I was reminded today that I’m supposed to write a user manual for myself. It’s a document written by you, about you, for the benefit of the folks you work with. I like the idea and I’ve done it before, but it got me thinking: why don’t I just (or also) answer the same sort of questions right here.

It’s specific enough that I’ve created a new place for it (for now). It’s not linked from any navigation, but I might send people to it, and I might point robots at it. For now, I’ve answered a single question—and the most meta one—about why the space exists at all, which... I’m now repeating. Oh.

Anyway, that’s how I spent a bit of my time this evening, at the end of a day that’s felt both very long and very short. I made the changes with the OpenAI Codex app, which was both fun and completely pointless. Then again, I think I might want more fun, pointless moments in my life.

Aneesah’s traveling to England this evening on the red-eye. I drove her to to the airport, and as I was putting her suitcase in the boot of the car, I asked whether she thought it was more sad to be the one leaving, or the one staying. We agreed that it feels worse to leave—although now that I’m sat here alone, missing her, it doesn’t feel much better.

I say “alone” but Cacio is here with me, which does make it less lonely. Right now she’s walking all around the house looking for Aneesah though, which doesn’t make her great company, but does make feel something. She loves Aneesah so much, and I assume likes me enough. If we meet Aneesah from the BART she drags me all the way to the station.

It’s funny, whilst you’re busy doing things your brain lets you forget that you’re apart, but it saves up all of the feelings for when you climb into bed, and wonder how you’ve ever fallen asleep by yourself. That might be a bit dramatic, but I suppose that’s just what love feels like. The good kind of drama, I mean; a West End show, not an argument.

An excellent collaborator gave me a great gift today: the chance to talk, without taking a breath, about things that I’m interested in. They asked a simple question at some point: “what is MCP?” I could have answered with something like “an abstraction between tools like Claude desktop and external systems, like your computer or an API.”

Instead, I said something like “let me take a step back and talk about training and tokens in the context of large language models... and then I’ll talk about tools and tool-calling... and then I’ll talk about schemas and grammars and sampling... and then, finally, I can talk about MCP... but only briefly, before talking about APIs.” Good lord.

Truly, what a gift it is to just let someone talk at length about something they’re intrigued by. What grace you must have to sit and listen, knowing that there must be a better way to answer the question. What patience and kindness you must have to not point that out. We should all be so lucky to be given this time and attention.

Aneesah and I headed into San Franciso in the evening to see Ed Gamble perform at Cobb’s. We paid him attention, but of course we did. We paid money to pay attention. You might say we paid twice, in that case, but we definitely paid attention. Great show, very funny, but sat here now thinking about the stage that many of us have.

It feels good to get up on stage—real or otherwise—and perform for a bit. Not all the time, but sometimes. Feels better with an audience, even (especially, perhaps) an audience of one. I don’t think that a sold-out Cobb’s wants to hear me talk about MCP for an hour. Then again, it is San Francisco—maybe that’s exactly what they’d want.

The heating hasn’t been working today, and it’s suspiciously cold in the East Bay. You realize how much you take warmth for granted when you don’t have it. I’m typing this under a comforter on the couch, with the dog flopped over my left leg and my body awkwardly twisted so that I can still type. She was curled up under this same comforter today.

I’m reading about machine learning, so that I can read about deep learning, so that I can read about language models, so that I can build a language model. I both love learning new things and wish that I could learn everything I need to know immediately. I just love making stuff, but I’m in the stage of this journey where I mostly feel really dumb.

I’m thinking of getting a printer that’s more geared towards printing lots of pages as quickly and cheaply as possible. I suspect I’m going to be reading a lot of research papers over the next few months, and I sort of hate the pinching and panning of PDFs on screens. People swear by large e-ink readers for this, but I’m not totally sold. Paper is just... nice.

One of the benefits of reading them on some sort of screen today, of course, is that you can ask the magic internet brain about them. For that reason I’ll probably end up back on a screen, but maybe I’ll intentionally try a first pass without a computer in sight. I think learning happens when your brain hurts, and you actually let it. Embrace it, even.

I took the day off today. I’d contemplated not taking the time because, like many of us in the year 2026, I’m plagued with the seemingly-incurable illness of “never feeling like I’m doing enough.” Completely self-inflicted, I should add. Anyway, I took the time in the end, despite the illness.

