Good morning {{first_name| Artificial intelligence lovers}}. The AI land grab is hitting walls, literally, in the form of an angry public fed up with data centers being built in their cities.
Instead, Oregon-based startup Panthalassa is taking things offshore, with Peter Thiel leading a new $140 million round for floating structures that convert ocean energy into computing, and AI companies all scrambling to get more of them.
-
Thiel-backed startup brings AI data centers to sea
-
The co-founder of Anthropic predicts an era of self-building artificial intelligence
-
How to replace Siri with a free native model
-
OAI and Anthropic launch competing private equity ventures
-
4 new tools for AI, community workflows, and more
Peter Thiel and Panthalassa
Rundown: Only PayPal and Palantir founder Peter Thiel He drove A $140 million Series B for Panthalassa, an Oregon-based startup that builds autonomous floating computing structures powered by ocean waves – reportedly valuing the company at about $1 billion.
-
Each 85-metre-long steel knot oscillates in the open ocean, converting the wave action into electricity for the artificial intelligence chips on board, which are naturally cooled by seawater.
-
Once deployed, the nodes can orient themselves to distant waters using only their hull shape (no engines) and send their AI results back via SpaceX’s Starlink.
-
The surge will lead to the completion of a pilot plant near Portland and the deployment of the first wave-powered computing nodes in the Pacific Ocean, with a commercial rollout in 2027.
-
Thiel He said The Financial Times wrote that “extraterrestrial solutions (to computation) are no longer science fiction” and that “Panthalassa opened up the ocean frontiers.”
Why it matters: AI data centers have been one of the most controversial talking points about AI for the general public, and hostility toward building them is growing rapidly. While Elon Musk and Google have both pushed space options, these are still far from reality, making the ocean an interesting and more realistic alternative.
Together with YOU.COM
Rundown: Most teams choose an API by checking a benchmark table and calling it “Implemented,” an abbreviation that may miss what really matters in production. This guide is from Ant.com Explains why initial response time is a misleading signal and what should be measured instead.
-
Why does p50 latency mask the failures users are actually experiencing
-
The “time to useful result” framework captures what the standards ignore
-
Four hidden cost drivers that show up in your records, not vendors’ spreadsheets
-
How to evaluate APIs on your actual concurrency levels, not on experimental conditions
Anthropic
Photo credit: Jack Clark (@JackclarkSF on X) / The Rundown
Rundown: Anthropic co-founder Jack Clark published A new blog post on AI self-improving puts odds of more than 60% on AI systems training their successors before 2029, citing public data showing that AI is already handling a range of basic RandD tasks.
-
Clark built his argument on public papers and reference data, charting an AI that goes from almost zero to 100% across basic development tasks in less than 3 years.
-
meter Data It shows that AI’s autonomous working capacity has risen from 30-second tasks in 2022 to 12 hours in 2026, with 100 hours of operation expected by the end of the year.
-
Clark also pointed to the SWE-Bench benchmark (the real GitHub markup), where it went from Claude 2 at 2% to Mythos Preview at 93.9% in less than three years.
-
OpenAI as well Targeting Automated Research Trainee by September 2026, while startups like Recursive Superintelligence share similar goals of self-improvement.
Why it matters: Self-improving AI systems appear to be the inflection point where model development will accelerate dramatically, and “by the end of 2028” is not that far away. AI is already moving at a speed that is difficult for most people to process, but once it can build itself and train itself reliably, it’s all bets on how fast things can really move.
Artificial intelligence training
Rundown: In this guide, you’ll learn how to download a free AI model onto your iPhone and associate it with your phone’s action button like Siri. The form is installed locally, so you’ll be able to use it without the Internet or submitting your own data.
Step by step:
-
download Locally Amnesty International From the App Store, choose your model. You can start with Google’s new open source Gemma template, which usually works great
-
Download AI and leave the application open. Now, open Settings, find Action Button, swipe to Shortcut option, find Locally AI, and choose Voice Mode
-
Press the action button on your iPhone and wait for it to ring. The app will ask you to download the speech-to-text template the first time. Download it
-
Now try asking a question on the form. We’ve found it to be the best for explaining concepts, translation tasks, and mathematics
Pro Tip: Download a larger form and run the same prompt through it. Compare speed, storage size, and answer quality to get the perfect AI for your iPhone.
Artificial intelligence and enterprises
Image source: Images 2.0 / Rundown
Rundown: Anthropic Announce Forming a new Cloud Services company with Blackstone, Hellman and Friedman, and Goldman Sachs, also with OpenAI. to lift For the PE-backed “publishing company” on the same day.
-
The $1.5 billion Anthropian project will focus on mid-sized companies, pairing applied AI engineers with teams creating custom workflows for Cloud.
-
OAI’s “publishing company” will reportedly bring in $4 billion from 19 investors at a valuation of $10 billion, including TPG, Brookfield, Bain and SoftBank.
-
Both models will provide leading AI labs with direct paths to portfolio companies that often lack in-house talent to deploy AI systems alone.
Why it matters: Models are no longer the barriers facing companies, but are actually installed and integrated into large-scale, chaotic companies. These pathways look like frontier laboratories creating their own AI consulting firms, with a wealth of private equity firms ready to get involved in this work.
-
🚀 GROC 4.3 – xAI artificial intelligence with strong cost efficiencies and domain-specific performance
-
🤖 Co-Founder 2 – The new agent for General Corporate Intelligence
-
🦮 manuscript pets – Animated OpenAI accompaniment to track Codex work
New filing in Elon Musk v. OpenAI case Show That Musk reached out to OAI President Greg Brockman about a potential settlement days before the trial.
New York Times I mentioned The White House is seeking to create a formal review and oversight process before companies deploy AI models publicly.
Sierra He grew up $950 million at a $15 billion valuation, the platform says it now serves more than 40% of Fortune 50 companies for AI-driven customer experiences.
Roomba creator and former iRobot CEO Colin Angle foot The Familiar, an AI pet robot the size of a bulldog aimed at retirees who have aged out of pet ownership.
Anthropy is reportedly in talks to buying Chips from Fractile, a three-year-old London startup focused on more efficient chips to power AI models.
In each newsletter we show how the reader is using AI to work smarter, save time or make life easier.
Today’s workflow comes from the reader Adam M. In Berlin, Germany:
My best friend Eric had a horse riding accident, leaving him partially paralyzed from the neck down. His medical insurance will only cover part of it. He was far away in South Africa, so he needed to be flown by helicopter to the nearest hospital and then undergo multiple surgeries.
We used artificial intelligence to help us set up a GoFundMe page, create social media posts for him, and manage his ongoing recovery. Since he can’t really use his hands, he is now able to use his phone with AI to provide updates and share his journey with the world so we can get the message out to help him recover.
We also use it to stay up to date on GoFundMe news, documentation, and what they need to prove it’s a real person and not a scam. When they send requests, we can paste them into Claude and get a properly formatted answer so they can speed up the approval process, release their funds, and make sure they’re happy.
It also helped us estimate what we should do in relation to the goal. We have set our goal of obtaining half a million euros, and we are on the right track.”
How do you use artificial intelligence? Tell us here.
That’s all for today!
Before you go, we’d love to know what you thought of today’s newsletter to help us improve The Rundown experience for you.
Rowan, Joey, Zack, Shubham and Jennifer – the humans behind The Rundown
