If Vibe Coding is valued in Billions; what is the PSL/i value?
July 9, 2026
Vibe coding is persuasive because it gives people something they can understand quickly. A person describes an application in ordinary language, and the system turns that description into something that looks and behaves enough like software to make the idea feel real. Anyone who has waited on an engineering queue, or watched a small internal software idea get dragged through meetings, revisions and scheduling, understands the pull of it. The person with the need gets closer to the thing being built.
The valuations follow from that pull
Replit announced a $400 million Series D in March 2026 at a $9 billion valuation, which it said was three times its valuation from six months earlier.[1]
Lovable has reported enormous revenue growth, with TechCrunch reporting that it crossed $400 million in annual recurring revenue in early 2026, and later reporting that it reached $500 million in annualized revenue.[2]
Forbes reported in June 2026 that Lovable was in talks to raise money at a $12 billion valuation, up from a reported $6.6 billion valuation in December.[3]
Wix bought Base44, an AI app-building platform, for $80 million in cash in June 2025.[4]
Figma’s July 2026 acquisition of the team behind Bud, formerly Orchids, belongs in the same conversation, because Figma is trying to stay close to the point where product intent becomes working software.[5]
The market is not paying those prices because the current products make nice demos. The market is paying for the possibility that vibe coding becomes a new starting point for software creation. If a founder, an operator, a designer, a product manager, or a small business owner can create usable software by describing what is needed, the old path from idea to application changes. The old process does not disappear, especially inside serious companies, but it no longer controls the first move. The first move is what investors are buying.
There is value in giving more people a direct way to build, and there is value in shortening the distance between knowing the work and shaping the tool. Most software starts because someone sees a process that does not work well enough. When the person who understands the problem can get much closer to building the answer, the economics of software creation change.
But seeing an application appear does not prove the application can be depended on. To me, it proves something far narrower. It proves that an idea can be turned into a working artifact, which is useful. It does not prove that the artifact belongs inside a business, inside a regulated process, or inside a system where people rely on the result. A demo can make the surface believable while leaving the much deeper questions unresolved.
Those questions are where the real test begins. Who approved the workflow? What information is the tool allowed to use? What systems is it allowed to change? What happens if it misunderstands the user, or if the same user gives a different instruction next week, or if one generated tool conflicts with another? Who is responsible when the tool takes the wrong action? None of that is abstract. Those are the questions that show up when software touches customers, records, money, contracts, employees, or operations that cannot simply be undone.
The research around prompt-to-app systems points the same way. One 2025 paper described vibe coding as a shift from deterministic instruction toward probabilistic inference, which is a more formal way of saying that the machine is now interpreting intent rather than executing a precise specification. That opens the door to faster creation, and the same paper identified risks such as black-box codebases and responsibility gaps.[6] A separate 2026 paper on design-to-code systems found that proprietary models can achieve stronger visual fidelity while remaining limited in layout responsiveness and code maintainability.[7] Those limits are exactly what a good demo hides.
The valuation question gets more interesting from there. If vibe coding is valuable because it may change how software gets created, the next question is what makes that created software usable beyond the first impressive moment. Once creation gets cheaper and faster, the harder problem moves. It moves to whether the generated thing can operate with the right authority, the right memory, the right limits, and the right connection to the person or organization using it.
This is the logical opening for PSL/i. A Personal Semantic Layer or Organizational Semantic Layer architecture would not be another prompt-to-app tool sitting beside Replit, Lovable, Base44, Bolt, v0, or Figma Make. It would sit underneath generated capability as the layer that gives it context, boundaries, and continuity. Vibe coding can help create the functional object. PSL/i would help determine what that object means, what it is allowed to touch, what requires approval, what becomes durable, and how it fits into the operating life of a person or organization.
The difference is where each one starts. Vibe coding begins with the prompt. PSL/i begins with the person or the organization behind the prompt, carrying the history, permissions, commitments, domain meaning, and operating context that a generated tool does not naturally possess. Without that foundation, a generated tool may still be useful, but it stays local to itself. It runs on its own assumptions and its own interpretation of what the user meant. PSL/i would give that capability a place to live and a structure it does not get to invent for itself.
