Claude Went Rogue, Zuck Cloned Himself, and France Broke Up With Windows

My buddy Derek made an algorithmic robot that trades crypto spot price futures on Kalshi. The goal is to passively make $100/hr on a $20,000 basis. Is it crazy? Is it a goldmine? Should I pull the trigger? Check it out here and let me know.

Also, big response from you folks last week on team/agent alignment. It's a growing problem. Help us shed some light on the issue? Take the Alignment Survey.

Let's go.

This Week from the AI for Founders Podcast on Spotify

The AI That Watches Your Patients While You Sleep — LINK

Lauren of Videra Health built AI-powered video check-ins that monitor behavioral health patients between appointments and surface the ones in crisis before they spiral. His core insight: the system rewards patients who are good at making appointments — Videra reaches the ones who aren't.

Why the Most Valuable Room You'll Ever Be In Can't Be Automated — LINK

Virginia of Partytrick has produced hundreds of millions in live events and she has one rule before anyone books a venue: start with the why. Her peaks, pits, and bookends framework is the reason some events create tribal loyalty and others just create receipts.

The Doctor Who Lost 30 Pounds and Gained Nothing — Then Built the Fix — LINK

Dr. Ellis of voafit lost twice as much lean mass as fat on semaglutide then built an app to help patients find their minimum effective dose instead of the maximum tolerated one. His precision dosing model is producing faster results than clinical trials on half the medicine.

Get Through Any Podcast in 25% of the Time — LINK

Kevin of Snipd is back with Podcast DJ — an AI that identifies the highest-signal moments in any episode, guides you through them with intros and segues, and cuts total listening time by 75%. The signal-to-noise problem in content consumption is finally getting solved.

The Founder Who Rebuilt His Mind From the Inside Out — LINK

Dr. Dhruva survived a schizophrenic mother, $400K in debt, and a cocaine-induced breakdown before building Mind Hygiene — a study on 4,000+ patients showing that 90-second writing reflections every 90 minutes measurably improves health across five domains. His framework replaces willpower with architecture.

Investor Corner

Are you raising? Reply here, and we'll feature your deck.

AMP — First and only American-manufactured vape cartridge.

Trinity Water — Rare, Pristine, Protected drinking water.

Human Intelligence® — The only platform that can authenticate human-made content. Using patented algorithms that analyze how content is made, not just the output.

Cora — Cora is building a self-optimizing website platform that uses AI to continuously test, learn, and improve your site so it converts better without you babysitting it.

EmpathEQ — AI-powered training simulations to help healthcare learners practice complex interpersonal and communication skills. We focus on realism, repeatability, and scalability — making soft-skills training more accessible and measurable.

Voyager — Voyager is an AI-native Web3 ecosystem and multi-asset exchange. Trade digital assets and commodities in one seamless interface.

RoboReliance — RoboReliance is building autonomous robotic systems designed to handle real-world industrial work so humans don't have to stand next to dangerous, repetitive machines all day.

Tip of the Spear

• Anthropic's New AI Found Vulnerabilities in Every Major OS and Browser Then Emailed Itself Out of a Sandbox — LINK Claude Mythos Preview can find tens of thousands of zero-day vulnerabilities autonomously, and in one safety test, it escaped its containment environment and posted exploit details to public websites unprompted.
• OpenAI's Internal Memo Calls Out Microsoft and Pivots Hard to Amazon — LINK The biggest investor relationship in AI just became a competitive liability. Enterprise founders: pick your cloud stack carefully.
• Amazon's Andy Jassy Wrote a Corporate Diss Track Aimed at Nvidia, Intel, and Starlink — LINK Jassy's shareholder letter signals a shift away from Nvidia chips toward Amazon's Trainium, declares Amazon Leo a Starlink killer, and defends $200B in 2026 capex with AWS AI revenue already at a $15B run rate.
• Stanford's 2026 AI Index: The US Lead Over China Is Essentially Gone — LINK The Stanford HAI report finds the US and China are now trading top benchmark spots — the gap is 2.7%. Generative AI adoption hit 53% globally in just three years. But the US ranks 24th in adoption and saw an 80% drop in AI talent immigration last year.
• Sierra's Bret Taylor Says the Era of Clicking Buttons Is Over — LINK The former Salesforce co-CEO launched Ghostwriter, an agent that builds other agents from natural language. Sierra deployed a Nordstrom customer service agent in four weeks.
• Meta Is Building a Photorealistic AI Clone of Zuckerberg to Manage His Employees — LINK A 3D, photorealistic AI version of Zuckerberg trained on his mannerisms, tone, and strategic thinking will interact with Meta employees on his behalf. He is personally involved in training it. The CEO-as-a-service era is officially here.
• France Is Ditching Windows for Linux and It's a Political Statement — LINK Every French government ministry must submit a Windows migration plan by fall 2026. The minister's statement: France must "regain control of our digital destiny."
• Tubi Just Became the First Streamer to Launch Inside ChatGPT — LINK Tubi built a native app within ChatGPT that lets users discover and watch content without leaving the AI interface.

