Clarity — Pre-Build Intelligence

LLM wrapper that turns a vague founding idea into a build-ready prompt: a 7-step pipeline from context through a scoring matrix to a build spec, cloud and local mode, pay-per-session with no subscription. Concept, prompt system and full-stack build from one hand — in production since February 2026.

Starting Point / Problem

Building has never been easier. Replit, Lovable, Bolt and Cursor turn an idea into a running app in minutes. The bottleneck has moved: failure today rarely happens at the code level — it happens at the decision of what to build in the first place.

Feed any AI a vague idea and it produces a vague result. Garbage in, garbage out — the problem isn't the AI, it's the missing clarity before you use it. No AI tool asks who is building first: background, available time, tech level, network. An idea that's brilliant for a ten-person team can be poison for a solo founder with three hours a week.

My Role / Contribution

Sole concept lead and developer — from methodology to production:

Pipeline design: a 7-step flow (Context → Ideas → Selection → Landscape → Assumptions → Product → Build Spec), guided by versioned prompts rather than an open chat interface
Scoring methodology: a Technical Scoring Matrix across five weighted dimensions (market potential, differentiation, critical assumptions/risk, AI-native feasibility, founder fit) — configurable weight profiles for indie founders, corporate innovation managers and technical builders
Full-stack build: frontend in React 19 / TypeScript / Vite / Zustand, Local Mode against Ollama/LM Studio, Cloud Mode via a Vercel Edge Function calling the Claude API (Haiku for speed and cost), embedded Stripe checkout with no account required
Positioning & pricing: from an internal experiment ("Ode Afortunada") to a monetized product — a pay-per-session model (€4 single session, €15 for a pack of 5), subscriptions deliberately rejected, a go-to-market brief and a promo-code strategy for the validation phase

Solution / Result

Clarity is not an idea generator and not an app builder — it's the layer in between.

Context before creativity:

Before any idea, Clarity builds the founder's profile: background, interests, available time, tech level. That profile isn't decoration — it's the filter the rest of the pipeline runs through
The pipeline generates ten contextualized ideas, prioritizes them strategically, analyzes the competitive landscape of the shortlist, and surfaces critical assumptions before a single line of code is written

Structured evaluation instead of gut feeling:

Every idea is scored against the same five dimensions (1–5 per criterion), weighted by profile — the result is a reproducible ranking, not an opinion
Reducing self-deception doesn't happen through confrontation, but through structural exposure

Clarity — output example: the "StandUp" build spec with problem, target user and core assumption, generated from context and idea

From conversation to build spec:

The output isn't an analysis report — it's a structured prompt, ready to paste into Replit, Lovable, Bolt or Cursor. It describes the goal, the problem, the three core MVP features and the basic interface logic
A session takes about 30 minutes, with no install and no forced account

Two paths, one pipeline core:

Local Mode: free, offline, your own model (Ollama/LM Studio) — for anyone who doesn't want this to be anyone else's business
Cloud Mode: no setup, in the browser, mobile included — Claude Haiku on the backend, pay-per-session via Stripe, no subscription

Clarity — positioning against other AI tools: asks "who are you" first, optimizes for decision quality instead of execution speed

Impact / Value

In production since February 2026, monetized: Cloud Mode live with promo-code access for selected cohorts (e.g. CLARIDAD26) during the validation phase
One head, four disciplines: scoring methodology, prompt system, full-stack build and pricing strategy without handover losses — from concept paper to a running Stripe session
Disciplined positioning: not an idea generator, not an AI builder, not a chat assistant — the layer that connects the founder's context to their next build decision
Business model as a design decision: pay-per-session instead of a subscription lowers the decision threshold to a single, immediately completed purchase — the subscription option was deliberately rejected
Built toward a standard: a roadmap to Portfolio Mode — evaluating up to five ideas in parallel, a Decision Control Panel, a gradient Clarity Score™ instead of rigid thresholds
Speculative Lab as origin: started as an internal tool for idea validation, now a standalone, paying product — an example of a Lab experiment becoming a product line

Role

Product Concept & Scoring Methodology · Prompt & System Design · Full-Stack Developer · Brand & Positioning

Skills

Guided LLM Pipeline (7 Steps) · Versioned Prompts · Technical Scoring Matrix (5 Dimensions) · Claude API (Haiku) · Local Models (Ollama, LM Studio) · React 19 · TypeScript · Vite · Zustand · Vercel Edge Functions · Stripe Checkout · Product Positioning · Pricing Model (Pay-per-Session) · UX Writing · Go-to-Market Briefing

Clarity — Pre-Build Intelligence

Starting Point / Problem

My Role / Contribution

Solution / Result

Impact / Value

Visit projectClarity — Pre-Build Intelligence.

Visit project
Clarity — Pre-Build Intelligence.