Inside that structure, vibe coding could become more useful. It could help generate or extend lenses. A lens is not a simple workflow. It can define a domain, and it can carry the language of that domain, its operating rules, its connectors, its validation logic, and even a specialized model where that makes sense. Vibe coding could help build pieces of that capability faster, while the lens still operates inside a larger PSL/i architecture that governs what it can know, what it can do, and what becomes authoritative.
That is a more realistic role for vibe coding than asking it to become the whole architecture. Vibe coding can create capability quickly. PSL/i can determine whether that capability is trusted enough to use, limited enough to be safe, and connected enough to matter. The generated lens does not have to solve every hard problem by itself, because it operates inside a structure built for the problems that appear after the demo works.
This is why the “pretty demo” criticism is fair but incomplete. The demo matters, because it proves that software creation can be compressed, and that is a real shift. What the demo does not prove is that dependence has been solved. The market may be valuing vibe coding on the belief that the generated surface eventually becomes the system. PSL/i raises a different question. What if the generated surface should not become the system by itself? What if it needs a foundation underneath it before it can safely become part of real work?
The question gets sharper as AI moves from assistance to action. A generated app is one thing. A generated workflow that acts across email, calendar, CRM, finance, documents, customer records, or internal systems is something else. At that point the value no longer sits in generating the workflow. It sits in knowing whether the action is allowed, whether the user approved it, whether the organization has a rule governing it, whether the result should be remembered, and whether someone can understand later what happened.
Figma’s move into AI-assisted product creation shows how seriously established platforms are taking the shift. The acquisition of the Bud team matters because Figma is no longer only defending the design canvas. It is moving closer to the point where an idea becomes a working product. Its own homepage now describes Figma as a place where “design, code, and AI come together,” from first idea to shipped product.[8] Replit, Lovable, Base44, Bolt.new, v0, Claude Code, and Figma are all pressing on the same boundary. The line between describing software and building software is moving.
The open question is whether the creation layer is enough by itself. For some categories it may be. Personal tools, simple internal apps, prototypes, and temporary utilities may not need much more. But as the generated work becomes more important, the foundation underneath it becomes more important too. Serious work needs a way to know what is being done, why it is being done, who allowed it, where the authority sits, and what the system should remember.
This is where PSL/i may be more valuable than vibe coding. Vibe coding is valuable because it may change how software starts. PSL/i is valuable because it addresses what happens after software starts, and how AI-generated work becomes something people and organizations can trust enough to use repeatedly.
Markets often pay for the visible breakthrough before the underlying structure is settled. The current valuations reflect growth, excitement, fear of missing out, and a believable claim that software creation is being re-platformed. They also raise a better question. If investors are willing to assign billions of dollars to tools that make software easier to create, what is the enabling layer worth if it makes that generated software safe enough and coherent enough to depend on?
Seeing creates belief. Dependence requires more. If vibe coding is the visible proof that AI can generate software from intent, PSL/i may be the layer that determines whether those generated systems can be trusted to operate in the real world.
Notes
[1] Replit announced that it raised $400 million at a $9 billion valuation, which it described as a threefold increase in six months.
[2] TechCrunch reported Lovable had crossed $400 million in ARR, and later reported that it reached $500 million in annualized revenue.
[3] Forbes reported that Lovable was in talks to raise funding at a $12 billion valuation, up from a reported $6.6 billion valuation in December.
[4] TechCrunch reported Wix acquired Base44 for $80 million in cash.
[5] TechCrunch reported Figma acquired the team behind Bud, formerly Orchids.
[6] “Vibe Coding as a Reconfiguration of Intent Mediation in Software Development” describes the shift from deterministic instruction to probabilistic inference and identifies risks including black-box codebases and responsibility gaps.
[7] “Figma2Code: Automating Multimodal Design to Code in the Wild” finds that models can perform well on visual fidelity while still facing limits in layout responsiveness and code maintainability.
[8] Figma describes itself as a place where design, code, and AI come together, from first idea to shipped product.
