§ Four Claims · Four Comparisons

We are not playing the same game.

These vendors are excellent at what they do. Each comparison is not a ‘we’re better’ claim. It is a statement that we are doing something architecturally different from the named alternatives — and an explanation of why that difference matters.

Claim 01

Web Apps, Not Websites.

Lovable, Bolt, v0, Replit, and Vercel ship single-page websites. We ship multi-window, persistent, desktop-grade applications.

Generative coding tools have converged on the same output: a single React page rendered in an iframe. That works for a marketing site or a CRUD form. It does not work for the software that actually runs a business — file managers, IDEs, design tools, ERP screens, debug consoles — the applications that keep multiple states alive and let the user move fluidly between them.

The constraint is not what the AI can write. It is what the runtime can host. A single-page app cannot meaningfully manage parallel windows, persistent sessions, or cross-window communication. We host applications that can.

Frames Studio Lovable / Bolt / v0
Output shape Multi-window web app Single-page website
Persistent windows Across sessions None
Prompt-to-deployed app End-to-end Code only
Claim 02

MDI as Architecture, Not Nostalgia.

The Multiple Document Interface is not a UX preference. It is the only paradigm in which an AI agent and a developer can collaborate at the same speed.

When an agent generates a component, the developer needs to view the generated code, the rendered output, the database schema, and the debug console — simultaneously. A single-page interface forces sequential attention. The browser becomes a desktop OS: virtual z-index management, focus and activation rules, cross-window messaging, persistent layout.

The industry abandoned MDI when laptops and phones forced single-task UIs onto every developer. The era of AI-assisted development restores the case for parallel attention — and the architecture that made it possible.

Frames Studio Cursor / VS Code / Replit
Window model Persistent, parallel Linear, modal
AI agent integration surface Database Editor + shell
Cross-window communication Native None / hack-only
Claim 03

Code as Data, Not Files.

An AI agent cannot reliably manage a Git repository. It can read and write database records with perfect fidelity. The substrate determines the ceiling.

The brittleness of agentic coding tools today is not a model problem. It is a substrate problem. Files in a repository require the agent to coordinate across shells, package managers, version control, build tools, CI/CD APIs, and cloud provider SDKs. Each tool has a different interface, different failure modes, different retry semantics, and different ways of being subtly wrong.

When code is rows in a SQL database, the agent has one interface. Reads, writes, history, deployment, and rollback are all SQL. Tests run against the database. Promotion between environments is a single statement. The reliability of the system is bounded by the reliability of the database — which is bounded by decades of production hardening.

Frames Studio Devin / Cursor / OpenHands
Tool surfaces the agent must master 1 6+ (shell, fs, Git, CI, cloud APIs)
Deploy operation SQL INSERT Build + push + pipeline
Failure modes during deploy 1 (transaction) Many (build, lint, deploy, infra)
Claim 04

Removing a Cost Category, Not Cutting a Line Item.

$450K–$700K of operational overhead per 10-developer team is not a sticker on a vendor invoice. It is the sum of the things you do not realize you are paying for.

License fees are the visible part of the cost. The hidden part is everything that surrounds them — the DevOps headcount required to run the pipeline, the cloud minutes burned by builds that fail before they touch business logic, the onboarding weeks lost while a new hire learns a build system, the security audits required to vet the supply chain.

Removing the toolchain does not produce a cheaper version of the same thing. It eliminates a cost category. That is the difference between a 30% margin improvement and a structural change in how the unit economics of an engineering organization work.

Line item Unit cost Annual total (10-dev team)
IDE & dev tools (per dev) $3,000 $30,000
Version control & collaboration $25 / dev / mo $3,000
CI/CD pipeline cloud minutes $80,000–$150,000
Security & dependency scanning vendor stack $40,000–$80,000
DevOps headcount 1.5 FTE $250,000–$375,000
Onboarding (per new dev) 2–4 weeks $50,000–$60,000
Total / yr (10-dev team) $450,000–$700,000