Return of excitement
I am excited again.
My excitement usually comes in waves, and tends to crest when my interest in my job starts to enter a valley. When this happens, I throw myself in a bunch of different directions trying to find something that's more creatively fulfilling. Right now I have a handful of things I'm putting my energy into with the goal of finishing most of them by the end of they year (lol sure). Writing this is a partial plea for accountability.
Here's some things I'm excited about:
Building a couple apps for swimming
I've been tinkering on a couple of swimming apps for nearly 10 years, never actually finishing them. The closest I got was last year trying to finish one of them by the end of October, but it's still not done; I usually get distracted or bored. Claude code has dramatically changed this and my ability to quickly build things I want. The UI on my projects still leaves much to be desired, but I've been chipping away at two different small apps that I think I can get to the finish line in a month or two with a bit more focused effort.
The first is a swim meet & race tracker, in the same vein as our old Track app for running, with a few twists. Since many parents track their kids times, managing multiple swimmers was part of it. The second is comparing against published time standards around the world. Right now I only put in the US time standards for all age groups, plus global masters standards. This is a helpful way to track how you're doing. When I was in high school, we had the Ontario time standards pinned to the side of the fridge and I would look at before & after every swim meet. Sadly it never made me any faster.
The second app is a combination of a workout visualizer and generator. Part one is to give a better analytics interface to swim workouts from your apple watch. I tried building this a few years ago but my ability to create & design charts was terrible. But thanks to claude, I'm set. I've been doing some research into some more advanced analysis I might be able to do such as evaluate your turn efficiency and how your technique deteriorates over time. The workout generator is what it sounds like, leveraging something like ChatGPT as a swim coach to generate a specific workout that syncs to your Apple Watch to give you a set in real time while you swim. The generation capabilities need a bit more testing, but I haven't actually been swimming on my own in a long time.
AI & LLMs
A few months ago I read Dwarkesh's book about the scaling era of AI and felt compelled to develop a better intuition around large language models and what it takes to scale them. Since then I've been doing research and passively writing a long-form piece exploring some of the constraints to scale (data, compute, energy) and what directions seem promising for each of them. This is mostly about me building an intuition, but I'm also curious about tangible things that can be done to make an impact. The extended essay builds on a project some of my students worked on while I was at TKS where they looked at cutting data center costs for Microsoft.
As part of building my intuition I've also been trying to learn basic calculus (whatever is relevant for neural nets), go through Andrej's tutorials on building an LLM from scratch, and reading some of the seminal papers hoping to understand them a bit better.
All of this will hopefully feed into the piece I'm writing which currently sits around 4,000 words and still needs a lot of massaging and research.
The second AI project I'm exploring is the build upon a post I wrote in January about creativity and latent space. I'm curious to expand upon the idea of a ui or tool to more systematically explore options within a design space. The simplest way to think about this is some sort of tool to visualize "show me this logo cranked to 10/10 on professional, 4/10 on funny, 7.4/10 on creative, and 5/10 on uniqueness". Except it could be applied to anything, not just a logo. This is more of a "research" style project, something I can finesse with Claude into an interesting outcome.
I have a couple other things that broadly fall under the buckets of: building an ai portfolio + building small products, but those should wait til I finish these ones.
Btw, what are you excited about these days? Working on anything cool?
Currently reading
Tai-pan - by the guy who wrote Shōgun, tells a story (fictional history) of british traders settling the island of Hong Kong. It's got pirates, the triads, trading, and some political posturing for power. I'm just about done, but I enjoyed it. It's not as good as I remember Shōgun being (I read it ~15y ago), but still a fun read. There's also a short nasty passage about the practice of foot binding which had me squirming.
A Biography of the Pixel - written by Alvy Ray Singer, one of the technical cofounders of Pixar. An epic history of how we got to digital images from the science of waves, the origins of computing, and more. Super interesting. It tangentially touches on the other things I've been reading about early computers & AI.
--
As always, I'd love to hear back from you.
Steven