Meta is building data centers in tents, OpenAI says chat is dead, and Anthropic is hacking China.
Hi,
Anthropic just released Claude Fable 5.
Which means I'm about to give Claude a task, go to lunch, take a meeting, forget about it entirely, and return to discover the machine has already done the work, checked the work, questioned the work, and written itself a performance review.
We've been telling AI what to do every 14 seconds like an anxious parent at a middle school dance. "Did you do it?" "How about now?" "What about now?"
Fable 5 appears to have looked up from the keyboard and politely suggested we go touch grass.
It's not that it's smarter. It's duration. It also takes 2x the usage of Opus 🙄
Let's go.
This Week from the AI for Founders Podcast on Spotify
The salesperson of the future never forgets you — LINK
JP Grace of Endear built an AI opportunity engine that surfaces the five best customer outreach moments for each associate every morning, pre-drafts the message, and cuts friction to seconds. Early results: 6x more outreach in six weeks and a 35x return on delivered messages across 2,000 stores.
She had no maternity leave, so she rewired how she worked and accidentally proved her own worth — LINK
Ashley Gross of AI Workforce Alliance used AI to claw back time with her newborn, then discovered she was the "human Google" her company couldn't function without. Three months after handing over her playbook, her team hit a $25 million pipeline overachievement. Then came the newsletter, the Maven course, and the leap.
The inbox is dead and the people emailing hardest are the ones killing it fastest — LINK
David Connors of The Swarm sold his recruiting startup to Sequoia, walked out with a thesis that trust is the only currency AI can't fake, and built the relationship graph LinkedIn will never open up. Warm intros convert 10 to 20x better than cold. His agents now run that motion in ten minutes a week.
Nina (trynina.co) creates SEO-optimized content that ranks on search engines and gets cited by LLMs like ChatGPT, driving the kind of organic traffic that actually converts.
Investor Corner
Are you raising? Reply here, and we'll feature your deck.
Dining Out — DiningOut unites restaurants, food lovers, industry suppliers, and brands through a self-reinforcing ecosystem of live events, targeted media, and proprietary technology.
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 — 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, with a focus on realism, repeatability, and scalability.
Voyager — An AI-native Web3 ecosystem and multi-asset exchange. Trade digital assets and commodities in one seamless interface.
RoboReliance — 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
Microsoft breaks up with OpenAI, launches seven in-house AI models — LINK
At Build 2026, Microsoft unveiled MAI-Thinking-1 and MAI-Code-1-Flash, its first proprietary reasoning and coding models, built to run on Azure and cut payments to third-party providers. Satya Nadella called it a shift from consuming frontier models to fully participating at the frontier, which is a polite way of saying they are done writing OpenAI blank checks.
SpaceX files the largest IPO in history at $75 billion — LINK
Elon Musk is taking SpaceX public at $135 per share, targeting a $1.77 trillion valuation on Nasdaq under the ticker SPCX, with debut expected June 12. He retains 82.4% of voting control. He is going public without actually giving anything up.
Anthropic is helping the NSA hack China. It also wants everyone to slow down. — LINK
The NSA is reportedly deploying Anthropic's Mythos model for offensive cyber operations, with Anthropic engineers embedded at the agency. This is the same week Anthropic published a report calling for a global AI development slowdown. The gap between what AI companies say and what they do is now wide enough to fit a cyberweapon through.
Google cuts a $920 million monthly check to SpaceX for compute — LINK
Google has committed $920 million per month to SpaceX for compute capacity, citing unexpected demand from recently launched AI products. A search company paying a rocket company nearly a billion dollars a month for servers is the sentence that best summarizes what AI has done to the infrastructure economy.
OpenAI says chat is dead, superapp is coming — LINK
OpenAI employees are openly saying the chat interface is finished and the company is accelerating toward a full superapp handling tasks, commerce, and daily workflows in a single product. This is no longer a model company. It is competing with Apple, Google, and every SaaS tool your team currently runs on.
Meta is building AI data centers in tents — LINK
Meta has erected six 125,000-square-foot fabric structures in New Albany, Ohio to house AI servers, cutting deployment time from three years to three months. One data center tracker called it the AI race entering its "Mad Max phase," which is also a reasonable description of 2026 generally.
Apple is revamping Siri at WWDC this week — LINK
WWDC 2026 kicks off today with Apple expected to finally deliver the Siri overhaul and Apple Intelligence upgrades that underwhelmed last year. Tim Cook is under more pressure than at any WWDC in recent memory, with every AI competitor having shipped meaningfully while Apple has been quiet.
Framework: Agents Need Management, Not Mythology
Enterprise agents need permissions, controls, logs, and oversight. Treat them like employees.
Everyone is deploying agents. Almost nobody is managing them. This week, Uber burned through its entire annual AI budget in four months and had to cap it. That is not an agent problem. That is a management problem wearing an agent costume.
01. Give every agent a role, a scope, and an owner. Write down what each agent is authorized to do, which systems it can touch, and who is accountable when it acts outside those boundaries. Without those three answers, you do not have an agent. You have an unmanaged process with a chat interface attached.
02. Set spend and action limits before you deploy, not after. Every agent should have a hard ceiling on token consumption, API calls, and third-party costs, plus a kill switch that requires a human decision to override. Default conservative. Raise limits deliberately, based on performance data only.
03. Log everything before you automate anything. If you cannot see what your agent did, when, and why, you cannot improve it, defend it, or debug it. Build the logging layer first. It is not a compliance afterthought. It is the foundation of a manageable agent operation.
