A cohort crash course in designing with AI · 2026

Goal-Driven Design

You can build anything now. The question is whether it feels like yours.

Starts July 275 weeksMon & Wed · 18:00 CET3–5h practice per week

Functionality has become nearly free — and everything built with AI looks the same. The advantage is no longer being able to code something; it's understanding what your customer needs and having the taste that makes a product feel considered.

Goal-Driven Design is a 5-week lab: a five-step method (goal → job → build → taste → ship), two live sessions a week, and a shipped product at the end. Starts July 27, a cohort of 8–15.

01 · The problem

Features are nearly free, and everything looks the same

The same gradient hero. The same rounded cards. The same purple SaaS palette, the same Inter everywhere, the same glowing dark mode. Five apps, one template — only the accent color and corner radius changed.

Apps and sites are easy to vibe-build now. The advantage is no longer being able to code something — it's understanding what your customer needs, learning to serve it, and speaking their language. Only a human has the taste that makes a product feel considered.

02 · The method

Outcome first. The look follows the goal

We start from what we want to change in the world with our product — not from how the interface should look. The method is five steps in strict order: the tools are interchangeable, the order is not.

1 · Goal
Name the single meaningful outcome and the constraint that protects it.
2 · Job
Understand the job the product is really hired for.
3 · Build
A working prototype built through the whole pipeline.
4 · Taste
Style and anti-values written as rules an agent follows.
5 · Ship
A live, working product — for customers, or for a problem of your own.
03 · Curriculum

Idea to launch — several times over

Week 1 · Constraint & taste
JTBD and The Goal: a one-page brief — goal, switching forces, anti-values.
Week 2 · Brand to assets
Palette, type, brand voice; on-brand image generation instead of stock.
Week 3 · Code-native design
DESIGN.md drives a real UI: your brand becomes a file and AI tools obey it.
Week 4 · Quality gates
Taste written as text and variables, enforced by an agent.
Week 5 · Ship
The design system as code; the certificate is a live demo link.

Before week one (or async): we set up the agents. Never worked with agents? The Foundation pre-training gets you to the start line.

04 · Format

A lab, not a course

Two live sessions a week

Theory Monday and practice Wednesday, 18:00 CET, two hours each, for 5 weeks starting July 27.

40% talk, 60% practice

Frameworks and live demos, then work on your own project. Never more than 30 minutes of lecture in a row.

A small cohort of practitioners

The group shares questions, blockers, and progress on real work.

Support between sessions

A private WhatsApp group, weekly agent-driven homework, recordings of every session.

05 · Outcome

You ship a product, you don't sit through classes

A reusable DESIGN.md
Your brand as version-controlled text.
A working prototype
Built through the whole pipeline.
A deck that presents the project
On-brand, from the same tokens.
The certificate is a demo link
Not a diploma.

Right now you ship fast and it looks like everything else. After the lab you start from the goal — and the work looks considered, yours, unmistakably so. The difference isn't the feature count: it's judgment, and a system that makes it repeatable.

06 · Instructor

Gleb Kalinin

Portrait of Gleb Kalinin holding a camera

Ex-CPO of Ozon Travel and Ostrovok, 25 years in IT building products and leading teams across startups and corporations. Solopreneur, AI product developer, and educator — more than 450 people have been through my labs.

These days I build AI-first apps and services (Cenno, Cull, a genetic toolkit, and more) and teach the way I work: start from goals and human needs, keep taste the whole way through.

07 · Who it's for

Who this is for

Designers going agentic
You've tried Claude Code or Codex and want your craft to survive the transition.
Product managers, founders, and small-business owners
Turning an idea in your head into a shipped product.
Career growers
Staying on the technology frontier — and dreaming of launching something exceptional.

Not for you if a certificate matters more than a shipped product, or you'd rather outsource taste than own it.

08 · Proof

What participants already shipped

This isn't theory. Everything below was built by Claude Code Lab participants — the cohort this course grows out of: product people, founders, analysts, and people who had never written code. In weeks, not months.

A 20-year-old print shop · no coding background
"I still can't believe it — my son says nobody would believe his mother is programming at 3am and sending him terminal commands." Payroll went from a full day to minutes; prepress checks run without reprints; a new business unit runs on its own lightweight ERP instead of corporate software.
Production studio · a prototype for clients and investors
"On the first try it produced a working prototype we show to clients and investors — and they go: wow, this looks great." Two passes, about $50 in tokens — instead of months of development and tens of thousands of dollars.
Agency · Google Ads analyst
A web interface on the Agent SDK: negative-keyword and competitor analysis, ready-made recommendations and slides. Weekly reports run on parallel agents — one gathers, one checks, one writes the summary.
HR · candidate-scoring dashboard
"Claude Code now feels permanently in my context window." A forty-question survey became an analytics site any hiring manager can use to compare candidates. Built in ninety minutes.

