Is vibe coding worth it? What the numbers say

Controlled trials show vibe coding accelerates fresh builds by 26% to 56%, and a six-month-old vibe-coding product sold for an initial $80 million plus earn-outs. Tools like Cursor and Claude Code have reached billion-dollar revenue run rates, and JetBrains' 2026 survey reports the highest developer loyalty scores for these tools. The data supports vibe coding as a viable method for solo builders to rapidly create working products.

The Study · Explainer Is vibe coding worth it? What the numbers say Yes: controlled trials measure 26% to 56% faster on fresh builds, the exits and revenue ramps are documented, and the two habits that protect the win cost an afternoon each. The verified case, with every vendor flag left on. Yes. For the job a solo builder hires it for, getting from idea to working product while the window is open, vibe coding is worth it, and 2026 is the year the receipts caught up with the demo. A vibe-coding product sold for an initial $80 million plus earn-outs https://www.wix.com/press-room/home/post/wix-further-expands-into-vibe-coding-with-acquisition-of-base44-a-hyper-growth-startup-that-simplif at six months old. The terminal agent half of the stack passed a $2.5 billion run rate https://www.anthropic.com/news/anthropic-raises-30-billion-series-g-funding-380-billion-post-money-valuation by February 2026. And in the freshest field data there is, JetBrains’ April 2026 survey of over 10,000 developers, the core vibe-coding tools hold the highest loyalty scores in the category https://blog.jetbrains.com/research/2026/04/which-ai-coding-tools-do-developers-actually-use-at-work/ . This piece is the case, made with verified numbers, plus the two habits the data says separate the builders who keep the win from the ones who hand it back. The short version The method makes real money. The exits, the revenue ramps, and the adoption inside the world’s most selective startup batch are documented below, with the deal terms read carefully. The speed is measured, and it peaks on your use case. Controlled trials put the gains on fresh, self-contained builds between 26% and 56%, and the famous slowdown result is about a different scenario, on 2025 tools, with the researchers’ own 2026 update pointing back up. The year-after bill is knowable. The best longitudinal code data shows where AI-era codebases accumulate weight, which means you can budget for it the way you budget for hosting. One afternoon of verification protects the whole asset. The security incidents in this category share a single root cause, and it is the cheapest thing on this page to fix. Everything below is the evidence, with the vendor flags left on. The receipts Start with the deal everyone quotes, because it survives a careful reading. Wix’s own press release confirms an initial consideration of approximately $80 million plus earn-out payments through 2029 https://www.wix.com/press-room/home/post/wix-further-expands-into-vibe-coding-with-acquisition-of-base44-a-hyper-growth-startup-that-simplif , which makes the famous number the floor of the deal. About $25 million of it went to employee retention, and the solo-founder, six-months-old, profitable-before-the-sale details come from founder Maor Shlomo’s own posts and TechCrunch’s reporting https://techcrunch.com/2025/06/18/6-month-old-solo-owned-vibe-coder-base44-sells-to-wix-for-80m-cash/ rather than the release. Trimmed to what the documents support, it is still the category’s defining receipt: one builder, one vibe-coding product, an eight-figure exit inside a year. The tools themselves, the same slots priced in the stack guide /primer/vibe-coding-stack/ , posted the same curve. Cursor reached $1 billion in annual recurring revenue https://www.cnbc.com/2025/11/13/cursor-ai-startup-funding-round-valuation.html by late 2025. Claude Code passed a $2.5 billion run rate by February 2026 https://www.anthropic.com/news/anthropic-raises-30-billion-series-g-funding-380-billion-post-money-valuation . Lovable crossed $100 million ARR eight months after launch https://techcrunch.com/2025/07/23/eight-months-in-swedish-unicorn-lovable-crosses-the-100m-arr-milestone/ , a company-disclosed figure, so season it. Money at that scale is a crowd of working builders voting with renewals, and the 2026 loyalty data says the votes are enthusiastic: JetBrains found Cursor and Claude Code tied for second place at work, with Claude Code carrying a 91% satisfaction rating and an NPS of 54 https://blog.jetbrains.com/research/2026/04/which-ai-coding-tools-do-developers-actually-use-at-work/ , the highest in the survey. The sharpest adoption receipt comes from the most selective room in startups. Inside Y Combinator’s Winter 2025 batch, managing partner Jared Friedman said a quarter of the startups had codebases that were 95% AI-generated https://techcrunch.com/2025/03/06/a-quarter-of-startups-in-ycs-current-cohort-have-codebases-that-are-almost-entirely-ai-generated/ , remarks from YC’s own discussion rather than published data. These are funded, technical founders who could build any way they want, and one in four builds this way. Across the wider field, 80% of the 49,009 developers https://survey.stackoverflow.co/2025 in Stack Overflow’s 2025 survey use AI tools at work. The method is the mainstream now. The speed, measured The controlled evidence is strongest exactly where a vibe coder lives: new code, fresh project, defined goal. In a controlled experiment, 95 developers built an HTTP server from scratch and the group with AI finished 55.8% faster https://arxiv.org/abs/2302.06590 , with a wide confidence interval 21% to 89% and Microsoft-affiliated authors, so read it as a strong signal with a vendor flag. The largest field evidence is calmer and still emphatic: across 4,867 developers at Microsoft, Accenture, and a Fortune 100 firm https://papers.ssrn.com/sol3/papers.cfm?abstract id=4945566 , AI access produced about 26% more completed tasks, with the biggest gains for the least experienced, per the paper’s own characterization. Newer builders gain the most. That is the vibe-coding demographic, described by peer-reviewed research. The result skeptics quote is METR’s randomized trial https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/ , where experienced open-source maintainers took 19% longer with AI, and it deserves its context: sixteen experts, working on mature repositories they knew intimately, using early-2025 tools. That is the single scenario furthest from a solo builder shipping something new, and METR itself warns against generalizing beyond it. More telling for 2026: METR’s February 2026 follow-up https://metr.org/blog/2026-02-24-uplift-update/ reads the newer data as weak evidence that developers are more sped up now than in 2025, with the honesty to flag its own selection-bias problems. The slowdown was a snapshot of early tools in the wrong scenario; the trend line, on the researchers’ own reading, points the right way. The full anatomy of the split is its own piece /study/does-ai-coding-make-you-faster/ ; the version that matters here is one sentence long. On the work vibe coding is for, every controlled measurement says faster. The year-after bill, and how to budget it Winners plan for the whole cost of the asset, so here is the part the launch threads skip. GitClear, a code-analytics firm that publishes its raw tables, tracked 623 million changed lines through mid-2026 https://www.gitclear.com/the ai code quality maintainability gap and found the AI era’s codebases gaining weight: moved code, its proxy for refactoring, fell from 21% of changes in 2022 to 3.8% in the 2026 data, copy-paste rose from 9.4% to 15.7%, and code rewritten within two weeks of landing is up 15% against the 2023 baseline. Carry the flags: GitClear sells analytics, the data is correlational with the AI era, and an independent reviewer notes the 2022-2024 layoffs as a possible confounder. For a solo builder this is a budget line, and a known one beats a surprise. The prototype that validates your idea costs a weekend. The product that survives will ask for cleanup passes, and the data above tells you which kind: deduplication, structure, the pruning the model skipped while it was busy shipping you a working app. Schedule it like an oil change and the compounding never gets ahead of you; the subscription side of the budget stays tracked in the price index /palette/ai-tool-price-index/ . The builders who treat the codebase as disposable until the idea is proven, then treat it as an asset the day money arrives, get the best of both curves. The one afternoon that protects everything The security record in this category has a single lesson, and it is cheap. The headline incident is CVE-2025-48757 https://nvd.nist.gov/vuln/detail/CVE-2025-48757 , rated 9.3: Lovable-generated apps shipping without row-level security, their databases readable and writable from the open internet. The disclosure timeline https://mattpalmer.io/posts/2025/05/CVE-2025-48757/ was rough, vendor notified March 21, 2025, fix shipped in early June, and two flags belong on the record: Lovable disputes the CVE’s framing, holding that app data is the customer’s responsibility, and the discloser works at Replit, a competitor, though NVD and independent firms corroborate the technical facts. The ecosystem moved: the pipeline was fixed, and the platforms have been hardening ever since. In October 2025, security vendor Escape scanned 5,600-plus vibe-coded apps https://escape.tech/blog/methodology-how-we-discovered-vulnerabilities-apps-built-with-vibe-coding/ and found 2,000-plus vulnerabilities and 400-plus exposed secrets, vendor research with a disclosed method, and a useful census of what unreviewed apps leave open. The pattern in every incident, including the builder who told his followers his SaaS was built with zero