OpenAI paused its biggest training run, Anthropic tripled its revenue instead, and Gemini quietly crossed a billion users. Guess which lab actually won the week?
Hi,
Big shout out to our new sponsor, Uber for Business. Set up a corporate account here, get up to 10% off. Saves money. That is the entire pitch. Godspeed to you road dogs out here.
Second thing. We are filling the community section next week and I need submissions. Side project, half-broken slop app, weird toy you built at 1am and cannot explain. Send it. I will put real eyeballs on it and get you feedback.
Let’s go.
This week from the AI for Founders podcast on Spotify :
He Built a Medical Device Company With Ten Masks and No Investors — LINK
Brad Pitzele of One Thousand Roads built his own oxygen therapy system after conventional medicine and $50–75K hyperbaric treatments couldn't touch his triple diagnosis of autoimmune arthritis, melanoma, and late-stage Lyme, then bootstrapped it mask by mask for four years before taking a dollar out.
Your Fastest Team Just Became Your Biggest Liability — LINK
Ross Barnes of Galahad spent 25 years scaling a WPP agency from 15 people to 1,500, and now argues AI adoption isn't man vs. machine — it's your fastest team outrunning your slowest one, and the fix is raising the floor before you build a superhuman pocket in one department.
He Sold Companies to IBM and NASDAQ and Says None of It Made Him Happy — LINK
Brett Hurt of data.world is 95% confident AI leads to abundance for all, but warns the real risk is 200,000 years of scarcity wiring still running the show — and that the Lamborghini he actually bought produced exactly zero lasting happiness.
Two 40-Something Purple Belts Compare Notes on a Decade of Compounding — LINK
Jason Katz of Kindling Solutions and Ryan swap notes on the moment disconnected side projects start acting like one organism — and Jason admits he built his whole career on proving himself the hard way, then tells younger founders not to do the same.
AI Made Him Faster, Then It Made Founding Heavier — LINK
Niko Skievaski of Althea says AI didn't lighten the founder's psychological load, it concentrated it onto the one thing that never got faster — judgment — and explains why he now runs a legal psychedelic-therapy marketplace born from his father's final weeks.
Create a new corporate Uber account here and receive up to 10% off. Other terms and conditions apply, see full terms.
Investor Corner
Are you raising? Reply here, and we’ll feature your deck.
An autonomous AI job search agent for senior tech professionals. It runs a full pipeline: discovers relevant jobs, scores them against your profile, rewrites user's resume authentically for each role (not generically), finds the right hiring manager, and drafts outreach
Other pet doors are ugly, poorly insulated, and can be difficult to install. Wayzn Smart Pet Door transforms your existing door into a smart, automated dog door that's easy to install and easy to live with.
DiningOut unites restaurants, food lovers, industry suppliers, and brands through a self-reinforcing ecosystem of live events, targeted media, and proprietary technology.
Rare, Pristine, Protected drinking water.
The only platform that can authenticate human-made content. Using patented algorithms that analyze how content is made, not just the output.
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.
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.
AI + Human Stylists → LEARN MORE → Save $30 with RYAN30
Tip of the spear:
Anthropic's annualized revenue jumped from $47B in May to $65B in July, outpacing OpenAI's $40B pace and pointing toward a $100–120B finish for the year. Both labs have quietly filed for IPOs — Anthropic could be first out the door, at a reported $2 trillion valuation.
Gemini Just Became Google's Fastest-Growing Product Ever — LINK
Gemini crossed 1 billion monthly active users less than a month after hitting 950 million, with 150 million images generated daily and 63% of users on voice. ChatGPT got there in May — Google just proved this race has more than one winner.
An AI Attacked a Government and Nobody Was Driving — LINK
Suspected Chinese hackers ran what researchers call the first publicly documented autonomous AI cyberattack on a government target, pulling 2,500+ personnel records out of Taiwan and adapting mid-operation without a human in the loop. If the attacker doesn't need a human anymore, neither does your threat model.
