2024 has felt like four years compressed into one. It's been the least silent year of my life, in more ways than one. Somehow it did that without being loud.

This was another year where the progress in AI - both in my personal and professional lives, and in implications for the world - has kept accelerating. It's been the longest unbroken period of growth I've witnessed in the industry.

This was also my last year at the company and home for the past half decade. More on that later. It was also the year I launched, wrote, swam, ran, and flossed the most out of any year of my existence - but I get ahead of myself.

§Act I - the first threshold

The year started with one of my favorite things I've ever built, and one of the best releases we've had at Greywing: WalkingRAG and Proteus. It was arguably one of the first agentic retrieval systems to exist - just seeing it come to life and tackle hard questions in ways I couldn't have done myself, still gets me excited.

It also started, as the next one likely will, with a review of days past. 2023 was the year of bagels, building and bombers.

January had me posting a few ideas and predictions that excited me, like data transformations using models, and why the web is the best platform for modern LLMs. Little do I know that I'll be forced to put my life behind those predictions before the end of the year.

The first month also marked the first time I'd put myself on camera, in the public eye. I have been blogging (and giving the occasional talk to a small audience) for many years now, but this was different. Somehow along the way I realised there were more people listening. This blog still feels like a private space, somewhere we can just talk. Twitter and Youtube felt like everyone else, and you find yourself battling the eyes you feel on you long before you've hit publish. Something about facing a camera yourself, with multiple takes to no one and everyone at the same time, felt really hard.

It went well! I was overwhelmed by the number of people that reached out (and still do). Next year I'd like to keep doing more, for both selfish and unselfish reasons. I'd like to get better at expression when you don't have the tight feedback loop of body language from a real person. I'd also like to contribute what I'm learning, to the same platforms I learn.

In a few months I'll admit to myself that I've burned out, but looking back, it hasn't started yet. Right now I'm having more fun than I've ever had - and trying to talk about it, and finding so many people building new things. It felt like the days of increasingly commoditising SaaS were over (for a while), and completely new ideas were everywhere.

Feb is when I start feeling a pull from different directions. The world is definitely changing, but we don't know where. I begin to get the sense that the structures of companies, people, social networks, and technology were also beginning to waterlog and crumb away. However, I also have a job, and a responsibility to the people at my company, even as I try and figure it out. As I do so, I begin to wonder if I'm being helpful.

March sees me depart as CTO - nothing functional is expected to change, it's more of a recognition that our paths are diverging. It's also an opportunity for me to try and recover. After months of fundraising, product development, a number of specific customer interactions left me feeling pretty burned out. For the first time in five years, I realise that I'm not looking forward to work, while marveling at the fact that I had done so continuously for 1900 days. I still love the people at my company - I see them often - and all the things I've seen them do, but I feel like the wrong fit for this industry.

§Greywing

I've thought many times about writing a review of my time, but it feels strange to be doing that when Greywing is still a living organism that is in-process. Maybe some day I'll take on that gargantuan task. Right now, I have nothing but fondness for my time with all the wonderful people that made Greywing what it is, and are taking it where it is going. Not everyone gets to build the things they want with some of the smartest and most inspiring people in the world, but I did. If I could go back, I'd choose to spend the second half of my 20s once again with Nick, Greywing, Hendy, Bob, Meghna, Garima, and everyone that gave their time and their life to it.

§Burnout and recovery

The good news - as I fail to resist the urge to be the insufferable silver-lining guy - is that being more aware of my mind and my present moment made burnout (partially) a good thing. It helped me get high-resolution frames of what it's like. It helped me see the wonderful people around me as they jumped into my hole with no thought to themselves, to see if we could figure it out together. Looking back it felt like one of those slow-motion videos of crash tests, except you're the dummy. You get to see every causal chain, every exacerbating factor, but with very little felt control. I wouldn't recommend it, but it was fun. I found a lot of new friends, a lot more fellow travelers who'd been here - once I could admit it to myself and everyone else.

