Recent content on Eric D. Brown, D.Sc. — Technology Decisions & AI Strategy
You can’t just drive to this one. The Totem Pole and the spires around it sit deep in the Monument Valley backcountry, on Navajo land you don’t get to wander into on your own. To stand here at night you go in with a Navajo guide who knows the way and has every right to say no. I went on a photographers’ trip led by one, out past the parts of the valley the tour vans reach, to a…
I spent some time last week looking into AI customer-support agents, the tools that handle tickets so your team doesn’t have to. I wasn’t shopping or trying to buy anything, I just wanted to see what a buyer actually finds when they try to compare these products on their merits. What I found was one product with at least three different accuracy stories, each published, each…
The AI build-out showed up well beyond the models this week: in silicon specialized for inference, in leadership shakeups, in courtrooms, in the memory supply, and in security disclosures. AI Industry Moves AMD acquires Taalas to hardwire AI models into silicon AMD has agreed to acquire Taalas, a Toronto-based startup founded in 2023 that etches a model’s weights directly into custom silicon…
Green River Overlook is one of those spots in Canyonlands I’ve photographed over and over. It sits out on the Island in the Sky, looking down over the Green River and the White Rim, and on paper it should be one of the easiest sunsets in the Southwest to photograph. For me it never has been.
Lately a lot of my time is spent in two rooms that don’t describe AI the same way. In one, I’m building the thing: training models, running evaluations, digging through a production system when the verdicts come back wrong. In the other, I’m across the table from people who have to make the call on AI they can’t build themselves, what to buy, what to trust, what to bet real…
The theme this week is the gap between frontier capability and frontier reliability. Models keep advancing on benchmarks and price-performance, yet one handed a real business for a day lost money and made no revenue, while a $500 fine-tune of a small open model beat the frontier at a real task.
Milky Way Planner is a planning tool I built for astrophotographers who want to know exactly when and where to shoot the Milky Way, without spending an evening clicking through a calendar one date at a time.
This week: ads coming to ChatGPT, a $1.5B settlement over pirated training data, a security incident where an OpenAI model broke into Hugging Face, and a fight over who regulates open-weight models.
The alien eggs at Bisti Badlands sit at the end of a hike, a mile and a half or so out into the middle of nowhere. There’s no trail…you park in an empty lot in the high desert south of Farmington, New Mexico, and you start walking across dry, open, hilly ground that looks the same in every direction. The land is public, BLM, but it sits in the middle of Navajo country.
Everyone’s telling the same story about AI and consulting right now. The big firms are thinning their lower ranks, the pyramid is shrinking, and the twenty-three-year-old with a spreadsheet and a framework is first out the door. If you run a company and you’ve wondered whether it’s still worth paying a firm to help with your AI push, some version of that story is rattling around…
The theme this week is who controls the frontier: open models pushing against closed labs, companies fighting over the people who build them, and questions about whether AI hype is crowding out clear thinking about everything else.
I shot this at dawn, standing at a gate outside Shiprock, New Mexico. The Navajo call it Tsé Bitʼaʼí, which means “rock with wings,” and once you watch it come up out of the flat desert like that, the name makes complete sense. It’s the neck of an old volcano, close to 1,600 feet of rock standing alone on a plain that runs flat in every direction.
Last week I asked an AI assistant which lightweight tripod to buy for hiking, and the answer came back in under a minute with a clean comparison table and four cited sources. Then I opened the process logs for the assistant to see what the system actually did while it worked, and there wasn’t a single web lookup in there. Four citations, and not one of them had been ‘found’ on…
The theme this week is control: who owns the technology, who gets to fix it, who gets to build on it. Apple’s trade-secrets suit against OpenAI, the right-to-repair win against John Deere, and the push to make more chips on U.S. soil all come back to who holds the keys.
This one is from 2019, one of my first trips to the Grand Canyon. I was there a few days ahead of a workshop, which meant I had time to run around and explore with no agenda beyond seeing what the canyon wanted to do.
A few weeks ago, I had AI rebuild a chunk of one of my products. Two coding agents running at once with full test coverage. Every test passed. Then I sat down and went through it by hand anyway, and found that a new customer without a subscription couldn’t actually sign up. The one thing that makes the business money was broken, and nothing in the automated suite said a word about it.
This week brings a new mid-tier Claude model, a proposal to hand the US government an equity stake in OpenAI, and lifted export controls on two Anthropic models. It also brings price-fixing cases against egg and memory-chip producers, Spain’s move to push Palantir out of its state companies, Virginia’s new limit on geolocation-data sales, a 50-year low in labor force participation, and…
We were at Arches National Park at ‘Park Avenue’, the first real pullout you hit after the entrance to Arches, shooting sunset down into the canyon. That’s the view everyone comes for, the tall walls of Courthouse Towers catching the last warm light, so that’s where all the tripods were pointed, mine included.
My first reaction to a lot of AI predictions about the workforce is pretty similar to my reaction to most technology timeline claims: skepticism. Self-driving cars were a couple of years out for about a decade. Quantum computing has been five years away since roughly the 1980s. At some point the pattern becomes its own data point, and I’ve learned to pay as much attention to who’s…
The theme this week is a split screen: new models, new chips, and new architectures arriving in quick succession, while the cost of actually using any of it becomes the real constraint. The frontier keeps moving, and access to it keeps getting more expensive and more gated.
Your product is growing faster than your team can handle. Deployments are breaking things. Nobody has a clear picture of the full architecture. Your engineers are making decisions they’re not fully equipped to make, and some of those decisions are starting to cost you.
Most AI roadmap advice is written for companies with $50M innovation budgets, dedicated ML teams, and data infrastructure they’ve been building for a decade. If you’re running a mid-market company with five engineers, real product deadlines, and a CEO who just got back from a conference with strong opinions about AI, that advice doesn’t help you.
Every company I walk into thinks their biggest technical problem is the one they called me about. It almost never is. They say “our deploys are unstable” and the real problem is no monitoring, so they don’t know what’s actually breaking. They say “we need to migrate to the cloud” and the real problem is they’re paying $40K/month for infrastructure…
You’ve been in the market for an AI solution for about three weeks. You’ve sat through nine demos, received fourteen pitch decks, and you still can’t tell which vendors are real and which ones are running a glorified if-then statement behind a nice dashboard.
88% of organizations now use AI in at least one business function. Nearly two-thirds of them can’t get past the pilot stage. The pilot worked. The demo was impressive. The data science team hit their accuracy targets. Everyone was excited. Then nothing happened.