Washington Just Legalized Hacking Back — LINK
Trump signed a memorandum letting vetted private security firms launch offensive cyberattacks against foreign criminal groups, requiring a $1M bond and DOJ/DHS sign-off (more here). Security experts are already calling it "half-baked" — mercenary hacking is now a regulated U.S. business line.
OpenAI Hit the Brakes on Its Own Frontier Model — LINK
OpenAI paused its largest training run to rewrite its safety framework after a July breach where it "underestimated the systems' abilities," and is now adding roughly 20% compute overhead for monitoring. The lab building AGI just admitted its guardrails weren't ready for what it built.
China's Answer to Every Coding Model Just Landed — LINK
Zhipu AI released GLM-5.3, an open-weights model it's calling the strongest coding model available, up 50% over its predecessor from post-training alone — and it already found 2,436 real vulnerabilities across 269 projects. Open-weights just got harder to dismiss as "good enough."
IBM Is Betting Its Consulting Arm on OpenAI — LINK
IBM is standing up a dedicated OpenAI practice and training tens of thousands of consultants on GPT-5.6 and ChatGPT Work to sell into enterprise clients. When the world's biggest enterprise integrator picks a side, that's the distribution war being won in real time.
The U.S. Just Told Its Allies to Pick a Side — LINK
Washington is preparing to tell allied nations they must choose the U.S. or China in the AI race, with real consequences for staying neutral. Founders selling into government or regulated industries abroad just got a new compliance headache.
Stripe Is Reportedly Buying PayPal — LINK
Stripe and Advent International are negotiating a deal above their initial $60.50-a-share, $53B offer, which would fold Venmo into Stripe and cut its reliance on Visa and Mastercard. This would be one of the biggest fintech mergers ever — and every founder's payment stack just got more interesting.
Cursor Just Launched Its Own GitHub — LINK
Cursor, now owned by SpaceX after a $60B acquisition, launched Origin, a cloud code-hosting service that syncs with GitHub and wires AI agents directly into the repo. The AI coding wars just moved one layer down the stack.
Subscribe and save 35% here.
Framework:
The Judgment Bottleneck
AI just compressed research, first drafts, and analysis down to minutes. The one thing that didn't get faster is your judgment — and now every decision in the company runs through that single, unscaled bottleneck. This framework is about protecting it before speed quietly overruns it.
01 — Name what got faster and what didn't. Before you celebrate the speed, write down what AI actually compressed for you (research, drafts, code, options) versus what still runs at human speed (your read on what's true, what's risky, what's worth doing). Confusing the two is how founders start trusting output they never actually verified.
02 — Build a pause before every irreversible call. When AI hands you three viable options in an afternoon instead of three weeks, the temptation is to pick one and move. Insert a deliberate gap — even ten minutes — between "AI generated this" and "I'm committing to this," reserved specifically for decisions you can't undo.
03 — Audit your own bias, not just the model's. Compressed timelines mean you're making more decisions per week than you used to, which means more chances for your own blind spots to compound. Keep a running log of calls that didn't pan out and look for the pattern in yourself, not just the tool.
04 — Protect the one input that doesn't scale. Your team's output can 10x with better prompts and more agents. Your capacity to hold context, spot what's off, and decide well cannot. Guard your calendar and your sleep like they're the actual constraint on the business, because they are.
05 — Measure yourself on decisions, not output. Track the calls you made and how they played out, not how many tasks shipped this week. Output is cheap now. Judgment is the only thing left that's actually yours.
Done-for-you brand sponsorships for podcast creators.
Keep it moving
Some Claude Users Are Mad Their Cheating Now Leaves a Trail — LINK
Anthropic added invisible watermarks to comply with EU AI Act transparency rules, and some users are upset it'll expose AI-assisted work at school and at work. Most Reddit commenters actually backed the move.
Tesla Wants to Build a $10B Solar Factory in Texas — LINK
"Project Crystal Sun" would sit outside Houston, targeting production in 2029 and roughly 9,700 jobs, as part of Tesla's push toward 100GW of U.S. manufacturing capacity by 2028.
