Provider boundaries
Mock mode is deterministic. Live MiniMax and OpenAI-compatible providers are server-side opt-ins with host allowlists, timeouts, body caps, and schema checks.
Portfolio case study
LaunchLens AI turns founder ideas into evidence-grounded GTM decisions.
This is the reviewer-facing map for the project: what problem it solves, how the AI workflow is constrained, how cloud persistence works, and which gates prove the production demo.
Product loop
The model helps, but evidence controls the decision.
- 01
Founder brief
A raw product idea starts the flow, with deterministic examples for B2B SaaS activation, clinic operations, and creator commerce.
- 02
Structured GTM workspace
The provider returns target users, pains, MVP scope, backlog, launch content, pricing, risks, assumptions, and execution tasks.
- 03
Validation evidence
Each assumption becomes a stable experiment so evidence, confidence, decisions, and next actions do not drift when the plan changes.
- 04
Cited decision brief
The decision copilot can only cite recorded evidence IDs and rejects stale or invented citations before the UI accepts a brief.
- 05
Cloud recovery and sharing
Private snapshots, recovery keys, tenant isolation, RBAC, and public-share projections prove the workflow can leave local storage safely.
Engineering signals
What a reviewer should inspect first
Evidence discipline
AI-generated plans are editable, but validation evidence is human state. Decision briefs must cite exact evidence IDs.
Cloud persistence
Neon-backed snapshots, recovery migration, quotas, and idempotent migrations give the project a real data layer without breaking local-only mode.
Privacy projection
Public shares expose summary status and decision state while excluding founder input, evidence notes, sources, owner credentials, and private briefs.
Release operations
CI, hosted visual regression, post-promotion verification, cloud smoke, and release evidence turn the project into an inspectable artifact.
Reviewer path
The live demo, this case study, the written case study, and the demo script are now connected through an explicit portfolio verifier.
Evidence map
The case study is backed by commands, not just prose.
`verify:portfolio` keeps this public page, the written case study, README, demo script, runbook, and production packet wired together.
GateCommandProof
Local quality gate
npm run qualitylint, tests, typecheck, evals, build, auditPortfolio package
npm run verify:portfoliocase study, demo script, README, runbook, and production packet stay linkedProduction demo
npm run verify:production-demopublic status plus browser e2e against the live URLCloud release
npm run release:cloudmigration, schema verifier, workspace smoke, tenant smoke, and RBAC smokeRelease evidence
npm run evidence:releaseignored Markdown and JSON proof packet for the current production SHAProblem
Founders can generate ideas quickly, but the hard part is tying positioning, launch scope, evidence, and decisions into one loop that survives handoff.
Product answer
LaunchLens creates a structured GTM workspace, lets humans add evidence, and then asks AI to synthesize only evidence-cited recommendations.
Next stage
The next work is not more random features. It is reviewer evidence indexing, commercial-readiness planning, and visible eval drift signals.
Next milestones
What comes after this portfolio release
- Shorten the reviewer path with a one-click evidence index and current workflow-run links.
- Use the Commercial readiness plan for onboarding, plan limits, and billing re-entry design.
- Extend eval history with latency and quality drift signals that are visible from the public case study.
- Turn the case-study page into a durable release-note surface once product analytics exist.