I made coffee, did some chores—mostly vacuuming the places that I don’t vacuum and that collect dog hair as if it’s a precious resource—and then walked Cacio in the rain. We walked to the bookstore so that we had a destination, and I bought a copy of the Atlantic because... well... I had to buy something. Stuffed it in the pocket of my Barbour and forged on.

I listened to an audiobook while I walked: Empire of AI by Karen Hao. So far, it’s covered events that I could have probably written about myself because, like many of us in the year 2026, I’m terminally online. A good book nevertheless. Pleasing narration, even at 1.75x. Whenever I commit to learning something deeply, I tend to shove maximum adjacent stuff into my eyes and ears in the hope that it’ll all glom together in the end.

The main book I was following turned out to assume machine learning knowledge I don’t possess, so I picked up another couple books to dig into alongside it. I’m going to keep building, because I think that you should always learn through building (at least). You can read and read forever, but at some point you just have to start, and to make stupid mistakes over and over again. That’s where the real learning happens, I think.

Got back and read half of the one article I was actually interested in from the Atlantic, about AI and the future of work. It’s probably a great article, but I felt like I’d read it a few times. I’m glad to support the Atlantic though, because they published one of my favorite essays on the future of technology a long time ago: As We May Think by Vannevar Bush.

So in the end a day of learning, and learning that’s relevant to my day job. Maybe that means I didn’t take the time at all, but it turns out that work:life is more complicated when your day job includes many things that you’d do regardless of whether it was economically useful to society. I’m lucky in that regard, obviously. I feel lucky every day, in fact.

Woke up super late today. Must have needed the sleep. Was going to tidy up the garden and work on some side projects, but it started raining so I joined Aneesah on some errands in San Rafael and Mill Valley.

We popped to the restored Mill Valley Lumber Yard, a cute collection of stores and restaurants that makes you feel like you’re cosplaying a lumberjack, but one who buys overpriced beanies and only drinks perfect flat whites. It was chucking it down, but we grabbed an early dinner and sat under the heater outside with Cacio. We watched the rain pour down in sheets with the most beautiful backdrop. I loved every second.

I love the rain in general. I love being warm and dry but being viscerally aware of the wet and cold. There’s something about knowing that the warmth shouldn’t be taken for granted. That you’re warm despite the weather, not because of it. I like being in the rain too sometimes. I even like camping in the rain a bit. Growing up in England does that.

I really can’t imagine a better feeling than being wrapped up warm in front of an open fire, in a lighthouse on a remote island, whilst wind batters the windows and the rain and sea rages outside. I’m certain that I’m supposed to be a slightly moody lighthouse keeper, and that I was simply born at the wrong time and in the wrong place. I might change my mind if I ever actually got the chance, of course. Still, I can dream.

It also reminded me that running errands can be the good stuff. The good times. Living in the Bay Area tends to make this more true I think, because regular commutes might happen through awe-inspiring landscapes, but it wasn’t just about that today. It was about spending a simple, quiet moment with my family on a slow Sunday.

“Happy Valentines Day, how are you doing tonight?” The server suddenly appeared at our table with a warm smile. We like this server, she seems kind but also like you could talk shit about the world with her. We’d popped out for dinner at a local spot that doesn’t take reservations. We thought it would be a long wait because it normally is, but we were lucky.

We ordered too much on purpose, because there are some situations where you’re supposed to lean into abundance. The food was great, as usual, if not a little confusing at times. We ordered a ragu that tasted nothing like ragu, but did taste like pasta and meat cooked in butter, and I’m not going to complain about anything that tastes like it was cooked in butter. Butter makes everything better. Seasoning does, too.

I ended with affogato because it’s the best choice when you can’t decide between coffee and dessert. You’d think that both ice cream and coffee would make it harder to sleep, but as soon as I got home I just completely passed out on the couch. That’s rare for me. I’m writing this the next day because I was absolutely incapable of writing it yesterday.

Now that I think about it, I did start the morning with a 6-mile 8am run along the Berkeley waterfront. That might go some way to explaining why I was more tired than usual. I’m going to sign up for the Golden Gate Half, because after the 6-mile run I really thought that I could do the same again. It would be a slow time, but I’d have done it, and that’s something.

