2025: hard year, good year.

Work was… pretty crazy. Too much to go into here, but at Google I/O my team launched Veo 3 and a product (built with our friends in Google Labs) around it, Flow; later in the year Nano Banana, our latest and greatest image model, shipped (then its Pro variant); and around these we opened up the Music AI Sandbox (a suite of tools for pro musicians) and put out some interesting APIs, open source music experiments, and a model. We also started working more closely with professional filmmakers, premiering ANCESTRA, the first of three short films we’re making with Darren Aronofsky, at Tribeca. Generative AI is an increasingly contentious space, generative media also, and I’m proud of both the progress we’ve made, and the contrasts we’ve drawn between our work and that of our competitors. I look at ANCESTRA, or the delightful animations of Christian Haas, or SuperHero Sam (a couple of Sunday mornings with B), and I’m excited by the clay that the creators of today and tomorrow will have to work with.

All this definitely stretched me in new directions, and naturally that wasn’t always comfortable. It was the hardest year at Google so far - the pace in AI is intense, and in most ways I like that, but at points I noticed sleep and health start to fray in ways they hadn’t since the harder days of running Future Platforms, nearly 20 years ago. I took a month off in November to decompress a little before moving to a new role - still at DeepMind, still around generative media.

When time allows I’ve been noodling with some ideas - fairly abstract, noncommercial stuff around creating a robot which exhibits curiosity, and using this as a jumping-off point to explore simpler forms of intelligence, inspired by what I took from A Brief History of Intelligence last year. My GitHub history tells the story of when I’ve managed to find time for that.

It was a solid year for karate. I didn’t train as frequently as I’d like - averaging 1-2 classes a week, but this average is thrown off by travel (between work travel and a summer in the UK, I was out of town for about 3 months of the year). I had another fun weekend in Petaluma with Rick Hotton, whose teaching style I continue to enjoy and whose body mechanics I continue to marvel at. At the end of October I traveled to Okinawa with a group from our dojo, and retook my shodan grading (see 2023), this time passing - to much relief. I’m still helping with the Saturday morning kids classes (which I continue to find extremely rewarding), and B has continued to train, herself moving up the ranks and towards being one of the more senior kids, including a couple of gradings this year and taking part in our annual demo at the Cherry Blossom Festival - #prouddad of course.

Work took me to New York and Los Angeles frequently, London and Zurich a little. Kate and I spent 6 weeks of the summer camped out just off Worthing sea front, giving B some grandparent time. In retrospect it was too long - I spent the week days in London, but even there being 8 hours distant from most of my colleagues for so long didn’t work so well - next year I’ll do a much shorter trip for me. Kate and I made it out to Iceland for a few days while in Europe, which was phenomenal - in particular the boat tour of the glacier lagoon was an experience I’m carrying with me.

Outside of these trips, I spent quite a few odd days hiking around Marin and Point Reyes (which I have inwardly declared to be My Happy Place), continuing to gasp at the beauty of the Californian countryside, and made it up to Tahoe for a few weekends of skiing with B and friends.

We enjoyed a few sets of visitors this year, including an old friend from university days having her first trip to the USA - which gave me an excuse to wander around this beautiful city, and chance to see it all a little fresher.

I read a little less this year personally, but managed to keep The San Francisco Cognitive Science Reading Group alive. We met 8 times this year, mostly discussing books, each mostly split over a couple of sessions. This was OK, but next year I want us to get more into papers for more breadth and depth.

• Active Inference (Parr, Pezzulo and Friston) - explaining Karl Friston’s ideas uniting sensing and action. Friston has kept coming up in other readings, and we’ve enjoyed grappling with him.
• The Self Evidencing Brain
• The Book of Minds (Jakob Hohwy) - mostly a literature review, notably for kicking off a discussion of “can crabs play chess?”
• Where Minds Come From (Michael Levin) - a video rather than a book, but we wanted to dip into Levin’s work after he came up in Book of Minds
• Biological Underpinnings for lifelong learning machines (Kudithipudi et al) - read for the second time because neither I nor other members of the group noticed we’d done it before… perhaps a commentary on our own cognitive faculties. That said, it’s striking comparing my notes how I picked out very different details this time around.
• What Is Intelligence? (Blaise Aguera y Arcas)- by popular vote (though I spent 5 years working in Blaise’s research group a while back so would’ve read it anyway). Covers a lot of ground, I think I took more from the earlier sections (around his work with artificial life) than the later more philosophical ones.

