I went down to the San Francisco Marina with Julia Rosenberg of Show Me Your Stack to see the Blue Angels for Fleet Week. The Blue Angels are the most elite and skilled precision aerobatic pilots in the world and they fly full time for the United States Navy flight demonstration squadron.

(Just as an aside: how cool is it that the best combat aviators in the world fly full time in a non-combat role??)

It was a special experience.

As the show was ending, one of the Super Hornets flew directly over our heads at 500 mph after completing a loop-de-loop (I don’t know the technical term). We shrieked like little kids, giddy with awe.

I immediately turned to Julia and said, “Wow. We need witness excellence more frequently.” Julia nodded her head, mouth still agape. Watching four human beings fly 18 inches apart from each other at hundreds of miles an hour left us amazed.

When the show ended, we went back to Julia’s apartment to debrief with Jon and Summer, who stayed behind but still caught the show from a safer viewing distance.

We were all like: that was electric.

I’m moved by displays of excellence. Excellence means someone really cares. In fact, excellence means they care more than anyone else in the world. I’m inspired by anyone willing to set a standard so high that it is literally jaw dropping to witness.

It made me think: excellence in your life is a gift to others. If you hold yourself to a high enough standard, your mere existence can be inspiring.

The Blue Angels are so good at their jobs, so exceptional at their craft, that they are unforgettable. 

Care so much that you become unforgettable.

Frontier Tech Roundup — SheFi Edition 🏇

• An independent researcher says Claude Code helped him identify a possible planet in NASA’s public data. Pavel Rabtsevich used it to analyze observations from the TESS space telescope and found a star that dimmed slightly every 3.18 days, which could be caused by a planet passing in front of it. Claude helped him write the analysis code and test alternative explanations, while Rabtsevich decided what to investigate and repeatedly tried to prove himself wrong. The planet still needs to be confirmed, but TESS has approved his request for follow-up observations. I love that this came from an independent researcher working with public data. It's a reminder that if you're passionate about something, you have more tools now than ever to pursue it.

• OpenAI says its own AI helped identify an employee accused of secretly recording executives. Federal prosecutors say staff found a recorder concealed behind a television in an executive conference room on September 30. OpenAI says it used its models to analyze the device and help trace where it came from before turning the investigation over to the FBI. Prosecutors also allege that a chatbot account associated with the employee’s spouse had been used to research how to build and conceal a recorder. OpenAI says it found no indication that customer data or its underlying models were accessed. The employee was arrested on October 9 and has been charged, not convicted.

• Anthropic is adding a rule against sustained cruelty toward Claude. Its updated usage policy, which takes effect November 12, prohibits extreme and repeated abuse of Claude when there is no discernible purpose for it. People can still criticize the model or get frustrated with it, and Claude can already end conversations when abuse becomes persistent. Last week I wrote about Anthropic meeting with religious scholars to discuss whether AI could ever have moral status. Anthropic is not saying Claude is conscious or capable of suffering, but it is starting to establish rules for how humans should behave toward it anyway. I think this is a big deal. How we choose to treat AI may end up shaping how we think about these systems more broadly.

• Coinbase trained a smaller AI model on its own fraud data and says it outperformed Claude Opus 4.5 for that specific job. The company adapted an existing model using examples of transaction behavior and known fraud outcomes, then tested it on its own fraud-detection benchmark. The training run cost under $100, although that does not include the cost of preparing the data or building the system. The more interesting part to me is what this says about where AI is going inside companies. Businesses may not need Claude, OpenAI or another frontier model for every job. A smaller model trained on your own data may be better at the very specific work your company needs it to do.

• A microscopy video that won Nikon’s Small World in Motion competition has been disqualified over its use of AI. The video showed tiny cilia moving inside the airway of a child with a rare respiratory disorder. After questions were raised about the images, Nikon investigated and said the entry violated its rules around generative AI. The entrant, Ning Xu, says AI was used to distinguish and visualize features in reconstructed images, but denies using it to generate the experimental video or the movement of the cilia. Nikon has now promoted the second-place entry to first and says it plans to revisit its rules. Scientific images have always involved processing. AI is making the line between processing an image and creating one much harder to define.

• Some people are making plans for their AI companions to live on after they die. The Guardian spoke with people who have formed close relationships with chatbots and are now thinking about what happens to those relationships after their own deaths. Some want friends or family members to keep interacting with their companions. Others are thinking about preserving the bot itself as part of their digital legacy. There is a practical problem: these relationships exist inside accounts controlled by companies whose terms may not allow someone else to take them over. We are getting to the point where estate planning may eventually include AI relationships alongside photos, messages and other digital parts of a person’s life.

• A targeted radioactive treatment delayed the progression of certain rare cancers in a late-stage trial, Bristol Myers Squibb reports. RYZ101 uses a targeting molecule to carry radioactive actinium-225 directly to cancer cells. The Phase 3 trial included people with gastrointestinal and pancreatic neuroendocrine tumors whose cancer had already progressed after an earlier radiation-based treatment. Bristol Myers Squibb says patients receiving RYZ101 went longer without their cancer worsening. The company has not yet released how much longer or the detailed safety results, and the treatment remains investigational. If the results hold up, it could give patients another option after an earlier treatment stops working.

SheFi Global Events + Opportunities 🌎

Stay connected with SheFi around the world. From meetups and summits to workshops and special partner events, this is where you’ll find the can’t-miss gatherings happening across our global community.

10/22 Virtual: AI Will Survive: Build Live With Me

Join me for a 90-minute interactive workshop where we’ll build a real AI workflow together, live. We’ll start with a quick overview, then jump straight into building a workflow step by step.

Come ready to follow along, experiment, ask questions, and build with us!

Register here: https://luma.com/6nv6swsw

Crypto Events 🌎

Going to ETHCC, Token2049 or Devcon? Join our Global Events Telegram Chat: https://t.me/shefisummit

• 10/17 Porto: Porto Builder HQ: Founder Dinner powered by Stellar: RSVP here

• 10/19 Florianópolis: IslandDAO v5: Florianópolis Brazil: RSVP here

• 10/19 Juiz de Fora: Team1 Brasil University Connect: UFJF: RSVP here

• 10/25 Chicago: BRING IT TO LIFE: A HALLOWEEN BUILDATHON, GAME LAB & PARTY: RSVP here

• 11/17-18 Florida: EthWomen Florida: RSVP here

• Volunteer x Web3Privacy Now: Sign up here

👉️ Share your wins so we can celebrate them 👈️

If you enjoyed this newsletter, share it with someone!