The problem
An AI-generated launch plan is easy to produce and hard to act on. Founders need to see which assumptions have support, what remains uncertain, and what to do next.
My contribution
- Led the product requirements and choices around the idea-to-validation workflow.
- Used AI-assisted implementation, then reviewed the interface, evidence boundaries, and release checks.
Decisions & tradeoffs
Make the demo useful without a model account
- The choice
- Ship stable example workspaces and a deterministic provider; enable live generation separately.
- The tradeoff
- The default example demonstrates the workflow, not fresh market research.
- How it is checked
- Public fixtures and the demo script provide a repeatable path through the product.
Keep decisions tied to recorded evidence
- The choice
- Require decision claims to cite evidence IDs and invalidate a brief when its source evidence changes.
- The tradeoff
- Users must record useful evidence; fluent generated prose cannot replace that work.
- How it is checked
- Decision evaluation checks citations, source fingerprints, and response structure.
Make storage an optional step
- The choice
- Keep editing and private exports available locally; add cloud history and public sharing behind explicit actions.
- The tradeoff
- Browser-only data needs an export or recovery path, and shared views must expose less than private workspaces.
- How it is checked
- The documented cloud checks cover recovery, ownership, and public-share field boundaries.
See the work
Explore the work
Follow one idea into a decision
A walkthrough of the public, pre-seeded B2B SaaS sample. Evidence records are fictional demo data; no model runs here.
The brief
A weekly fix list for a small SaaS team
The public B2B SaaS activation sample turns support tickets, session notes, and churn reasons into weekly product fixes. Its audience is seed-stage founders, product managers, and customer success leads without a growth analyst.
Constraint: two people, no data warehouse; the first demo must work with uploaded CSVs and copied support notes.
What the work shows
The public project includes a working interface, stable sample scenarios, export formats, and an engineering case study. A reviewer can follow a decision from its supporting evidence to an execution task.
Follow the evidence
What this does not establish
- Demo mode and live model generation are separate; a sample output is not evidence of a live model call.
- Cloud, team, and billing capabilities depend on deployment configuration. No customer adoption or revenue claim is made.