This was April. It took a good few months before I could - as one of them put it - hear the birds again. It helped that we continued to be in one of the biggest technological shifts in human history. I built something (actually many somethings) - and walked down a painful path. But that's Act II.

§Act II

Somehow after a month or two of picking up some new things, like running and skiing, seeing family again, and doing all of the things you didn't realise were on pause for a while, I felt a little better - only to instantly fall back into the old causal chain. It's hard to stop working when you've been showing up every day for years. It's also easy to misjudge recovery.

So here's what happened: I built something, talked to people about it, got offered funding and customers, and got on a flight to New York. (Too soon, I know.)

I'll leave that story for another time. Suffice to say that it turned out to be once again one of the most stressful months of my life. A speedrun through building another company ended in a semi-peaceful agreement to walk away.

The silver lining turned out to be this: Circumstances (outside of any single individual) asked me point-blank, again and again, the question I've often wondered about: do you have a number? Do you have a number you'll put aside passion (or calling or being helpful) for, or cross some amoral lines to get?

This is one of those questions you think about. You think you know the answer, but you also know deep down, that until the day comes when it stares you in the face with real stakes, you won't know what you'll do.

On my first date with my now girlfriend of 8 years, I almost drowned - and she saved my life. Before she did, there was enough time underwater when I was sure this was it. It was a fortunate moment - because I got asked a very similar question. Are you okay with death? Turns out, I was. I definitely wanted to live, but the moment came and I felt myself be okay with going back to nothingness. It was also something that saved both of our lives. It kept me from drowning the wonderful person half my weight that was trying to save me.

As it turns out, I don't have a number. It took a week of intense stress, but deciding to walk away gave me a lot of clarity on life ahead.

§Fixed points and transitional comfort

So now we're around August. At this point we've flown twice around the world on far too many flights, moved houses, lost and gained multiple professional paths, while being once again exposed to the raw uncertainty of life. By the midpoint of this year I'd lost both a fixed income and a fixed home, of the constancy I've had since 2018. Neither of those were really truly fixed - the last place was a rental, and my last income was a startup - but I'd learned to rely on them.

Somewhere along the way I felt the original goal of 2022 - of being at peace in chaos - really come true. Somewhere along the line, my perspective shifted from this:

to this.

They're the same graph of a butterfly attractor in real life (from this old paper I wrote in college).

I feel I've found a kind of transitional comfort zone, of really connecting with change as the true constant. It feels like a hard-won perspective that feels even harder to keep around, but I'll try.

Another way to put it, is that it feels like a state of continuously updating priors. Travel - especially non-leisure travel - forces this on you, with more people, more cultures, different timezones, schedules, etc. But it's especially present in a fast-moving space. What was true tomorrow may not be true today, and it's often to your benefit to learn that early.

§Expression can be a life-raft.

What helped was a lack of silence. Better put, it was proactive communication. I think - actually I know (look at the chart!) - I wrote more this year than I ever have in my life. I spoke more, and at more depth, and communicated with the people closest to me, more than ever.

Writing has been one of the most helpful things. It retains a timeless, slow quality that I keep coming back to as a way to really understand what I think - or to expose my ideas as disorganized.

Taken together, it's really not all that many words. They represent about 18 hours of reading, at an average reading pace - about 2 hours for every year I've been writing here.

Communicating and connecting through expression has made all the difference in 2024. I find it hard to take sole credit (or debit) for almost any decision or turn this year - in a good way. A lot more things have had multiple contributors, in ways that have made those junctures better than the sum of their parts.

§Act III

New beginnings. My nature is often to put my head down and focus on the problems of the day, to make progress on smaller units while moving in the right direction. Pre-GPS, I remember direction finding in a new city: my natural guidance leaned more towards finding moves that got me closer to the target (hopefully not in an asymptotic fashion), instead of routing the entire path.

