Good morning. It’s Friday, November 13. GitHub became the place where developers build software together. What happens to it if AI agents write most of the code?

Plus today’s top stories: What Jensen Huang’s resignation means for NVIDIA’s open AI models, and why the Fed is watching AI’s effect on jobs.

And in today’s Friday Fact: the math puzzle that took 800 processors to solve.

→ Make enterprise content more useful at every step with Box

A lot can go wrong in seven hours and 47 minutes.

On August 17, GitHub suffered a worldwide outage that took down repositories, pull requests, GitHub Actions, APIs, authentication and Copilot. If you were planning to ship software that Monday, there was a decent chance you did something else instead.

Apparently, a critical infrastructure component failed to scale when traffic hit a new peak. GitHub knew it was bad. "We let you down," the company wrote afterward. Things were about to get worse, though.

In the middle of GitHub's very bad day, Cursor released Origin, an early-beta code-hosting platform featuring repositories, pull requests, code browsing, integrations with deployment & integration tools, and GitHub syncing.

In other words, Cursor released a starter pack of the things developers have spent the better part of two decades relying on GitHub for. The timing was almost comically bad.

→ Read the full story.

A simple-sounding puzzle sat unsolved for decades, and when the answer finally came, it arrived with a catch no mathematician had seen before. More at the bottom.

NVIDIA co-founder Jensen Huang resigned as chief executive and left the board after 33 years at the company. Chief Financial Officer Colette Kress is serving as interim CEO while the board searches inside and outside NVIDIA for a permanent replacement; the abrupt transition sent the stock sharply lower.

Federal Reserve Governor Michael Barr said society should prepare for the possibility of serious short-term labor disruption as AI spreads through the economy. He expects the technology to raise productivity over time, but warned that workers could feel the pain before the gains arrive.

Anthropic CEO Dario Amodei told prospective investors that AI could make cancer, Alzheimer’s disease, heart disease and mental illness relics of the past within a decade. Medical experts say AI can accelerate research, but complete cures on that timeline remain unsupported by clinical evidence.

New from Box: Reimagining business processes with content and security at the core

→ Make Enterprise Content More Useful at Every Step With Box

Colette Kress has stepped in as NVIDIA’s interim CEO after Jensen Huang’s sudden resignation. She joined NVIDIA as chief financial officer in 2013 after senior finance roles at Cisco, and has helped manage the chipmaker through its rise from a gaming company to the central supplier of the AI boom.

Kress now has to steady employees and investors while the board searches for a permanent leader with both technical authority and global political reach. Her first public test comes with NVIDIA’s earnings report on November 17, when the company will have to show that its strategy can survive without the only CEO it had ever known. → Read the full report.

Jensen Huang made open models a major part of NVIDIA’s strategy, arguing that freely available systems would expand the market for the company’s chips. NVIDIA has released model families for agents, robotics and physical AI, backed open-source infrastructure and invested heavily in Hugging Face.

That commitment now faces a leadership test. A new CEO could keep the strategy because it drives demand across the ecosystem, or narrow it to focus on NVIDIA’s most profitable products. → Read the full article.

AI agents spot possible Maya site: 48 Google AI agents scanned NASA LiDAR data and flagged an unmapped site in Mexico now under investigation.

SpaceX plans $74B for compute campuses: SpaceX said it plans to invest billions in supercomputing campuses at Starbase, Texas, and Arizona.

Notion AI chats exposed in vendor breach: A Fireworks AI incident exposed Insights chat transcripts, affecting under 1% of workspaces.

Most production model swaps take weeks: 69% of firms that replaced a production AI model spent weeks on it; 36% needed over a month.

Fake products poison AI shopping advice: Chinese regulators penalize firms flooding the web with fake rankings to sway chatbot recommendations.

AI is breaking consulting’s billable-hour model: Firms are testing outcome-based pricing and reshaping graduate roles as AI automates junior work.

In May 2016, computer scientists Marijn Heule, Oliver Kullmann and Victor Marek settled a question Ronald Graham had put a $100 bounty on in the 1980s. Can you color every whole number red or blue so that no Pythagorean triple, like 3, 4, 5, is all one color? You can, up to 7,824.

At 7,825, it becomes impossible. Proving that took 800 processors running for two days on the University of Texas's Stampede supercomputer. The resulting proof was 200 terabytes, roughly the size of all the digitized text in the Library of Congress.

The team compressed it to 68 gigabytes so others could check it with an independent program. The answer is trusted because one machine verified another.

Here's how you voted: Opinions were split, with 47% believing NVIDIA is too big to fail, while 45% disagreed.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team