Ask HN: What's the ideal stack for a solo dev in 2025 I’m a marketer turned dev. When I built my first SaaS 3 years ago, I went with ROR because it seemed like the ideal stack for a solo dev. But with all the development in AI, I wonder, is it still the best choice? Most AI tools have JS or Python integration but the Rails ecosystem seems a bit behind. But at the same time, a lot of the JS ecosystem feels all over the place. So what tech stack should a solo dev use in 2025? |
The ideal tech stack for a solo dev to use in 2025 is also the one that is most suitable to the problem that you are trying to solve.
Unless those two are wildly different from each other — such as one being Ruby and the other being Haskell — learning the same concepts that you already know in a new language takes a few weeks at best.
Personally I picked up Python for a couple of small LLM demos that I have given, but ironically the sample application is otherwise written in .NET (because I am most comfortable hacking GUI applications with .NET), with the Python parts being invoked as a subprocess.
When you hit a problem ask people (most likely still doable as it's just REST api endpoints) and if not, possibly re-evaluate.
At worst you'll know if your idea is worthless (probably lol) and if the tools you used were good or not. Rinse-repeat.
Good luck!
Both are totally valid, but they're somewhat orthogonal.
For the uninitiated, Laravel is an opinionated, "batteries included" stack that lets you get right to the meat of the application without spending time working on authentication, routing, html templating, and so much more.
There are also several (paid) resources the organization provides that allow you to get your app hosted pretty quickly (see Forge and Cloud).
HTML/CSS/JS won’t go out of style (could use templates like Askama, utility classes like tailwind to speed things up)
Axum is a nice option for backend because it’s much less buggy than the Python/JS stuff
Redis / Valkey for a KV cache
PostgreSQL database
If all your integrations with AI are via HTTP API's then it does not matter what stack you choose, they all can call API's. Maybe some stacks will have pre-built libraries but really, AI like cursor, copilot or chatgpt can create an AI HTTP API library if that's what you're doing. If you are building AI/ML yourself then you'll have narrower choices of stacks.
If you choose a stack that is widespread and has been around for years, like ROR or python/django, then chances are AI will be better trained to be your helpful assistant that is sometimes wrong.
I build most projects in Elixir/Phoenix these days, but wouldn't flinch at ROR if you are comfortable with it.
There's lots of libraries for calling out external AI services. e.g. something like FastMCP[2]. From the sound of it that's all you need.
I'd use Hotwire[3] for the frontend and Hotwire Native if you want to rollout an app version quickly. I'd back it with SolidCache, SolidQueue, etc
I'd use Kamal[4] to run it on cheap hosting using on something cheap from Hetzner.
Indeed. Not everything even has to be a "web app" in the first place.
The language you use has nothing to do with your ability to participate in "AI". Use what you know. 99% of working with this technology is making a JSON HTTP call to a chat completion endpoint. If your ecosystem supports HTTP and strings, you should be good to go.
Inevitably, your AI will get stuck and you will have to roll up your sleeves and dig into the problem. In what framework/language are you most quickly able to read (likely overly-verbose) AI code and dig into the root issue? The one you're most familiar with.
If you’re calling out to AI tools, you’ll likely use worker processes and these can be in whatever language you want.
Don’t fall into the trap of selecting the perfect stack. Just get started and iterate. You’re already working against the odds enough.
I'm really burnt out by the current state of JS, and the way rails 8.0 does server-side rendering feels like a complete breath of fresh air. The ability to just add slight JS functionality with stimulus and action cable seems like just the right level of abstraction for most web apps.
If you are doing full stack development, this works much better. The npm ecosystem is amazing. You can create common libs, share types, don't have to context switch and all future developers can do some full stack.
This is the boring stuff, but works and scales.
Generally speaking, typescript everywhere. LLMs know it quite well. If you are doing agentic work, you can do a FastAPI encapsulated python service.