I work in AI text detection, so I spend most of my time on the content side of this question: whether a paragraph can be traced back to a model. Vibecoding asks the same thing one layer down, about code, and the answers people give are completely different.
Nobody labels a commit. There's no convention for it, nothing in a README, and the honest reason might be that a label wouldn't survive anyway. Once generated code has been reviewed, renamed and refactored into the rest of the codebase, whatever was distinctive about it is gone. One editing pass does the same thing to a watermark in text.
So does provenance matter for code, or only for content?
Where I've got to so far, and I'd change my mind easily on this:
We run four providers behind one chat and they all format history slightly differently, so a lot of our code is adapters. Adapters are exactly the kind of thing you hand to an agent and never look at again.
Here's the problem. If the same session writes the adapter and the test for the adapter, the test asserts the shape your code produces. Not the shape the provider actually sends. It passes on day one and it passes forever, because it was never checking upstream, it was checking that your code agrees with itself.
I ve made a bunch of actually useful web apps recently using lovable.dev and v0.dev, and here s the thing: I don t know how to code. At all. No CS background. No tech degree. Just a regular person scrolling through tools and products like everyone else.
But around 3 months ago, I gave v0.dev a try... and I was genuinely shocked. Not just by how well it worked, but by how fast it s improving. It s kind of scary in the best way.
Sure, it can t build a full app from start to finish yet... But take a second to think about it This kind of thing didn t even exist 2 years ago. At the pace it s moving? Give it 1 year and solo builders will be able to do some absurd stuff.
We now look at actual code less and less. What does your developer experience look like now that we are getting closer to a non-IDE world? I am using @Superset and am loving it so far. In my corporate job, I have 10-15 repos going at once and it's all super organized. What are you all using?
I don't have a CS degree. Never shipped a product. Never started a company. One month ago I didn't know what a Next.js route was.
I built Four-Leaf.ai, an AI career prep platform with voice mock interviews, resume tailoring, and negotiation coaching. It's live, it has users, and I launched it on Product Hunt today.
If you use any of the current coding agents you've seen it: you ask where something lives in the codebase, the tool announces it's "spawning an explore agent", goes quiet for a minute, and comes back with a tidy two-sentence answer. The feature usually gets explained as teamwork, which hides the useful part.
The clone is a disposable context window. Your agent's memory is finite and expensive everything it reads stays in its working memory and gets re-billed on every turn after. So instead of reading 60 files into its own head, it forks a copy. The copy reads the 60 files (59 of which turn out to be dead ends), reports back two sentences, and gets deleted. The dead ends never touch your main session. That's the entire feature, and most of what people call "multi-agent" is this trick applied repeatedly.
The catch: the clone knows nothing you didn't write down. It can't see your conversation. The library you rejected an hour ago, the API version you're pinned to, the thing your teammate said in standup all invisible unless the delegation prompt spells them out. Which is why fan-out works great for reading and searching, and keeps failing for parallel writing: two clones writing code make small conflicting decisions about naming, style, and edge cases, nobody wrote the assumptions down, and the merge is where it dies.
Cost reality check, since this is the part nobody quotes: Anthropic's engineering blog put agents at roughly 4x the tokens of a chat session, and multi-agent systems at roughly 15x. Parallel intelligence is something you buy. A deep audit of an unfamiliar codebase can be worth 15x. A typo fix can't.
I m curious what you all devs and founders are relying on day-to-day in 2025. With the flood of new ai tools, it feels like every tool looks different depending on industry and workflow.
What s ai tool working well for you right now?
Which AI tools actually save you time?
Which ones did you try but drop?
Would love to see how other folks are stacking their tools this year.
