1 / 18

Free, step-by-step slide guide

The AI R&DSystem™

Move from research to reality with a six-stage, evidence-first system for validating, building, and launching an AI product or service.

6 stages4–8 weeksAI-firstProgress saved

The finish line

A live product with a launch system

You will leave with market evidence, a validated wedge, a functional app, and a repeatable plan for turning attention into revenue.

Progress is saved in this browser.

Free, step-by-step slide guide

The AI R&DSystem™

Move from research to reality with a six-stage, evidence-first system for validating, building, and launching an AI product or service.

6 stages4–8 weeksAI-firstProgress saved

The finish line

A live product with a launch system

You will leave with market evidence, a validated wedge, a functional app, and a repeatable plan for turning attention into revenue.

Progress is saved in this browser.

What you will build

Every stage ends in evidence you can carry into the next decision.

Customer language and 10 interview records

A live validation page with qualified signups

A focused roadmap PRD

A behavior-led prototype

A functional wedge workflow

A live launch and revenue activity plan

The system deliberately delays expensive implementation until demand, audience, and the first valuable workflow are concrete.

Before step 1 · setup

Put the agent toolkit where Claude Code can use it.

Terminal

cd ~/.claude
git clone git@github.com:technical-leaders/agents.git
cd agents
ls -la

Ready when

Git and Claude Code are installed.
Your GitHub SSH access works.
The agents repository is visible inside ~/.claude/agents.
You have a real product or service idea to test.

The guide

Research the problem. Validate the bet. Build only what earns the next step.

Step 1 of 6 · Week 1

Validate demand and find a defensible market position.

Find demand before features

The fastest way to waste an AI build is to automate a problem nobody is actively trying to solve. Start with observed demand, not an imagined feature list.

Do this

1Find 10 online communities where your ideal customer already gathers.
2Observe for one week and document the five most repeated pain points in the customers' own words.

Step 1 of 6 · Week 1

Validate demand and find a defensible market position.

Map trust and interview the market

Do this

3Build an inverted Dream 100: the people, newsletters, podcasts, and communities that already hold your audience's trust.
4Conduct 10 customer interviews and save the patterns in icp-interviews.md.

Remember: Do not summarize the market into generic language. Keep the phrases real customers repeat.

What you leave with

market-research.md

icp-interviews.md with 10 interviews

Step 1 of 6 · Run it

Start the demand research

Do this

1Find 10 online communities where your ideal customer already gathers.
2Observe for one week and document the five most repeated pain points in the customers' own words.

Sea of Demand command

Use sea-of-demand on my PRODUCT_OR_SERVICE for ICP_IDENTITY with PROBLEMS_OR_RISKS

Step 2 of 6 · Week 2

Prove interest before you build the full product.

Turn the thesis into a measurable bet

A signup is stronger evidence than a compliment. A small landing page turns your market thesis into a measurable bet before you spend weeks building.

Do this

1Turn the strongest problem and desired outcome into one clear landing-page promise.
2Add an email capture and a specific early-access invitation.

Step 2 of 6 · Run it

Build and distribute the bet

Do this

3Promote the page inside the communities and relationships you mapped in Step 1.
4Get at least 10 qualified signups before committing to the full build.

Remember: If the right people do not sign up, revise the promise or audience before adding features.

Validated Bet command

Use validated-bet

What you leave with

Landing page

Email capture system

Step 3 of 6 · Week 2

Prioritize the smallest valuable product that can win.

Choose the wedge

Your wedge is the first job you can solve noticeably better than the alternatives. It keeps the roadmap narrow enough to ship and valuable enough to buy.

Do this

1Translate interview evidence into Jobs-to-be-Done statements.
2Compare the current alternatives customers use, including manual workarounds.

Step 3 of 6 · Run it

Write the roadmap PRD

Do this

3Choose the wedge feature with the strongest mix of urgency, differentiation, and monetization.
4Define the first release, explicit non-goals, and success measures in a product requirements document.

Wedge Roadmap command

Use wedge-roadmap

What you leave with

roadmap-prd.md

Step 4 of 6 · Week 3

Design a prototype that makes the desired behavior easier.

Design for behavior

BJ Fogg's model says behavior happens when motivation, ability, and a prompt converge. A useful prototype removes friction instead of merely looking polished.

Do this

1Map the shortest user journey from trigger to useful outcome.
2Use B = M × A × P to identify motivation gaps, ability barriers, and missing prompts.

Remember: B = Motivation × Ability × Prompt. Improving ability is often the fastest lever.

Step 4 of 6 · Run it

Turn the journey into testable stories

Do this

3Turn the journey into wireframes and complete user stories.
4Put the prototype in front of real customers and note where they hesitate, misunderstand, or stop.

Prototype PRD command

Use user-researcher to create a PRD

What you leave with

ux.md

Step 5 of 6 · Weeks 3–6

Turn the validated prototype into working software.

Prototype, get feedback, integrate

Short build-feedback cycles keep the implementation tied to customer evidence. Each pass should reduce risk, not just add code.

Do this

1Prototype: feed each complete story from ux.md into v0 and build the visible experience.
2Feedback: show the working prototype to people from the waitlist and record what changes their behavior.

Step 5 of 6 · Weeks 3–6

Turn the validated prototype into working software.

Build one thin slice at a time

Do this

3Integrate: give the final v0 prototype and approved stories to Claude Code.
4Repeat one thin slice at a time until the wedge workflow works end to end.

Remember: Do not treat every suggestion as a requirement. Prefer repeated evidence tied to the wedge outcome.

What you leave with

Functional app

Step 6 of 6 · Weeks 6–8

Deploy the product and turn validation into revenue.

Deploy the product, not just the code

Shipping is not the finish line. The launch system connects the product to repeatable revenue-generating activity and tells you what to improve next.

Do this

1Deploy through GitHub and Vercel, connect the custom domain, and verify the live customer journey.
2Create a launch kit from the evidence, promise, screenshots, demo, and customer language collected in the previous steps.

Step 6 of 6 · Run it

Build the revenue activity plan

Do this

3Choose the revenue-generating activities you can repeat consistently.
4Track signups, activation, conversion, retention, and the qualitative reasons behind each result.

Launch strategy command

Use launch-strategist to create revenue generating activities

What you leave with

Live product

Launch kit

Revenue activity plan

The loop continues

You shipped the wedge. Now make the system compound.

Use live product behavior and customer conversations to decide the next research bet, not a backlog built on assumptions.

Build better

The AI Builder Playbook

Turn validated stories into a bounded prototype with verification and handoff evidence.

Open training

Prompt better

Fortune 100 Prompt Library

Use proven prompts for analysis, strategy, operations, and execution.

Open training

Ship with us

Join Technical Leaders

Get the training, community, and accountability to keep shipping.

Open training