Am I making a mistake by switching startups?

Sorry, going to be a somewhat long post. I'm an MLE with roughly 2.5 YoE after a physics PhD (with a sprinkle of ML). I'm leaving my job at my current startup just after 8 months. It's been quite stressful, everything is on fire all the time and pretty much everyone is working overtime regularly (allnighters, weekends, etc). I'm physically exhausted. The job itself is also not so interesting to me. I'm a bit disappointed in the product, it started to feel to me it's some BS application of LLMs. A lot of prompt engineering, and building an app on top of it. Some fine-tuning, but it's also not inspiring, as you're mostly figuring out what dataset mixture to feed to an LLM. In the beginning I also thought it would be nice to learn more about "real" SWE, so I was picking up tasks in the backend. Even though it was nice I had an opportunity to do that, but the pace and lack of guidance made it very uncomfortable for me. I missed the days, when I actually worked on ML algorithms, implementing novel ideas from recent papers, doing training, etc. So another startup reached out to me about an opportunity. They are more research oriented and are working on optimization algorithms. My background fits well, I interviewed, I felt I clicked with the team, they extended me an offer, and I accepted it. With them I will be working on scaling up their ideas, experimenting with distributed training of big models, etc. Business-model wise it feels similar to what MosaicML has been doing. I'm much more excited about the new opportunity. WLB wise it seems like it will be better as well at least in the beginning, no crazy sprints, no daily standups for now. Initially I'll be working with researchers on a big project without a concrete deadline. The only thing is I feel it would be nice to work on SWE tasks, so that in addition to modelling/algorithmic development (which I feel I'm naturally more inclined to) I also have nice engineering skills under my belt. The startup I'm leaving was founded by ex-FAANG folks, they are quite smart and are very good SWEs, and I hoped I could learn from them, but alas. Versus in the new company I will be their first person on the applied ML SWE team. Engineering tasks are going to be down the line, but first I'll need to help them to achieve some milestone in training capabilities. I feel the new startup might be more risky long term, because it's in a more research field, but at least I will be working on something fun for me. However, not sure, how much riskier, maybe LLM hype will go down in a couple of years, and my current startup will join the likes of web3/crypto companies... I think I'll ride out a bad job market there, and then try to jump to a big company. Right now it's brutal, and I'm so exhausted from the pace at my current startup, I don't have energy to prepare for interviews. What do you guys think? TC at both: $200k cash + stock options.