We also moved venue at the end of the year to Frontier Towers on Market - somewhere I want to spend more time next year, an enjoyably shabby and enormously energetic co-working space / art + AI + robotics + bio-hacking lab. I’ve attended one other event there - the Wetware meetup - and had a couple of tours. Biology is not my thing, but the whole place feels special; perhaps I’ve spent too much time in Google offices, though I remember loving the more back-to-basics atmosphere in Android’s Building 44 too… so maybe this is a personal aesthetic.

Personal reading from this year - a bit less than normal, reflecting my working schedule and the tendency towards books at the reading group:

• Don’t sleep, there are snakes (Daniel Everett) - a missionary linguist takes his family to live with the Amazonian Pirahã people, to study their (unusually simple) language, translate the New Testament and convert them to Christianity… losing his own faith (in both Christ and Chomsky) along the way. Really really enjoyed this.
• Service model (Adrian Tchaikovsky) - amusing dystopian science fiction about a lost and effusively logical robot’s wanderings through a rotting world, after the death of its master.
• Fully Automated Luxury Communism (Aaron Bastani) - I felt bad for not having already read this but found it mixed. Opens with a set of cliched attacks on capitalism, goes through an amazing mid section plotting the economic effect of abundance across a few sectors (energy, food, minerals, health) and 180’s into a closing admonishment to enact left-wing staples (local credit unions). I’d love to understand more deeply the future they envisage - the technological and political seem quite separate visions. e.g. what are the cooperatives that workers share ownership of in a world of abundance where prices drop to zero?
• The Genius Myth - enjoyable wander through the presentation, celebration and reality of historical smartypants by the always-insightful Helen Lewis. Good zinger count (“Great discoveries create the conditions for their own under-appreciation”), finishes with a couple of good admonishments (stop thinking of genius as a transferable skill, apply the label to acts not people)
• Shop Class As Soulcraft (Matthew Crawford, AKA the British version I read, the much better titled The Case For Working With Your Hands) - really well written philosophical tract on our relationship to manual creation and consumerism, by an author who ran a think tank before leaving to work in a motorcycle repair shop, with no regrets. Mashes insecurity-buttons for this carpenter’s son who lives 90% massaging bits not atoms.
• What Art Does (Bette Adriaanse and Brian Eno) Really good, fun, short book exploring what art is, what it is for, how to engage with it. Read it myself, since used it as a bedtime story with B, who enjoyed it enough to keep asking for it.
• Abundance (Ezra Klein, Derek Thompson) - TLDR we’ve made it deliberately hard to do things which require a lot of coordination, for good reasons (climate impact, getting a handle on corruption etc) - but in doing so, have implicitly created a vetocracy and should, to paraphrase Graeber, choose otherwise. I loved this one, though it plays to many of my preconceptions (“let’s science our way out of this”).
• The Great Automatic Grammatizator and other stories (Roald Dahl) - I’ve only read the main story but it’s astonishingly prescient for 1953 when it was written : a frustrated writer invents a machine which encodes the rules of grammar and plot and can create novels. Themes he lays out are ones we’re grappling with now: can a machine be creative; unequal access to technology; AI slop (in the story, the machine is biased towards mediocrity); safety issues; engagement hacking; even a hint of AI take-off.
• Station Eleven (Emily St. John Mandel) - an aging actor falls dead on-stage during a performance of King Lear, and in the days following, a plague (“Georgian Flu”) kills almost everyone, leaving survivors wandering America or setting up encampments.

I failed to finish How to do nothing - the jokes write themselves.

I think 2026 will be about enshrining better habits, whether they be about preserving attention, health, boundaries or sociality.