It's only in periods like these - where you exit your previous forest into a clearing - that you can see the woods, at least for a second. You get to ask yourself if you want to keep doing the same thing, or to do something else.

I found myself asking the same questions over and over again, getting into the habit of talking to myself over long runs, coming back to write and talk, write and talk, making incremental felt progress along the way.

The more I thought, the more I couldn't get January out of my head. We built a lot of wonderful things at Greywing, but it was a sideproject - a little script called MagicETL that barely worked - that kept coming back to mind. As someone who's wrestled with data for over a decade, I couldn't shake the thought that these artificial minds we've built will change the world, if we can help them do it.

When you think about something long enough it feels like a categorical imperative, and there's often nothing left to do but to do it. You do it so the ideas will leave you alone, once they find their form.

We're now a month or so into a fledgling enterprise we're calling Southbridge - an attempt at unifying information, finding new ways to process it, to actually make use of everything we've been saving for decades.

It's hard not to see a thread stretching further back than this year, as something I've been thinking and writing about for a while - without knowing it.

I have something almost done that goes into more detail. For the time being, you can read this Wikipedia page on Southbridge, or any of the things we've written about or built.

It's made up of wonderful people, who consistently surprise me with more than sparks of GI and moments of genuine creation. I think this is the beginning of something awesome, with some masters of karate and friendship.

Remembering the LAN from David Crawshaw often comes to mind when I think of what we're trying to do. Both data and networks have been things that have been so hard for so long, that we've folded those problems into our lives. Just as they've solved one, I think we have a shot at solving the other.

§Everything else all at once

So much more happened this year. It's strange having a profession that is also an obsession, an addiction, a joie de vivre, and a source of inspiration. It can be easy to forget that life is still larger and that things connect if you look deeply enough. Some years it can be easy to forget to live, and to forget that life is being lived outside the world of pure information.

This was not one of those years.

It has been a year of many wonderful moments - from running along the Hudson listening to Gus Dapperton, to having my first ever birthday party at 30, lots of mulled wine, and delirious jetlagged meals with friends.

§Positions reversed

I have been repeatedly told for some time that my 180s are discombobulating. There's often considerable whiplash to hear me reverse my position on something I've been vehemently for/against, and to find me once again passionate on the other side.

So it is that I've started putting down my 180s. I'm hoping it's a good thing, that it's really just strong opinions weakly held, but we're paying attention to make sure. In retrospect, I think I always suspect that I'm wrong about everything. 18-year old me was confidently wrong about too many things, and 25-year old me was no different. Unless something has fundamentally changed, 30 year old me is likely doing the same thing.

So what did we change our minds on?

The first thing is - embarrassingly - flossing. After spending most of my life arguing for its vestigiality (and being blessed with good teeth), I finally picked up flossing as a habit. I know - I know.

The second is Sonos. After a half a decade of fighting what I believed were overpriced speakers with wifi, I caved. I'm not sure I deserve much credit here though, I think things changed while I was holding on to my beliefs.

Something happened while I wasn't looking. Sonos somehow went from a company with okay speakers and an amazing app experience, to a company with really terrible apps and genuinely great speakers for the price.

I think the reason was the rise of true, object-based audio. Atmos (and other formats) crossed the threshold for me, where I couldn't see myself building a similar system with speakers and an amp for a lot less. I think it was true at a time when surround sound just meant routing channels to speakers, but today it means tuning, reverb, “throwing sound”, beam-forming, reflections, and far too many things I admit I do not know how to do.

The third has been running. Having never been able to run for more than a minute - and likely building a position against it as a result - I started running in May. It started as a way to improve cardio, after cycling at safe speeds became too time consuming, but it's grown into a true balm.

The wins here have been exciting. It was nice running my first 10 miles last week, at a faster pace than I ran my first 5k. The last 3k of the 10 miles felt exactly like my first 5k though.