hand-written code https://x.com/leojr94 /status/1901560276488511759 days before strangers maxed his API keys and bypassed his paywall, is the same: the ship happened and the check never did. A Stanford experiment https://arxiv.org/abs/2211.03622 saw it in the lab years before the term existed, with AI-assisted participants writing measurably less secure code. None of it is an argument against the method. It is the price list for skipping one afternoon: turn on row-level security, keep secrets out of the client, run one security pass before real users arrive, or pay someone one contract to do it. Against an asset that can sell for eight figures, that is the cheapest insurance in this economy. The verdict Worth it, and say it without flinching: the numbers back the method. Andrej Karpathy named it https://x.com/karpathy/status/1886192184808149383 as a way to build where you give in to the vibes and forget the code exists, and eighteen months later the scorecard reads: an $80 million exit, a quarter of YC’s batch, billion-dollar revenue ramps on the tools, the category’s best loyalty scores, and measured speed gains of 26% to 56% on exactly the work solo builders do. Simon Willison’s distinction https://simonwillison.net/2025/Mar/19/vibe-coding/ supplies the operating manual: vibe code to ship and validate, then review what won, the same supervised-beats-unattended split the agent research keeps finding /study/do-ai-agents-work-yet/ . The two habits above, a scheduled cleanup pass and one security afternoon, are the entire difference between the receipts at the top of this piece and the incidents in the middle. The field has already voted: 80% of developers use the tools https://stackoverflow.blog/2025/12/29/developers-remain-willing-but-reluctant-to-use-ai-the-2025-developer-survey-results-are-here/ , and the ones on the best tools report the highest satisfaction in the category. Build the thing. Read it before strangers do. That is what the numbers say. One email, when there's something worth sending Get the receipts in your inbox. No fixed schedule, no filler. You get an email when we've tested something, run the numbers, or found a tool worth your time. Free. Double opt-in, unsubscribe in one click. Did vibe coding pay for you? Compare notes in the forum ↗ https://community.okaneland.com Sources & how we researched this - Wix press release June 2025 , acquisition of Base44: ~$80M initial consideration plus earn-outs through 2029. wix.com/press-room - TechCrunch June 2025 , '6-month-old, solo-owned vibe coder Base44 sells to Wix for $80M cash.' techcrunch.com - TechCrunch March 2025 , 'A quarter of startups in YC's current cohort have codebases that are almost entirely AI-generated' Jared Friedman remarks . techcrunch.com - Anthropic Feb 2026 , Series G announcement: Claude Code past a $2.5B run rate. anthropic.com/news - CNBC Nov 2025 , Cursor reaches $1B in annual recurring revenue. cnbc.com - TechCrunch July 2025 , Lovable crosses $100M ARR eight months after launch company-disclosed . techcrunch.com - JetBrains research April 2026 , 'Which AI coding tools do developers actually use at work' 10,000+ developers . blog.jetbrains.com - Stack Overflow 2025 Developer Survey 49,009 respondents and results blog post. survey.stackoverflow.co/2025 - Peng, Kalliamvakou, Cihon & Demirer 2023 , The Impact of AI on Developer Productivity: Evidence from GitHub Copilot. arxiv.org/abs/2302.06590 - Cui, Demirer, Jaffe, Musolff, Peng & Salz 2025 , The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments. Management Science / papers.ssrn.com/sol3/papers.cfm?abstract id=4945566 - METR 2025 , Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. metr.org / arxiv.org/abs/2507.09089 - METR Feb 2026 , Uplift study update. metr.org/blog/2026-02-24-uplift-update - GitClear 2026 , The AI Code Quality Maintainability Gap, 623M changed lines 2023-2026. gitclear.com/the ai code quality maintainability gap - GitClear 2025 , AI Copilot Code Quality research, 211M changed lines 2020-2024. gitclear.com/ai assistant code quality 2025 research - Matt Palmer May 2025 , CVE-2025-48757 disclosure advisory. mattpalmer.io discloser is a Replit employee; affiliation noted in the piece - NVD, CVE-2025-48757 record CWE-863, CVSS 9.3, vendor-disputed . nvd.nist.gov/vuln/detail/CVE-2025-48757 - Escape.tech Oct 2025 , vulnerability scan methodology across 5,600+ vibe-coded apps security-vendor research . escape.tech/blog - Perry, Srivastava, Kumar & Boneh / Stanford ACM CCS 2023 , Do Users Write More Insecure Code with AI Assistants? arxiv.org/abs/2211.03622 - Andrej Karpathy Feb 2, 2025 , the original 'vibe coding' post. x.com/karpathy/status/1886192184808149383 - Simon Willison Mar 2025 , 'Not all AI-assisted programming is vibe coding'; and Oct 2025 , 'Vibe engineering.' simonwillison.net - Leonel Acevedo / @leojr94 March 2025 , Enrichlead build and attack posts first-person, archived . x.com/leojr94