04. Stage permissions from read to write. Start every new agent in observation mode. Only after you understand its behavior patterns and failure modes do you grant it the ability to act. The cost of an agent reading the wrong thing is almost always lower than the cost of an agent writing the wrong thing.
05. Design escalation in from day one. The best agent systems assume the agent will hit its limits. Build explicit handoff paths so that when it does, the human gets a context-rich summary and a clear next step. A well-designed escalation is faster than a human doing the whole task from scratch.
The mythology of agents is that they run themselves. The reality is that they run as well as the management structure you build around them.
Apply today — Agentic partnership acquisition for the world's top podcasts. gethowdi.com
Keep It Moving
Uber caps employee AI spending after burning through the full-year budget in four months — LINK
Uber encouraged staff to use AI as much as possible, hit its annual limit before summer, and has now introduced individual spending caps. Letting employees run free on AI subscriptions is not an adoption strategy. It is a budget liability with good PR.
Meta's WhatsApp Business AI agent goes global — LINK
Meta has made its AI agent for WhatsApp Business available worldwide, with businesses charged on a per-token basis. If your customers are on WhatsApp, which they are if you sell to anyone outside the US, this is now a real infrastructure decision.
AI agents are now buying things and fraud looks identical — LINK
Agentic AI traffic grew 7,851% in 2025 according to HUMAN Security, and the gap between legitimate bot commerce and fraud has shrunk to less than half a percentage point. Your fraud detection systems were not built for a world where the buyer is a script.
Microsoft's Scout personal assistant launches at Build — LINK
Microsoft launched Scout, an OpenClaw-inspired personal AI assistant built into Microsoft 365, at its Build developer conference. It is the clearest signal yet that Microsoft is competing in the daily productivity layer, not just the model layer.
Bots now outnumber humans on the web — LINK
Cloudflare data shows bots account for 57.4% of all web traffic, with AI training crawlers alone at 50.6%. Every analytics dashboard, ad campaign, and SEO strategy your company runs was built on the assumption that the audience is human. That assumption is now factually wrong.
Google rolls out fake call detection to fight AI deepfake scams — LINK
Google is rolling out on-device detection for AI-spoofed calls impersonating trusted contacts, employers, and authority figures. Deepfake voice scams are now common enough that Google built a native defense into Android. Worth flagging to any team that handles sensitive communications by phone.
Amazon faces class action over Ring's facial recognition — LINK
A class action filed in Seattle claims Ring's Familiar Faces feature stores facial images of passersby without consent. If you are building any product that touches biometric data in shared spaces, this lawsuit is the clearest preview of the legal risk you are assuming.
Meta's AI app has hidden facial recognition code — LINK
Wired found unreleased facial recognition code inside Meta's AI app, previously flagged as a candidate feature for its smart glasses. The pattern of shipping biometric features first and asking for permission later is becoming a legal liability fast.
Anthropic proposes a global AI development slowdown — LINK
Anthropic published a report warning that AI systems will soon be capable of building their own successors without human involvement, calling on companies to slow development. This is the same week they were reported to be embedded at the NSA running offensive cyber operations. The public and private positions are not aligned.
Chinese spies are using LinkedIn to recruit western sources — LINK
A US government advisory warns that Chinese intelligence is using job platforms to contact people with access to non-public information and build information-sharing relationships over time. If your team has any exposure to defense, AI research, or critical infrastructure, this is a real operational security briefing.
IBM accused of covering up multiple data breaches — LINK
A former cybersecurity executive has filed a whistleblower lawsuit accusing IBM and two subsidiaries of concealing breaches from the mid-2010s. If you are evaluating enterprise vendors for sensitive data infrastructure, ask this question about every incumbent.
Companies are rehiring the workers they laid off for AI — LINK
Fast Company is calling it the "AI boomerang." Managers who cut headcount expecting AI to fill the gap are discovering the technology cannot fully replace humans after all and are quietly calling people back. The automation thesis is getting stress-tested in real time.
ChatGPT is citing Reddit 24x more than at the start of the year — LINK
Foundation's analysis shows ChatGPT is adding "reddit" to its query fanouts 24 times more often than in January, pushing Reddit back to the top of AI citation sources. If your content strategy does not include Reddit community presence, you are not in the rooms where AI models answer questions about your space.
Google Maps now asks users whether reviews were paid for — LINK
Google Maps is surveying users to identify incentivized reviews, with data showing over 1,000 single-month deletions and retroactive audits going back to 2024. If your reputation strategy has ever included review incentives, this is a live enforcement issue, not a future risk.
Google publishes new guidance on third-party SEO tools — LINK
Google updated its SEO documentation to assert itself as the canonical source for AEO and GEO advice, warning that third-party tools have no access to internal ranking data and cannot guarantee results. If your agency is making promises about AI search rankings, Google just told you exactly how to push back.
Bootstrapped Lectric e-bikes grew while VC-backed rivals went bankrupt — LINK
While funded e-bike startups collapsed, Lectric scaled profitably without venture capital and has since launched three new brands. Capital efficiency is still a competitive advantage.
Snipd is the world's best podcast streaming app — try premium for free as a sub from this newsletter.
What I'm Thinking About
Before you send another cold email. — Clip
Uncover the AI tools here you didn't ask for, but now you're weirdly into.
Thanks for being here.
-Ryan
p.s. I write about Revops, Product, and Founder-led marketing on LinkedIn, X, and my blog.
