Startups across the technology landscape are wagering substantial resources on the emergent phenomenon known as “vibe coding,” and their recent financial footprints reveal just how strongly they believe in its potential. In a compelling collaboration, venture capital powerhouse Andreessen Horowitz—better known as a16z—joined forces with Mercury, a cutting-edge financial technology company that provides digital banking and payment infrastructure for startups. Together, they conducted a comprehensive analysis of transactional data drawn from a vast sample of more than 200,000 customer accounts during the period spanning June through August. The resulting report, published on Thursday, offers an intricate view into how emerging companies are channeling their spending in the artificial intelligence space, pinpointing the top fifty AI-native applications based entirely on this observed expenditure data.
Among the most noteworthy findings, a16z observed an unmistakable shift toward enterprise adoption of what it calls “vibe coding”—a new paradigm in software development in which users generate functional programs through conversational prompts or natural-language instructions rather than through traditional, line-by-line programming. In other words, it allows anyone—whether or not they have a conventional computer science background—to construct working software simply by articulating what they want an application to do. Such tools, exemplified by leading platforms including Replit, Cursor, Lovable, and Emergent, all secured positions within the upper echelon of the fifty most heavily used AI-native products. Replit, in particular, distinguished itself by ranking third overall in total spending by Mercury’s customer base, surpassed only by AI giants OpenAI and Anthropic. The authors of the a16z report emphasized that “vibe coding is no mere consumer trend—it has firmly established itself in professional, organizational contexts.” They further noted their curiosity about how the field will evolve over time, speculating whether it may eventually diverge into specialized subdomains offering tailored environments for particular classes of application building.
The extensive dataset also illuminated another pattern: a majority of the leading AI tools—approximately sixty percent—were horizontal in nature, meaning they were built to serve a broad range of users across business functions rather than being confined to isolated technical specialties. These general-purpose tools may include everything from AI meeting copilots to productivity-focused digital assistants. The remaining forty percent of top performers fell into vertical categories, created for narrowly defined professional or industrial applications. a16z explicitly placed creative software like Canva and voice-synthesis platform ElevenLabs, as well as the major vibe coding tools, into the horizontal group. Historically, creative software suites were the realm of marketing and design departments, while software development utilities were reserved almost exclusively for engineers. Artificial intelligence, however, has disrupted that division by democratizing access—making tasks like design, content creation, and even software development available to employees across departments regardless of their technical specialization. The report explains that AI has effectively opened up previously domain-restricted capabilities, enabling them to become accessible and functional for individuals in essentially any role within an organization.
Beyond the data, a16z declined to offer additional comment to Business Insider, which reported on the findings. However, the cultural resonance of vibe coding continues to multiply. Once a niche phrase within select AI and developer circles, it has rapidly become one of Silicon Valley’s most persistent buzzwords. The term itself first emerged publicly in February, when OpenAI cofounder Andrej Karpathy popularized it by describing vibe coding as “a new kind of coding” that involves an almost intuitive collaboration between human intent and AI generation—where developers “surrender to the vibes” and in many cases “forget that the code even exists.” This description captures the essence of a paradigm that prioritizes high-level conceptual thinking over syntactical precision.
Today, that once-experimental notion has transformed into a marketable competency. Leading global corporations—from Visa and Reddit to DoorDash—have incorporated vibe coding knowledge or proficiency with AI-assisted coding platforms into their hiring criteria, acknowledging its growing relevance to modern software workflows. Even Meta, one of the world’s most influential technology firms, announced in July that it would permit job candidates to consult an AI assistant during technical coding interviews, a policy shift reflecting the ubiquity of these emerging tools in professional practice.
Capital investment trends reinforce this technological momentum. In July, Anysphere, the company that created Cursor, secured a massive Series C funding round of nine hundred million dollars, achieving a staggering nine-point-nine-billion-dollar valuation. Similarly, the Swedish startup Lovable, which also specializes in vibe coding platforms, successfully raised two hundred million dollars in Series A financing, reaching an impressive valuation of 1.8 billion dollars according to PitchBook data. These figures highlight the enthusiastic confidence investors place in the persistence and profitability of this development model.
Yet, even amid soaring optimism, the technology’s limitations remain evident. While vibe coding tools can significantly accelerate productivity and empower non-technical users to prototype and implement functional applications rapidly, many executives caution that current AI systems still exhibit fallibility. These systems may produce inaccurate code, introduce unnecessary complexity by generating verbose or inefficient outputs, or fail to align with a project’s optimal architectural foundations. Rowan Trollope, CEO of the software company Redis, succinctly captured this cautionary sentiment in a Business Insider interview, noting that AI-assisted development, though promising, “is still not in a place yet where we would trust it with our core technology.”
Taken together, the findings and commentary reveal a fascinating juncture: vibe coding stands simultaneously as a democratizing force, a venture capital magnet, and a yet-imperfect evolution in how humans instruct machines. If current momentum persists, the concept could redefine not only who codes, but what coding itself means in the age of ubiquitous AI.
Sourse: https://www.businessinsider.com/a16z-startup-bank-mercury-data-spend-vibe-coding-tools-2025-10
