What's the next frontier for AI Coding?
At this point, all of the AI coding assistants are in the same neighborhood. Decent at "advanced autocomplete", OK at code generation sometimes, and most are somewhere in the process of incorporating code context mechanisms. But what's next? Agentic behavior? Something else?
My pet prediction is that we will see the emergence of a new programming language that's designed for use with AI and can be translated to a variety of popular languages. (Or if we're cursed, just javascript š)
Replies
Have you used Windsurf much? I'm basically acting as a product manager and telling it what to do and then refining its work. That seems far beyond "advanced autocomplete".
It would be interesting if a new pidgin-creole language were developed that was more like pseudocode ā where you could vibe your way to a refine product by intermixing programming and product concepts... perhaps intent-driven coding? (i.e. explain what I want it to do and how to evaluate the outcomes versus describing how to achieve those ends?)
Product Hunt
@chrismessina Yeah, that's I guess what I meant by "code generation". They all do that more or less. Seems to be a matter of preference (echoing text editor wars) which is better at what with no clear consensus emerging just yet.
And yeah, you can basically just write pseudocode and it'll do it now. But I mean something more formal. Very high-level, but structured. Intent-driven is a good way to phrase it. Lets you be conceptually precise but without bothering with all the details. I.e. "on message:arrival user.notify(slack, push, email, desktop)"
Alta
AI coding tools are solid right now: great at autocomplete, decent at generating snippets, and kinda figuring out context. But the real next step? AI that actually works like a dev.
Instead of just spitting out code, itāll write, debug, and optimize with less hand-holding. Maybe even a new programming language built for AI collaboration. Or, you know, just more JavaScript. š
Product Hunt
@jawad_tijani
Manicode does the code-test-fix loop, though it can get into the weeds.
Product Hunt
This might not be the next frontier but I'm looking forward to AI that can consume context and information from more and more sources. Right now most AI code tools are pretty siloed to looking at the code and maybe pulling things from the web. It'd be great if they could inspect databases, peruse slack or parse production logs to better understand the problem I'm trying to solve or the feature I'm trying to build.
Product Hunt
That's a really interesting thought, I haven't considered a new programming language for interacting with AI. Maybe on a similar thread I think we're going to see a new way folks interact with their devices, including development work.
We had to modify how we think about questions for Google, and now we're going to have a new way to input queries or request into @Perplexity, @Cursor, @Windsurf, or really anything like @ChatGPT by OpenAI. While not 100% a new programming language, I do think the way our technological behavior is being impacted will change how we "talk" to our devices, we're already seeing things with prompt engineering.
This is an exciting space. Iām not building for dev workflows directly..but I think the parallel shift is happening in team workflows.
Weāre building an AI Executive Assistant that handles email triage, prep, and comms routing ..and weāre seeing similar questions come up:
ā How much should AI guess?
ā How much should it explain?
ā How much agency do users want?
Curious if you think dev tools will need to show their logic more too.. or stay āmagicalā?