We Scored 42,000+ Product Launches on Whether They're 'Vibe-Codeable'
October 5, 2026 · Nicheloom
"Vibe coding" -- building a working app mostly by describing it to an AI tool instead of writing the code by hand -- went from a meme to a real way people ship products this year. So we asked a blunter question: of the 42,000+ product launches Nicheloom tracks, how many of them could a solo developer actually have vibe-coded?
The short answer: 61% of the 42,436 launches we've assessed look buildable by a solo developer with today's AI coding tools (Cursor, Bolt.new, v0, and similar) -- 20% look partially buildable (the core is simple, something specific makes it harder), and 19% genuinely aren't. But that number swings hard depending on where the product launched: 64% of Product Hunt launches and 60% of GitHub Trending repos assess as vibe-codeable, versus just 22% of Y Combinator companies.
How we assessed this
Every launch gets a verdict -- yes, partial, or no -- from an LLM that reads the product's own description and its real, live website text, and judges whether the apparent scope and complexity is something a solo builder could put together with current AI coding tools in a weekend. It's a judgment call based on what's publicly described, not a technical audit of the product's actual codebase (we don't have access to that for most of these). Think of it the same way you'd size up a competitor from their landing page -- an informed read, not a guarantee.
We checked for the obvious bias before trusting this number: does the split change depending on when a launch was classified, or which version of the classifier touched it? Neither. Every launch in this analysis went through the same classifier version, and the yes/partial/no ratio has stayed within a few points of the overall average in every quarter since we started tracking. The gap below is a real difference between where products come from, not an artifact of when we measured them.
The split, overall and by source
| Verdict | Launches | Share |
|---|---|---|
| Yes -- vibe-codeable | 25,754 | 60.7% |
| Partial -- the core is simple, something makes it harder | 8,530 | 20.1% |
| No -- not realistically solo-buildable | 8,152 | 19.2% |
| Source | Launches assessed | Yes | Partial | No |
|---|---|---|---|---|
| Product Hunt | 23,994 | 64.1% | 21.6% | 14.3% |
| GitHub Trending | 9,607 | 59.6% | 14.6% | 25.8% |
| Hacker News | 7,782 | 56.8% | 20.3% | 22.9% |
| Y Combinator | 1,053 | 22.3% | 33.9% | 43.8% |
Y Combinator isn't just a little behind the other three sources -- it's a different category of product almost entirely. Fewer than 1 in 4 YC launches we've assessed look solo-buildable, and 44% are flatly not. That tracks: YC companies are funded to go after problems that need more than a weekend, on purpose.
What "no" actually looks like
The YC launches that score "no" aren't vague or hand-wavy about why -- the reasons are concrete, and they're the kind of thing no amount of AI-assisted coding shortcuts past:
Ruma Care (1,857 upvotes) -- helps clinics get patients onto biologic treatments faster. The blocker: navigating real pharmacy-benefit-manager prior-authorization rules and clinical criteria, not a coding problem at all.
Scalar Field (1,413 upvotes), an agentic trading desk -- needs low-latency execution across equities, options, prediction markets, and decentralized exchanges simultaneously. The hard part is infrastructure and risk management, not the UI.
Piggy Robotics (753 upvotes) -- humanoid robots. This one's not even a software question: mechanical, electrical, and control-systems engineering for physical hardware under $1,000 isn't something you vibe-code at all.
What "yes" actually looks like
On the other end, AIHOT (4,944 upvotes), a GitHub-trending news aggregator, assessed as vibe-codeable because it's a standard pattern -- fetch RSS feeds on a schedule, summarize, display -- that AI coding tools handle well today. And it's not only small or unfunded products that clear the bar: Brickwise, an AI property manager that came out of YC (3,090 upvotes), still assessed as "yes" -- a reminder that funding and ambition don't automatically mean something is hard to build, just that it usually is.
What this doesn't tell you
This measures how a launch reads from its public description, not whether it would actually be easy to build well, scale, monetize, or defend once it existed. A "yes" here means the apparent scope looks solo-buildable -- it says nothing about whether the market is any good, whether it's already crowded, or whether "buildable in a weekend" is actually an advantage once you're competing with a dozen other weekend builds of the same idea (see our opportunity scoring for that part separately). Treat this as a shortlist signal, not a verdict.
Frequently asked questions
What does "vibe-codeable" mean? Our own classification of whether a launch looks buildable with an AI coding tool (Cursor, Bolt.new, v0, and similar) in a weekend, based on its apparent scope and complexity -- not a guarantee, a starting signal.
What percentage of product launches are vibe-codeable? Of the 42,436 launches we've assessed, 61% look vibe-codeable, 20% look partially vibe-codeable, and 19% don't. The share varies a lot by source -- see the breakdown above.
Why are Y Combinator companies less vibe-codeable than Product Hunt or GitHub launches? Only 22% of YC launches we've assessed look solo-buildable, versus 57-64% for Product Hunt, GitHub Trending, and Hacker News. YC companies are funded specifically to tackle problems that need more than a weekend -- regulated industries, physical hardware, and infrastructure-heavy systems show up disproportionately in the "no" column.
Does "vibe-codeable" mean an idea isn't worth building? No -- it's the opposite concern if anything. A "yes" verdict means the apparent scope looks buildable by a solo developer, which also means plenty of other solo developers can build the same thing. Buildability and opportunity are different questions; check a theme's opportunity score separately.
How is vibe-codeable assessed? An LLM reads the launch's own description and its real, live website text, then judges whether the apparent scope and complexity is realistically solo-buildable with current AI coding tools. It's a judgment call from public information, not a review of the product's actual code.