New year’s resolution for me was to post something once a month. Doesn’t matter what, ideally something meaningful / useful / tangible, but whatever.

The theme of January for me was a lot of agent-fueled ADHD. With the improvements in agent coding, my “productivity” has gone up a lot. I redid some old side projects and worked on several new ideas. It’s really amazing how something that took me multiple weekends and nights can be banged out in an hour or less now with zero manual coding. However, most of the new projects are still totally halfbaked. But what should I expect from almost zero effort? I imagine in a year from now, these will be 80% baked if not full-baked with the same level of effort.

I grew up on the mantra that ideas are worthless and execution is what matters. So I never valued any ideas I had unless I could implement them. And this unfortunately meant I started filtering ideas out of my head for sheer implausibility. I wouldn’t blog about 5 random side project ideas I had without actually doing them, that would seem silly. But now I can basically do just that because I got something workable out of most of them this month. They’re not polished and definitely could not be released to the public meaningfully, but something has clearly changed now because I derive genuine utility out of these tools.

Things this month:

• rocket money alternative, give it access to amazon so it can figure out my purchase categories on there
• surf forecast refactor from an app i made years ago
• timeline of shifting European relationships, inspired by reading about the Napoleonic Wars (available at http://blog.austinsteady.com/european-relationships-visualizer/)
• tool to scrape X and avoid the anti-bot measures
• redo a tool to book campsites for me after I had abandoned the old one after some new anti-scrape measures
• rank all the books mentioned by Tyler Cowen in his blog using LLM for sentiment analysis, build a basic search on top of it
• volleyball analysis app. probably the most involved, I built a labeling pipeline and trained a model to detect events within a beach volleyball match. abandoned it when someone on Twitter produced a better version
• supply chain analysis tool

Anthropic just published a study showing we’re not more productive with agents. I could believe it I guess. My peak coding ability was a few years ago when I was actively working on real side projects every day before and after work. I basically had the pandas API memorized somehow and could solve most LC mediums in under 5 minutes. The list of above projects maybe all could’ve been doable within a single day if I really tried? But the thing is - I don’t have to try now. I got a very viable POC with minimal effort. And I get to focus on the fun part. The volleyball app really is involved and you have to be thoughtful with how you label the data. But I didn’t have to waste hours on stupid playback issues or obscure off-by-1 bugs that would ruin my labels in some nefarious way.

I do hate how distracted my mind has become. The accessibility of the impression of productivity is very hard to ignore. There’s also something depressing and deflationary about the AI improvements in that I really have no interest in working too hard on a project, because I know it’ll be much easier to do in 6-12 months. In fact, that was the case with the volleyball app. A year ago, I had to use roboflow and/or write my own labeling tools. I spent a lot of time doing stuff other than training the model. Now, it was pretty trivial to get a decent model going. In a year, AI might be able to one-shot the project.

So a negative effect is that I feel much less focused. I am trying to be more intentional with deeper reading periods, but also as a personal matter, this is becoming harder to do with a very talkative partner and will be much harder in the future with kids. Maybe deep thinking won’t be a relevant skill in the future, but I feel too cluttered mentally without it.

I sometimes think about how people say they used to have to go to a library for a book report because they didn’t have the internet. I’ve always thought that was barbaric and I’m so grateful for the amount of information immediately available to me. Now that I am aging out of being cool myself, I find that there is something lost with how easy AI makes it to get an answer. I wrote my book because I was obsessed with knowing the details down to the lowest layer possible of how exercise works. It’s depressing to think how quickly I could “write” that book now with AI. But that period of time when I was researching for the book was one of the most satisfying in my life. I hope it’ll remain important and available to me to maintain that kind of deep thought in the future.