This room is the full readout of the commercial/private builder sweep: product families, capability taxonomy, OSS follow-on candidates, local evidence reused, and the bounded pilot stack that emerged without being mistaken for a build authorization.
Report language: findings, not conclusions.
Generic prompt → app is a category. Trusted composition for a client job is the wedge to investigate.
The reports distinguish generation from the hard parts that survive contact with a client: selecting a suitable scaffold, controlling source/provenance, operating software with no API, proving the result, and recovering when an action goes wrong.
Read by capability, not by hype ranking.
| Family / subjects | Observed capability surface | What it makes cheaper | Gap Actionist can test |
|---|---|---|---|
| Prompt-to-app Lovable · v0 · Bolt | Frontend generation, project iteration, design-system or registry context, export/build loops. | First-pass UI and scaffold production. | Outcome contract, vertical context, evidence and operation beyond generated code. |
| Agentic build environments Replit · Base44 · Manus | Agent loop, hosted execution, app scaffolding, task completion, preview/deploy. | From brief to runnable project with less setup. | Authority boundaries, recovery, client-owned data/deployment and no-API operation. |
| Data-native / internal tools Airtable Omni · Retool · Bubble | Structured records, forms, dashboards, workflow construction, business-facing UX. | Operations tooling and CRUD surfaces. | Reusable vertical atoms plus approval/read-back/audit semantics. |
| Visual / design-to-code Onlook and screenshot/design systems | Visual editing, design-to-code, tokens, component registry direction. | Shorter loop from taste to frontend starting point. | Image → token IR → registry-constrained code with screenshot proof. |
| Agent / workflow builders Zapier · CopilotKit · Cursor · Claude Code | Workflow primitives, embedded agents, code-native edits, tools, integrations. | Automation composition and developer leverage. | Typed action authority over legacy/UI-only systems and evidence receipts. |
| Sandbox / execution substrate E2B · browser / VM providers | Isolated runtime, previews, browser/computer-use execution, deploy surfaces. | Running untrusted or generated code. | Multi-tenant acceptance rig, rollback, cost normalization, and policy envelope. |
The long report contains the source index for every subject.
Use as a category boundary. The research argues a scoped, known-schema builder deletes much of the generality problem; it does not erase the need for deployment, tenancy, and evaluation.
Useful documented context-injection and design-system surfaces. OEM/white-label and production-ready buy boundaries remain held.
Evidence for how quickly prompt-to-project expectations are moving; not evidence that a client-owned governed loop is solved.
Shows how build, run, and agent operations converge in one product surface. Raises the bar for execution and cost control.
Important as a signal that broad builders can reach hosted app output quickly and become acquisition targets.
Agent + cloud VM + builder is not a future threat. Benchmark the no-API and evidence wedge instead of copying the whole surface.
Conversational forms, data, and automation show that the winning unit may be a job-specific control surface, not a code editor.
Reinforces the design-direction and token-contract idea, but does not close production adaptation or visual verification.
Cheap, constrained edits can be powerful inside a known scaffold; the builder needs a contract that keeps them bounded.
No candidate was admitted in this report.
| Slice | Candidate / evidence | Disposition | Why it matters |
|---|---|---|---|
| Image → code | abi/screenshot-to-code, MIT, active, high star signal. | liftable base | Provides a starting point for a visual compiler; still needs token/registry and build proof. |
| Scaffolds | boxyhq/saas-starter-kit, ultimate-backend, Refine references. | study / validate | Confirms the “10 repos” instinct; foundation contracts matter more than pretty screens. |
| Lovable clones | beam-cloud/lovable-clone, codecapsule/E2B templates. | study only | Useful architecture signals; many clones are unlicensed or incomplete. |
| License gate | ScanCode toolkit as industry reference. | required | License scanning must be executable in CI, not a prose claim. |
| Voice | Pipecat / LiveKit-based frameworks. | solved substrate | Voice transport is available; guided elicitation and structured approvals remain product work. |
| Image → tokens | No public OSS implementation found in the bounded sweep. | open field | Potentially novel, but requires an eval harness rather than a novelty claim. |
Validation target, not a conclusion.
One vertical → one guided spec → one design round → one archetype → one validated preview.
One dashboard-shaped vertical with a known data/deployment boundary.
One scaffold archetype plus roughly 50 curated components inside a registry.
Prompt → approved design → working app in one session with a complete artifact trail.
Corpus-wide retrieval, voice, marketplace listing, and factory scale-out remain phase 2+ until the bounded loop is validated.
These are the reports the overview previously compressed away.
Commercial matrix, taxonomy, OSS map, source index, local evidence, pilot stack, next milestone.
Open full report ↗Earlier commercial matrix, scope/confidence, product implications, and Actionist-owned gaps.
Open sweep ↗The companion AM platform room contains the immutable first-party dossier packets behind the shared comparison.
Open platform room ↗Local design documents for retrieval, image-gen front-end, guide agent, and block contracts.
Open design spec ↗Choose a report above. The local reader loads the copied source report in full.