We went to see James Acaster at the Masonic. Is there a more delightful job than making people laugh? I’m not sure. Probably a bit stressful too, but maybe not. Afterwards—and I’m not sure how or why this place exists—we ended up at Tonga Room, dancing to covers sang by someone on a floating platform in an... indoor lake, whilst water rained down from the... ceiling? Unsure if it actually happened or if it was just a fever dream.

There something so comforting about experiencing the comedy that you spent so long surrounded by. We loved going to comedy shows in London, especially the small ones in little pubs. The Bill Murray in Islington hosts some of the best comedians in the country as they practice new material. A dingy little room where you’re basically sat on the stage. It’s perfect. The Masonic isn’t that, but the show was great.

He’d been here for a month on tour, and he spent a few minutes talking about a consistent experience of mild... rudeness? Something like that. Like asking someone where the cereal is and them just staring, saying nothing, and walking off. It got some laughs, but it didn’t resonate with me. There are some curious cultural differences here, but I haven’t found rudeness to be one of them, at least in New York or California.

The show made me miss England in some ways. The mannerisms of people. The cultural in-jokes. The self deprecating nature of folks. It also made me glad for the things I experience more of here though. The general sense of optimism, in spite of the challenges that the country faces. The earnest nature of folks. The lack of complaining (or at least, the lower volume of it). How dare a comedy show make me introspect.

Tokens. Embeddings. Dimensions. Weights. Logits. Softmax. Loss. Backpropagation. Attention heads. Epochs. Feed-forward. Not all new to me, but I love the early phases of digging into some big new thing and teasing out all of the new language into concepts that you already understand. My favorite thing about learning in general is that you’re often learning concepts that transcend the thing you’re learning about.

Went for lunch with a friend and wondered by I don’t do it more. It’s such a great feeling to add more chapters to a day. I can get obsessed enough with whatever I’m doing that the hours disappear. The real joy though, of course, was simply spending time with a friend. We talked about work, life, family, parenting. I’m convinced she has a time turner, because she packs so much of it all into the same 24 hours each of us has.

One thing that we talked about is the need to be unreasonably optimistic when you’re getting something new and ambitious off the ground—or, in fact, doing anything that requires real perseverance. It’s the optimism that makes us keep going, through the parts that feel like swimming in treacle. We talked about how manifesting is real; how self-fulfilling prophecies work. Assume it will work, and it might just.

I think that brings me right back to the beginning: I’m not a machine learning engineer, but does that matter? Truly, you can just do things. You can just start playing and assume you’ll figure it out. That a little effort today, and the next, will build into something worthwhile and rewarding. I have to work hard at this, but: there aren’t many feelings greater than believing in yourself. Evidence helps, so I’ll be back tomorrow.

I was listening to some folks from Anthropic talking about interpretability whilst doing chores, and thinking about how totally strange and yet vaguely mundane the world is that we’re living in right now. I mean, it’s not mundane for many reasons, but I’m specifically talking about the way that language models are permeating every corner of western culture.

We’ve invented this strange new thing, and we don’t totally understand all of the ways in which it works, and we’re only now getting a good enough look at how it really works such that we can begin to steer it, and yet we’ve inserted it between ourselves and our everyday lives. Ourselves and our work. Sometimes, ourselves and our art. Are there any other examples of this, at such speed, from across all of history?

I’m writing this in IA Writer whilst a Jupyter Notebook peeks out from behind the window. I’m writing words that will make a machine (eventually) write words, but not any that I’ve told it to write. I’m writing in a text editor that has a first-class feature for annotating text as authored by AI. I’m mostly writing stream of consciousness; just predicting the next word, and then typing it—but it’s more complicated than that.

It turns out that it’s more complicated than that for language models, too. There’s enough evidence from research teams that the model might “think” a few words ahead, even if it’s only predicting the next word. We’ve jumped so quickly to labeling these things as either “just another tool” or “sentient silicon-person” (because either is comforting in some way) but the truth (as it often is) might be somewhere in between.

Anyway, that’s part of why I’m building a (small) large language model (I know)—to better understand this thing that we’ve made. The other reason, admittedly, is that it’s just fun. Non-deterministic computers are goofy and I want to play with them. Computers have done exactly what we told them to do for a long time. Probably good? Maybe not?