Thunderstruck
w.311 | Buc-ee's, Nuclear Criticality, Aalo, AI ROI, & FWEs
Dear Friends,
It’s been nearly seven years of (almost) weekly writing. The world has changed since then. ‘No one’ really reads long-form content anymore, and yet the quality of writing and reading has never been more important. So, I’m going to evolve the format of this newsletter by integrating all the pieces into a coherent narrative. I’ll experiment for a while and see what clicks. Feedback is always welcome.
With the US hosting the World Cup this summer, it’s hard not to feel pride as foreigners venture between venues and, in doing so, explore the depths of Middle America. And, for those in the know, I also recommend a stop at Buc-ee’s, a gas station chain (whose typical format features 100 pumps) known for its exceptionally clean restrooms, inexpensive Texas fare, and endless swag. True modern Americana.
The visit shown above was a few months ago and my first. It’s everything good about America: Positive, optimistic, welcoming, inclusive, friendly, outgoing. It’s for-profit, enterprising, and well-paying, too. Capitalism works, see?
The US Government (and all of us) notched a win with nuclear.
While AI grabs most of the headlines, one of the massive technology successes of the past two years has been the acceleration of the US nuclear fission industry. When President Trump came into office, he issued a flurry of executive orders to revive the nuclear energy industry, including revamping the approval processes. Without going into minutiae, three of the eleven selected projects are set to meet the July 4th criticality test deadline at Idaho National Labs (INL). The government (DOE, INL, NRC, etc.) has worked promptly and consistently to assure investors and founders of timely, safe regulations and approvals. It’s a blueprint for other areas of American industrialization and innovation.
Aalo Atomics, which I invested in last year on the thesis that AI Needs Nuclear Power Now, received approval this week to test the Aalo-X, a 10 MWe reactor.
Thunderstruck, a thrill-seeking anthem by AC/DC, an Australian band describing a night out in Texas (official video here), strikes this week as the walk-in music for John Wagner, the INL Director. Well deserved.
As an investor, right now, I like the energy part of the stack. It’s an input into everything, and we can’t get enough of it. The solutions are real and obvious. You just have to execute. Go Aalo! But let’s talk about the rest…
Fable, like a lightning strike, powerful but not for long.
A couple of weeks ago, my husband turned to me and said, ‘The Internet seems to think this Fable is a big deal.’ I was busy playing with the latest Claude model release, so I could only reply, ‘What? Not sure. Haven’t heard of it.’
Like everyone else, I was re-running every task and analysis to see the differences from the model upgrade and reflect on the output. It was a disappointment when it was taken away. The timeline for the re-release of Fable remains unclear.
While the defense implications and power of the models for cybersecurity and the like are all being worked out, executives across the spectrum are being told AI is the solution to their problems, but now access to the latest and greatest feels further away. When leaders have only so much bandwidth to follow the latest back-and-forth, we risk the modal response becoming ‘tell me when it is over.’
Is AI adding new revenue, or is it working to reduce costs?
The revenue side (unless you are selling into AI Capex or AI native services) is tricky. It requires creativity to sell more and pricing power to raise prices. AI is, almost by definition, not creative (see below).
On the cost side, today the efficiency gains for the enterprise are concentrated in coding and software development, with call center agents closing in.
The cost-efficiency-over-sales-growth claim is backed by analysis (see below) in a new report from Exponential View, which is worth your time.
True professional services are still a way off. It’s impossible to bill AI spend to your client like you can billable hours.
Enterprises all need help with adoption, so the consequences are the same as they ever were. A Forward Deployed Engineer is just a consultant with a different name and a slightly more technical orientation. Yes, the traditional FDE or TechIT consultants have been crushed in public markets (see below). But my MBB friends are doing fine, and we have fun new company names like DeployCo (OpenAI), Brain Co (strangest founding team ever?), and the OGs Palantir and Invisible.
At the consumer and solo-preneur levels, most people will do the work themselves using AI tools. Easier, cheaper, better. They will keep the surplus.
‘Self-help’ category is the original slop of old ideas reproduced for every context. Tim Ferriss noticed that his book sales are declining (-57% between 2025 and 2026) as he says: the market for information is collapsing into the chatbot. This mirrors the collapse in demand for executive coaching and, perhaps, for higher education.
As for me, I don’t use AI for writing & voice. You can tell. It’s pretty embarrassing when you read pieces from senior officials/leaders that are obvious AI slop. Either take the time and write something meaningful, or expose yourself for not having your own thinking and reasoning. AI does help with research and with editing. It can be an honest mirror and instant feedback - telling the truth quickly and often brutally, with the right prompt, of course.
Thanks for reading, friends. Please always be in touch.
As always,
Katelyn