So far this decade has seen quite a whirlwind of developments. From the pandemic at the beginning of the 2020's and the ensuing lockdowns things seemed to be very different everywhere and I think we all wondered if we would ever return to a sense of normalcy. I'd say slowly we did, but it took the vaccines and a good amount of time as it still seemed like we were dealing with lots of mutations of covid even after the peak. Now I feel like we don't think about it as much but I still keep up on my vaccinations anyway.

Another interesting development has been the creation of ChatGPT and LLMs more broadly. I've been pretty impressed with what LLMs have been able to accomplish. From the beginning (around the end of 2022) I seems like it has come a long way in a very short amount of time. In the beginning it was kinda basic but still seeing it come up with novel answers to all sorts of queries has been fascinating. Then when image models first came about they were also very basic and would have strange artifacts, to now when we see almost indistinguishable (from real) images being generated by the likes of nano banana.

Meanwhile, AI video generation also started out almost laughably bad but now it's actually getting pretty good. I think the tool use is what has impressed me the most in the past year. Previously AI had been useful for writing tests and some light scripts and things like that, but it was always difficult to get the right context for these tools to take on bigger tasks. It can still be difficult, but I think with things like Claude Code the tooling around working with LLMs has gotten a lot better so that given a query it can find the relevant context on its own and develop a plan to solve trickier problems.

For me, I had been working at Credit Karma for about 7.5 years and learned a ton as well as worked with some really great people during my time there. But eventually I wanted change and to continue learning about new things in different companies, so I left and went to Zip and then to Gamma.app where I am currently (as of writing this). I knew I wanted to work on products with AI infused into them and I think Gamma fits that bill well. I've slowly evolved my coding workflow over the past 6 months, I used to write most of my code and use LLMs for adding some tests and such, but now I've basically converted to having at least 2-3 Claude Code sessions going in different git worktrees making progress on various tasks which has been a bit of a transition for me but so far I'm enjoying it.

Additionally, and I didn't really realize this before, but it is very interesting and rewarding to work on a creative tool like Gamma. Seeing the things people create and try to do when given such a free form tool has been quite amazing, people make some really cool stuff and it's fun to see. Sometimes it's funny, sometimes it's kinda strange, and sometimes it's just very impressive. A few of my favorites that I can easily share are: https://lobster-launch-n3ehg0e.gammastaging.site/ and https://bonanhale.com/. Maybe I'll share more if there's interest.

Finally, I'll end with some predictions for 2026 and beyond. It seems like LLMs have reached some level of diminishing returns with pure model parameter count, which I guess should be expected, nothing goes on forever. But I'd expect the next set of gains to come from better orchestration, and I think we are already seeing this. Think smaller more focused models and models that can federate their answers out to other more specialized ones for certain queries. I'd also expect tool use for AI to continue to get much better, it feels like we are just scratching the surface. Another development I'd expect is getting LLMs that have persistent memory, I think we will see that become more common in the next year or two. This should help with context issues and make these systems more useful and relatable. And robots are another area where it seems like progress is being made rapidly. In the past they had been a bit rudimentary, but there are so many companies working in this area and the potential uses are very broad. I imagine we might look back on this time with a certain amount of nostalgia and wonder as I feel like things will likely be very different in the next 5-10 years.

The rate of change currently seems quite fast and it feels like many companies are racing towards visions of the future that feel like they are out of a Sci-Fi novel. From AI, and robots, to drones and flying cars, on top of all the numerous medical advances which seem promising. However, it does also seem like a bubble could be brewing and if it bursts could be quite disruptive. I'd say in another year or two we could see something pan out, if it takes longer it will likely be much more impactful but either way it won't be good. Still, I'm optimistic at this point that even if there is a bubble things will pan out in the long run, this is a very transformative and useful technology that I think will have a major impact on society for the foreseeable future, I'm just not sure that all of it will be good but hopefully most of it will be.