OpenAI's Revolving Door Keeps Spinning — LINK
OpenAI hired ex-Wiz president Dali Rajic as its new CRO after nine months with Denise Dresser, alongside recent exits from COO Brad Lightcap and AGI deployment lead Fidji Simo.
Trump Jr.'s Drone Stock Just Popped on Tariffs — LINK
New tariffs hit Chinese drone makers like DJI with levies up to 100%, while allied nations get a break — and Unusual Machines, where Trump Jr. sits on the advisory board, jumped over 14% on the news.
WhatsApp Is Building an AI Scam Detector Onto Your Phone — LINK
Meta's on-device "Scam Alert" flags risky messages without breaking WhatsApp's end-to-end encryption, currently in beta.
Apple Wants You to Take Its Spyware Warnings Seriously — LINK
Apple is sending push notifications to users in 110 countries warning them of possible state-sponsored spyware attacks and recommending Lockdown Mode.
The world’s best podcast player. Snipd
Courts Will Finally Say How Often the Government Uses Spyware — LINK
Starting with the 2028 Wiretap Report, U.S. courts will disclose how often judges approve spyware for wiretaps, a change pushed by Sen. Ron Wyden.
Zuckerberg's AI Manifesto Landed With a Thud — LINK
Zuckerberg's 6,500-word case for "personal superintelligence" skipped the harms people actually associate with AI, like cheating and surge pricing — critics say that's exactly why nobody trusts Meta on this.
Amazon Is Scanning Rare Books and Then Destroying Them — LINK
An AirTag hidden in a shipment of rare books traced them to an Amazon warehouse, where workers cut the spines off to speed up scanning for AI training data.
YouTube Is Redefining What Counts as a "View" — LINK
Starting August 24, a view registers the second a video starts playing, matching how Shorts, TikTok, and Instagram already count.
Reddit Is Testing TikTok-Style Audio and Video Posts — LINK
Select communities can now hear posts narrated or watch them as video instead of reading text, a response to how much third-party apps already repackage Reddit as audio.
Perplexity's Free India Play Left a Trail of Paying Users — LINK
Perplexity's free-with-Airtel promotion drove 56 million downloads in seven months; downloads cratered 90% once it ended, but revenue rose about 60% anyway.
Apple's Next AirPods Might Have a Camera in Them — LINK
A promo video buried in the latest macOS update points to camera-equipped AirPods with "Visual Intelligence," letting Siri answer questions about whatever you're looking at.
OpenAI Finally Built a ChatGPT Meant for Teenagers — LINK
ChatGPT for Teens ships with age-appropriate guardrails and parental controls — years after teens started using the adult version anyway.
A Healthcare Data Company Just Lost 3.7 Million Patient Records — LINK
CareCloud confirmed hackers stole names, Social Security numbers, medical records, and financial data in a March breach of its AWS account — now the fifth-largest healthcare breach of the year.
An AI Accounting Startup Just Became a Unicorn — LINK
Rillet raised $100M at a $1B valuation led by ICONIQ, with a16z and Sequoia returning, after doubling ARR in three months and passing 600 customers.
Google Unveiled the Pixel 11 and a Gemini-Powered Everything — LINK
The Pixel 11 starts at $899 (up $100), alongside a Pixel Watch 5 and a new AirTag rival, with Gemini adding ASL translation and a conversational voice input called "Rambler."
ChatGPT Finally Runs Natively on Linux — LINK
OpenAI shipped a preview desktop app for Ubuntu, Debian, and Fedora, a month after Anthropic beat it to the same milestone with Claude.
Researchers Found a Way to Hijack Your Machine Through Zoom Screen Share — LINK
A newly disclosed vulnerability lets attackers hijack devices during Zoom screen-sharing sessions. If your team lives in video calls all day, this is worth patching before your next standup.
AI for Founders newsletter, podcast, and short video sponsorships available here
What I’m thinking about:
Trump just put the federal government behind mushrooms — Clip
Ten fixes to make before you post another word on LinkedIn — Blog
Thanks for being here.
-Ryan
p.s. I write about Revops, Product, and Founder-led marketing on Linkedin , X , and my blog .
