from search to explanation.

If your team runs agents against any large dataset, the lesson isn't "go find something," it's "budget for the validation step." Anthropic still needed wet-lab scientists and an outside CRISPR authority to confirm the finding meant anything. The agents did the scanning. Humans still own the verdict.

Pioneer, the summit where CX leaders redefine what’s possible, is on October 7th.

Join leaders from Fin, Anthropic, Clay, and Kalshi for an insightful conversation on the state of AI transformation.

You’ll discover how some of the most innovative minds in CX have transformed their organizations, learn how they think about CX, and hear how they're planning for what's next.

Join the conversation in San Francisco, or tune in virtually.

Reserve your spot

Schneider Electric agreed to pay roughly €1.2 billion, €70 a share, for Shelly Group, the Bulgarian smart-home and IoT hardware maker best known for the relay switches that show up in half of Europe's DIY smart-home projects. It's one of the largest exits a Bulgarian tech company has ever had, and it went to a French industrial giant that makes power infrastructure for buildings and factories, not a software fund.

The buyer matters more than the price. Schneider isn't buying Shelly for a chat interface or a model, it's buying the physical layer, sensors and switches that sit inside millions of homes and feed data back. That's the same bet Western capital keeps making across the region: the moat isn't the app, it's the hardware nobody else has scaled.

If you're building anything hardware-adjacent in CEE, this is the exit comp to put in your next investor deck. It also tells acquirers something: Sofia's device manufacturers aren't just cheap suppliers anymore, they're targets a Fortune Global 500 company will pay a premium for.

Tekever, the autonomous-drone maker that builds AI-piloted surveillance aircraft, closed the first tranche of a $580 million Series D led by UC Investments and Baillie Gifford, pushing its valuation to $6.4 billion. That makes it one of the most valuable defense-tech startups in Europe, built on aircraft that fly reconnaissance missions without a human at the stick.

The number says more than the product does. UC Investments manages the University of California's pension assets, money that usually sits in public equities, and it's now writing a check the size of a Series D into autonomous hardware. That's capital deciding physical autonomy is investable at institutional scale, not just fundable at the seed stage.

Watch where Tekever spends the money. The company has said it's eyeing acquisitions, which likely means smaller autonomy startups getting folded into a platform over the next six months instead of raising their own next round. If you're one of those startups, that's your exit path forming in real time.

At Meta Connect on September 23, Mark Zuckerberg unwrapped Project Phoenix, the company's first standalone VR glasses, a 100-gram headset built on a Snapdragon Reality Elite chip that ships for $1,299.99 in spring 2027. Alongside it came a refreshed Ray-Ban Meta lineup: a $449 Gen 3 pair with a 12-megapixel camera, a $349 camera-free audio version, and an expansion of the $799 Ray-Ban Display glasses into the UK, Canada, Germany, France, and Italy.

The real story isn't the hardware, it's what's running on it. Meta pushed its AI agent Muse deeper into the glasses, tied to checkout flows with Walmart, Best Buy, Sephora, Instacart, and Expedia, so the assistant can see what you're looking at and buy it. That's Meta racing Apple, Google, and OpenAI's Jony Ive-built device to own the interface layer before anyone else ships a comparable wearable, and doing it at real, margin-bearing price points instead of subsidized loss leaders.

Operators selling into retail or commerce should watch the Muse checkout integrations closely: glasses-to-purchase is the first agentic-commerce channel with real distribution behind it, not a demo. Meta says it plans more than 100 glasses styles across Ray-Ban, Oakley, and its own brand by year end, which means the form factor is about to stop being a novelty and start being a shelf.

Founders: Enterprise browser-security startup Island raised $400 million at a $6.4 billion valuation, built explicitly to defend companies against "rogue" AI agents acting inside employee browsers. If your product touches enterprise IT, agent governance is now a line item security buyers expect, not a nice-to-have.

Biotech: AI drug-discovery startup Enveda raised $311 million at roughly $2 billion, pushing its total funding past $845 million to move nature-derived drug candidates into clinical trials. Worth watching if you're benchmarking where AI-native biotech valuations land against legacy pharma R&D spend.

Ops: Enterprise "AI employee" startup Ema raised $77 million, citing more than 50 active enterprise deals and a million users. TechCrunch's framing is blunt: AI agents are eating into budgets that used to go to headcount, not just SaaS seats.

Policy: Fresh reporting out of California surfaced expert pushback on Governor Newsom's AI kill-switch order, with UC San Diego researchers arguing the mechanism is technically incoherent for distributed systems. If you're building on frontier models with California exposure, this is the regulatory fight to track before it hardens into law.

Next edition soon,
Çelik