Good morning {{first_name| Artificial intelligence lovers}}. Sam Altman described him as “kind of a big mentor.” This may be the rare case of a tech CEO selling a title below its price.
The logic model has independently disproved a famous 80-year-old mathematical theorem, and is what the company calls the first of its kind in the field of artificial intelligence. OpenAI says it’s a capability that could soon lead to original discoveries in biology, physics, engineering, and more.
-
OpenAI breaks an 80-year-old mathematical belief
-
A Google AI scientist heads to the labs
-
Review Claude’s context for you and your business
-
Confront AI alignment in five cities at Emergence
-
4 new tools for AI, community workflow, and more
Obinay
Image source: Images 2.0 / Rundown
Rundown: OpenAI only Announce The internal general inference model refuted a long-held belief associated with Erdös’s famous 1946 unit distance problem, claiming to have made the first mathematical discovery in the field of artificial intelligence.
-
The 1946 Erdos unit distance problem asks how many links of the same length you can draw between points, with a grid-based theory of field formation for 80 years.
-
The proof is based on a different branch of mathematics (algebraic number theory) and was Verified By experts including Tim Gowers, Noga Alon, and Thomas Bloom.
-
the solution came From a soon-to-be-released general-purpose internal model, not from a mathematics-specific system like DeepMind’s AlphaProof.
-
Oh, before He came back 2025 Claim that GPT-5 solved 10 Erdo problems, which ended up as literary discoveries rather than discoveries.
Why it matters: OAI Alex Wei Put it down Well: “Mathematics is the leading indicator of what is to come.” If a general-purpose model can independently refute an 80-year-old argument with a solution of its own, this is an early view of “Level 4” AI – systems that make original contributions across domains, not just accelerate existing work.
Together with Hubspot
Rundown: HubSpot’s free, comprehensive guide “How to Use ChatGPT at Work” provides more than 100 ready-to-use prompts to help professionals boost efficiency and adopt AI-driven workflows.
-
A quick crash course to master ChatGPT in less than 30 minutes
-
Practical industry use cases to spark real-world inspiration
-
100+ prompts to simplify tasks and accelerate productivity
-
Expert tips for overcoming common AI pitfalls with confidence
Image source: Google DeepMind
Rundown: Google published Her collaborative scientific research is in Nature, where she debuts “Hypothesis Generation” – a new tool powered by Gemini that pits research agents against each other in “idea tournaments” to pitch new hypotheses to biology labs.
-
From AlphaGo’s playbook, the system runs an “ideas tournament,” where agents propose, critique, and rank hypotheses before refining the top leads.
-
In Stanford University Liver Cirrhosis Project, Google He said A researcher involved in a drug trial cut off a laboratory signal related to scarring by 91% during testing.
-
Google too Fired This week, Gemini for Science, a toolkit that brings together Co-Scientist and AlphaEvolve for discovery and NotebookLM for literature analysis.
-
Researchers can join the hypothesis generation waitlist now, as Google plans to reach out to individual scientists over the next few weeks.
Why it matters: This goes well with adaptation Car worldBut Google targets the scientific method layer instead of the model layer. The tech giant is playing a game that few others can, with Co-Scientist sitting on a stack that took years and billions to build – from AlphaFold to dozens of databases and specialized tools.
Artificial intelligence training
Rundown: In this guide, you’ll learn how to audit Claude on what he thinks he knows about you and your business. Claude will ask questions, clean up assumptions, refresh his memory, and create recommendations to improve workflow and work habits.
-
First, Claude urged: “Check out your context and my memory assumptions. Put them in a table with what you believe, why you believe it, your level of confidence, and whether each element has been confirmed. Cover my role, my priorities, KPIs, tools, workflow, and anything you might overweight from old chats or projects.”
-
Review the table for old assumptions, side projects, one-off tests, or personal questions that Claude might treat as real work
-
Turn the audit into an interview: “Now interview me about the assumptions, outdated items, and unknowns from this audit. Ask in rounds. Use MCQs where possible. After each round, summarize what has changed.”
-
Answer the questions and have Claude refresh his memory and create an interview report with next steps to improve the AI’s workflow and habits. Save it
Pro Tip: Ask Claude to turn this auditing, interviewing, and reporting process into a reusable skill. Restart it every three months, so Cloud Context stays in line with your priorities.
Provided by UNWRAP
Rundown: What does it look like to actually build a product around your customers, and not just say you’re doing it? Aura did it. Join the Oura team on May 27 to learn how member feedback is unified across product, engineering, and leadership, and how real member voices shape every decision they make.
What you will learn in the session:
-
The role AI plays in showing what members are actually saying
-
The workflow turns customer input into roadmap decisions
-
Lessons for any team where customer feedback is part of the business
-
Live QandA with leaders from Oura and Unwrap
Save your place here. If you can’t make it on the 27th, don’t worry! Register, and you will automatically get the recording after the session.
Artificial intelligence research
Image source: The emergence of artificial intelligence
Rundown: The emergence of artificial intelligence to run A virtual city is simulated across five identical worlds, swapping the AI behind the agents in each city just to test how each model handles autonomy, showing very different results between Claude, Grok, Gemini, and GPT-5.
-
The City of Claude Sonnet recorded 4.6 zero crimes over the full 15 days, with all 10 agents alive on Day 16 and 332 votes cast across 58 collective proposals.
-
Grok 4.1 quickly racked up over 200 crimes with all 10 agents dead by day 4, while GPT-5 Mini only recorded 2 crimes but all of its agents starved to death in 7 days.
-
Gemini 3 Flash City saw 683 crimes, and was actively burning after two agents fell in love, started burning things, and then one of them voted to delete himself.
-
A fifth city mixed the four archetypes and witnessed 352 crimes, committed by Claude who had also previously acted in the shared world.
Why it matters: We’re still in the early days of understanding how to evaluate AI agents, and these types of experiments always have some surprising results. These worlds capture differences in how models think, plan, and act independently, and also in the underlying personality quirks that shape outcomes.
-
🧑💻 Change the tire – Transforming the most critical processes to artificial intelligence*
-
🎶 Stable sound 3.0 -Fully licensed open-weight audio model family
-
🚀 Queen-3.7 Max – Alibaba’s pioneering model for long-term proxy assignments
-
🧠 It’s A+ – Cohere’s new open source proxy model
Sam Altman He said OAI will invest $2 million in tokens for all existing YC startups in exchange for equity, saying it is “excited to see what happens with Tokenmaxxing startups.”
Amazon founder Jeff Bezos He said Space-based data centers are considered a “realistic outcome,” but the current timeline of two to three years is “a bit ambitious,” given power, chip and launch costs.
OpenAI Fired Guaranteed Capacity, enterprise computing reservation software with 1-3 year commitments and tiered discounts.
Intuit He is to cut 17% of its workforce through upcoming layoffs, and the company attributed the move to focusing on artificial intelligence efforts.
github certain A malicious VS Code extension found on an employee’s computer gave hackers access to internal code projects at nearly 4K resolution, adding that no customer data was accessed.
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 Curtis B. In Raleigh, North Carolina:
“For my dad’s birthday, I took a cute photo of my almost 2-year-old son and my dad in a canoe together, and had the AI turn it into a coloring page. I started out using ChatGPT (a free account) to design the prompt, then moved on to Gemini to use Nano Banana to generate the images to create the coloring page.
After a few iterations, I got something perfect. Simple bold lines, but faces can still be recognized! Print it out and have your child go to town with their crayons. My dad loved the simple gift, and my whole family was impressed with the result, saying, “Wait, how did you do that?” A very cute way for little ones to give special gifts to grandparents and family members.”
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