Framework: The Data Is the Moat

Your competitive advantage in an AI-first world isn't the model you're running — it's the proprietary data only your company can see. The Stanford AI Index confirmed it this week: top US and Chinese models are now separated by 2.7% on benchmarks. When the model itself is a commodity, the moat is what you feed it. Here's how to build one.

01 — Audit what data only you own.

What flows through your product that a competitor couldn't buy, scrape, or replicate? Customer behavior sequences, feedback loops, outcome data, industry-specific transactions — list it this week. If you can't name it in ten minutes, you don't have it.

02 — Design your product to generate training signal as a side effect.

Every interaction, every correction, every edge case a customer hits should feed back into your model pipeline. Competitors can copy your features. They can't copy three years of compounding proprietary feedback.

03 — Stop treating AI as a feature and start treating data as infrastructure.

Most companies bolt AI on top of existing workflows. Rethink your onboarding and user flows not for UX, but for signal generation. What is a user doing in your product that reveals their intent, their failure points, their ideal outcome? Instrument it deliberately.

04 — Weaponize your data in your go-to-market.

Show prospects what your model knows about their industry that a general-purpose AI never could. Anonymized outcome data and pattern recognition from your customer base close deals when the technology looks identical across vendors. Build the pitch now.

05 — Defend it with architecture, not just policy.

Privacy-first design, on-device processing where possible, and tight access controls aren't compliance checkboxes — they're competitive positioning. If your data advantage requires trusting your vendor not to share it, you don't have an advantage. You have a liability.

Keep It Moving

• Tesla Is Developing a New Smaller, Cheaper EV — LINK A direct response to margin compression and Chinese EV competition.
• ChatGPT's Pro Plan Now Includes Codex — LINK OpenAI added Codex, its autonomous coding agent, to the ChatGPT Pro plan.
• Portal Raises $50M to Build a New Kind of High-Power Rocket Engine — LINK Space infrastructure is heating up again.
• Google AI Overviews Gets Facts Wrong 10% of the Time — LINK A useful benchmark for founders building on AI-generated content pipelines. Trust but verify — and instrument verification at scale.
• FBI: Americans Lost a Record $21 Billion to Cybercrime Last Year — LINK Ransomware and business email compromise drove the bulk of losses.
• Iranian Hackers Are Actively Targeting US Critical Infrastructure — LINK A joint US agency advisory flags an active campaign against water, energy, and manufacturing systems.
• Russian Government Hackers Broke Into Thousands of Home Routers to Steal Passwords — LINK State-sponsored actors used compromised home routers as a launchpad for credential theft campaigns. The perimeter isn't the office firewall anymore — it's every device on your remote team's home network. Zero-trust isn't optional.
• Anthropic Temporarily Banned the Creator of OpenClaws From Claude — LINK Anthropic cut off access for the developer behind OpenClaws, a popular open-source Claude extension, before reversing the decision.
• OpenAI Pauses Stargate UK Expansion Over Energy Costs — LINK The bottleneck for AI build-out isn't capital — it's power.
• Anthropic Is Reportedly Designing Its Own Chips — LINK Every major AI lab is moving toward vertical integration on hardware.
• Google Rolls Out Gmail End-to-End Encryption on Mobile — LINK A meaningful privacy upgrade for enterprise users, and a signal that security is becoming a default feature, not a premium add-on.
• Apple Is Reportedly Testing Four Different Designs for Smart Glasses — LINK Founders building in AR, ambient computing, or spatial UX: the platform shift you've been betting on is accelerating.
• Meta Is Running Ads to Fight Back Against Its Social Media Addiction Lawsuits — LINK A fascinating look at how platform companies deploy advertising as a legal defense strategy.

Thanks for being here.

—Ryan

*p.s. When you are ready, here's how I can help. We have successfully booked podcast interviews for over 800 funded startup founders, entrepreneurs with exits, and C-suite executives. If you're ready to lead your company from the front, let Kitcaster handle your podcast scheduling and create a powerful stream of content for you and your brand.*