It's been a wonderful way to make new cities feel less foreign. Something about running (more than walking) forces your brain to learn the implicit physical language of a new city. You get better with the things that contribute to felt micro-decision fatigue: people on the street you need to weave through, street signs, cars, crossings. Running speedruns you through learning these things. I find myself far more at home in a new city once I've run for a few days. (I'm also told it's done wonders for my backside)

§New things

Or new things devoid of Enantiodromia. (New word of the day for me 😅)

I learned to ski this year - or I should say I got through my first ski season without hurting myself. Having trained on motorcycles, it's still unintuitive to lean to the outside of a turn, but I'm learning.

AI has been an incredible companion for learning and making new things. I picked up sewing, and made three garments in a week - each tougher and better than the last. I can say I've gone from 'did you find that in a ditch' to 'clearance rack at Shein'. Claude was the most helpful here, pointing out the specifics of things that would've previously taken a lot of orthogonal reading until something clicked - like sewing on bias.

Made more fun code things from scratch, some of which got to #2 on trending! It was wonderful to see all the things that were made with Lumentis, and to get messages about it from people I've long respected.

I'm a month into learning to swim - somewhere along the line, as I cough out some more water from my lungs, I wonder if I've really learned to like sucking at things, to really enjoy the feeling of being incredibly bad at something that seems effortlessly easy to other people. It's fun.

§Things that continued

I've kept up the habit of making incessant TODOs. It's left my mind cleaner to wander in the present. Almost anything that doesn't need to be in working memory goes into Todoist. The numbers back me up:

It's left me with more mental space to fill with what actually needs to be there - good or bad - things worth thinking and worrying about, not that I need to remember to return that email.

I've also been able to read more. The few days I benchmarked it, I average around 50K words on a light day - between slow-moving papers at ~450 wpm and faster less dense things at ~980. Somehow my sources seem to have gotten better, and I'm seeing fewer repetitions and more new things I simply didn't know about. A measurement of reading speeds has provoked a lot of conversation among friends and family, and made me realise that it's a number that's tied into some amount of self-worth. Strange. There's a write-up coming from Hebe on this that I'll link in once it's done!

I also want to thank a dear friend for gifting me his copy of A story of philosophy. It's been a wonderful read that I've been ignorant of, and I keep going back to re-read it.

§Goals for 2025

I'll start with something small. I'd like to understand my food more. One of the things I've always envied about a very close friend is his ability to look at his food and know what's in it - macros, calories, nutrients. It's now been a month of tracking and cataloguing everything, but I feel I'm slowly starting to build that internal sense - and it's made a big difference.

It's been hard trying to eat healthy with just a general sense of 'more leaves' or 'more green'. As I measure and track, I'm starting to understand the nuance a lot better.

Something medium-sized: Do more talks, and more videos. I talked about this before, but it's one of those things where doing it is the only way to get better, same as swimming. It feels like one of the big themes of this year has been communication, and better communication (the third rewrite was almost entirely around this theme). I'm still finding ways to get better at it, so I jump at the recognition that I suck at something around expressing myself better.

Off to a good start with this one!

Here's the big one: I'd like to really learn to say no. I don't mean the big No's in life - I've rarely had a problem with those. I mean the small ones, the ones where discomfort with inhabiting a 'no' space matters a lot more. More comfort in 'no' spaces means better relationships and less time spent in misunderstood 'yes' spaces - or that's one of my currently strong assumptions, weakly held.

§Cat & a little treat

Let me know if this is too meta. Claude helped me make this little tracker that you can paste directly into the browser console, and keep track of your writing at a keystroke level.

Here's what it looks like writing this version:

It's fun seeing the stops and starts, and the backspaces. Oh hey look there's zipf's law!

As always, thank you for reading. These reviews have gotten longer to write as the years have grown, with less time to write them in.

But if you're just here for the cat - thank you anyway!