Skill20

SKILL20Master the art of beginning

Beginners quit a new skill in the first week, from self-doubt rather than a lack of information. Skill20 turns any goal into 20 hours of practice: five to eight subskills, the obstacles you are likely to hit, and a written reflection with every logged hour.

Three AI coaches help along the way: one for the whole skill, a tutor for each subskill and the Sherpa for obstacles. They see your real hours and share what they learn about you between chats. I built it end to end: product, design, web development and the AI.

My part

End to end: product, design, web development, the AI and the infrastructure.

Visit the site: skill20.app↗

In numbers

30 / 30

test runs on production that finished with a plan

14-23 s

to generate a plan, down from 41

69

obstacles in the approved library

28

ready programmes that open without AI

What I built

01

Development Plans People Finish

Any goal becomes a plan with a budget of hours, known obstacles and reflection. The coaches remember the person.

The problem

Beginners don't know what to practise first, and they quit in week one from self-doubt, not from a lack of information.

How it works

• A goal becomes 5 to 8 subskills with a budget of 20 hours, basics first. A safety filter turns down harmful goals and offers an ethical alternative.
• AI picks up to nine obstacles, three of each kind, from an approved library of 69 and rewrites them for the person. Strategies are copied word for word; the model doesn't invent them.
• Each subskill gets its steps when the person reaches it: a lesson, a mission and a way to tell it's done.
• An hour counts only with a written reflection.
• Three coaches, for the skill, for each subskill and for obstacles, see the real hours and remember the person between chats.
• 28 ready programmes open without AI, and a close one is offered when a generation fails.

Status

• Live: running in production30 of 30 test runs on production finished with a plan, in 14-23 seconds instead of 41.

Also fits

Individual development plans in academies, with the psychologist approving the obstacle library. Onboarding. Mentoring.

02

Design

Brand, design system, data on screen, reports and sites. Variants side by side on real content; the winner becomes the standard.

The problem

An interface made of template cards and default fonts looks like every dashboard. When every screen is solved differently, people relearn the product on every page.

How it works

• The subskill page picked from five drawn variants, web and phone side by side.
• 126 concepts before the choice: 88 layouts of the learning path, 7 hover states, 4 quest maps and 27 obstacle maps.
• The landing hero: a Prometheus figure holding fire drawn live in Three.js.
• Designed and built end to end by me, with AI tools.

Status

• Live: running in productionLive on skill20.app.

Also fits

Brand systems for clubs, agencies and media. Data-heavy dashboards. Reports that look the same on screen and in PDF. Cinematic launch sites.

Also inside

• Built: finished, not in daily use yet

  When the AI fails

  Drills for five ways the AI can fail. Before them, a skill could hang in processing forever. Now a failed generation says so and offers a close ready programme.

  

• Built: finished, not in daily use yet

  One codebase, two markets

  The market decides the language, payments and analytics, and the AI provider is set per deployment. The Russian version is built and not launched yet.