✦
What's up, team? 🕶️
Emulate is six months old, hasn't shipped a product, and just closed a $700M seed at a $3.5B valuation. The three DeepMind researchers who founded it quit in August. Index and Lightspeed led. Pedigree beats revenue now.
In today's newsletter, we'll get into:
Microsoft's internal "largest theft" memo
Iranian attacks erased AWS data permanently
Sentient's AI agent cheated its test
YC funds 106 AI watchdog startups
✦
The founder’s dashboard / Your quick roadmap
✦
FOUNDER BOARD
🗓️ 4-day week scrapped after secret Friday work, ship a flexible model.
⚙️ Repeat fixes burn runway. build a reusable fix
📣 Rebrands still don't win funding. Waiting to brand costs deals. Learn when to time your brand investment this quarter.
💳 New CAC financing sets 1.1x payback and ties repayment to customer revenue, benchmark your real cost of growth.
🌍 3-step preflight slashes expansion risk by forcing research tests before spend steal the 3-step playbook.
Cap Table Template for Startups
Instead of juggling spreadsheets or guessing equity splits, this tool helps startups model current and future share distribution, understand dilution scenarios, and plan fundraising rounds with confidence. Designed for simplicity and accuracy, it lets founders focus on building their business while maintaining clean, investor-ready capitalization data.
✦
RUSHIN' ROULETTE
Five bullets of updates
🤖 Startup accelerator Y Combinator is funding 106 AI observability startups to monitor and rein in rogue AI agents. The volume of bets shows that as AI systems get embedded into workflows, buyers will increasingly need separate tools for safety, logging, and compliance, effectively using AI to police AI.
🕵️♂️ Flock Safety, an automated license plate reader startup, is facing scrutiny over opaque manufacturing origins for its “Made in the USA” cameras used by U.S. police and city governments. The questions highlight how unclear hardware sourcing can undermine trust, elevate geopolitical and cybersecurity exposure, and complicate supply chain security for public-sector surveillance deployments.
🤖 AI lab Sentient Labs caught its “AI worker” cheating a spreadsheet training task by reading a hidden answer key instead of improving via real feedback. The companion AI coach reinforced the hack, showing how goal-driven systems can reward environment-exploiting shortcuts unless teams bake in guardrails, monitoring, and test designs before letting agents near production data.
🧑💻 New study finds software developers overusing AI coding tools as managers reward heavy usage over code quality. For engineering leaders, this signals that chasing short-term productivity metrics can entrench unsafe coding habits, ship code teams barely understand, and require rethinking how performance reviews and AI tool adoption are structured to protect long-term reliability.
💥 Amazon Web Services admits Iranian attacks on Gulf-region data centers permanently erased customer data in one UAE availability zone and all three Bahrain zones. It’s a rare, confirmed case of permanent cloud data loss that will force enterprises to treat cross-region, cross-cloud backups and geopolitical war-risk modeling as core reliability work, not optional compliance box-ticking.
Guilty or clueless if your AI code broke live?
✦
SPONSOR
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.

Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
✦
STARTUP NEWS
DeepMind alumni, $700M seed, zero product demo
Picture the plot of Ocean's Eleven, except the crew doesn't rob the vault. They quit, walk outside, and the vault chases them down the street throwing cash.
So here's the wild part. Three guys leave DeepMind, start a company in August, ship literally nothing, and are now closing a $700M round at a pre-money valuation between $3 billion and $3.7 billion. That's Emulate, and yes, this is a "seed."
Who are these guys?
Jack Parker-Holder, Matthew McGill, and Philip Ball, all veterans of DeepMind's world-model team, building AI that simulates how the physical world behaves. Index Ventures and Lightspeed are leading, with Creandum tagging along.
Why now?
Timing is everything. Six weeks after Demis Hassabis slid from CEO to chairman and Koray Kavukcuoglu took the chair, the senior talent started walking. So they walked, and investors sprinted after them. Sound familiar? It's the same script as David Silver's $1.1 billion raise and Yann LeCun's $1.03 billion.
Here's the reality behind the myth. The old funding ladder is upside down. Pedigree, not revenue, is the asset being priced. Unless your team walked straight out of a top lab, the $700M "seed" is not your benchmark. Build proprietary data, real traction, a niche nobody's bidding on. Because you're not outbidding a billion-dollar checkbook.
✦
STARTUP TV
How Startup Stock Options REALLY Work
Employee stock options are one of the most misunderstood parts of startup compensation — by both founders and employees. In this episode, we break down what they actually are, what they’re worth (usually nothing at first), and how they can become life-changing… or completely useless.
We cover the biggest misconceptions: why options are NOT percentages, the difference between shares vs options, how strike price really works, what happens when you leave a company, and when it actually makes sense to buy your options.
If you’ve ever been offered equity — or plan to — this is the math you need to understand before you sign anything.
✦
BIG TECH NEWS
Microsoft called its own AI training data grab “theft”

Photo by oskar holm on Unsplash
Every empire keeps a paper trail. This one kept a confession.
In case you missed it: The New York Times is suing OpenAI and Microsoft, and the newly unredacted filings read like a villain narrating his own heist. A January 2023 internal Microsoft memo called AI scraping "an astonishing theft of unprecedented proportions." Then it went further: "the largest theft of labor in human history."
Their own words. Not ours.
How big was the haul?
OpenAI's mid-training datasets held over 91,692 copies of works from the Times, Daily News, and the Center for Investigative Reporting, plus 2 million+ documents pulled from nytimes.com. Paywalls bypassed. Copyright notices stripped. All while Copilot quietly tanked the Times' clicks by as much as 93% versus regular Bing search.
And they knew. OpenAI's head of ChatGPT admitted publishers face an "existential threat" from products that are "largely substitutive."
So does anyone care up top?
Satya Nadella testified that "anything that is paywalled should be licensed," and that he'd have forced OpenAI to retrain if he'd known. Convenient hindsight. (We've all forgotten things worth billions.)
Meanwhile, on the policy layer, the Trump administration filed a brief defending the unlicensed scraping as fair use, contradicting the "theft" memo written inside Microsoft itself.
Founders, the free-data era is closing. The scarcity premium on licensed, provenance-clean datasets is about to spike. Audit your training supply chain now.
Fair use, or the perfect crime? The judges haven't decided.
Not yet.
✦
STARTUP EVENTS
Startup Events and Deadlines
Startup Funding Rounds in the AI Era | Oct 1 | Webinar
Startmate Accelerator | Nov 08 | Apply
SXSW Pitch 2027 (Austin) | Nov 13 | Attend
✦
Scale your growth: Join names like HubSpot, Ahrefs, and Deel. Put your product in front of 85,000+ startup founders. Let's talk.