Everyone shipped — from a worked-out idea to a working prototype you can open by link.

09 · What your company gets

An investment that stays with the team

A work project as the coursework
Your employee takes a real company task through the lab. The output is a working prototype and a DESIGN.md of your brand: a design system as code that AI tools obey.
An asset, not a certificate
The goal → job → build → taste → ship pipeline, the templates, and the session recordings stay with the team and transfer to the next task — prototypes without waiting on the dev queue.
Time under control
Sessions in the evening: Monday and Wednesday, 18:00 CET — outside core hours. Independent practice is 3–5 hours a week, for 5 weeks total.
Risk covered
A full no-questions refund within the first 7 days. Companies can pay by invoice.
10 · Formats

Pick your format

FormatPriceWhat's included
Standard€50010 live sessions + recordings, WhatsApp group, weekly homework, DESIGN.md templates, Claude Fable live on shared projects
Personal€1,200 · 3 seatsEverything in Standard + 4 private 1:1 sessions, a full review of your project, priority feedback
Team€2,700Everything in Standard, up to 7 people, 4 dedicated team sessions

The cohort starts July 27 — one small group, 8–15 builders. Invoice payment for companies. Full refund within the first 7 days.

ai-design.salient.community

11 · A letter to your manager

A ready-to-send approval request

Copy it, fill in the name, hit send — everything a decision needs is already inside.

Subject: Training approval request — Goal-Driven Design (5 weeks, €500)

Hi [Name],

I'd like your approval to join Goal-Driven Design — a 5-week lab on designing products with AI agents. It's run by Gleb Kalinin, ex-CPO of Ozon Travel and Ostrovok; more than 450 people have been through his labs.

What the team gets:

  • My coursework will be one of our real projects: the output is a working prototype and a design system of our brand as code (DESIGN.md) — both stay with us.
  • I'll learn the goal-to-ship pipeline with agents (Claude Code, Codex, Figma MCP) and can carry it over to the team's work.
  • Session recordings, templates, and materials remain available after the course ends.

The cost:

  • €500 for the Standard tier (invoice payment available).
  • Time: two 2-hour sessions a week (Mon & Wed, 18:00 CET — outside core hours) plus 3–5 hours of practice.
  • Risk: a full refund within the first 7 days if it's not a fit.

The cohort starts July 27. Program and details: ai-design.salient.community (this presentation lives there too).

Cohort crash course · 2026

Goal-Driven
Design

You can build anything now. The question is whether it feels like yours. A crash course in designing with AI: judgment first, tools second.

Starts July 27 · 5 weeks · Mon & Wed, 18:00 CET · 3–5h practice per week

The problem

Features are nearly free —
and everything looks the same

The same gradient hero
The same rounded cards
The same purple SaaS palette
The same Inter everywhere
The same glowing dark mode

Five apps. One template. Only the accent color and corner radius changed.

The shift

The advantage is
no longer the code

The advantage is understanding what your customer needs, learning to serve it, and speaking their language.

Only a human has the taste that makes a product feel considered.

The method

Outcome first.
The look follows the goal

01

Goal

Name the one outcome and the constraint that protects it.

02

Job

Understand the job your product is really hired for.

03

Build

A working prototype through the whole pipeline.

04

Taste

Style and anti-values written down as rules.

05

Ship

A live, working product — for customers, or for a problem of your own.

The tools are interchangeable. The order is not.

Curriculum · 5 weeks

Idea to launch —
several times over

1

Constraint & taste

Goal, switching forces, anti-values — a one-page brief.

2

Brand to assets

Palette, type, brand voice; on-brand image generation.

3

Code-native design

DESIGN.md drives a real UI; your brand becomes a file.

4

Quality gates

Taste written as text and variables, enforced by an agent.

5

Ship

The design system as code; the certificate is a live demo link.

Format

A lab, not a course

Two live sessions a week

Theory Monday, practice Wednesday, 18:00 CET, two hours each.

40% talk, 60% practice

Never more than 30 minutes of lecture in a row.

A small cohort of practitioners

Questions, blockers, and progress on real work.

Support between sessions

Private WhatsApp group, weekly agent-driven homework, recordings of every session.

Outcome

You ship a product,
you don't sit through classes

A reusable DESIGN.md
Your brand as version-controlled text.
A working prototype
Built through the whole pipeline.
An on-brand deck
Assembled from the same tokens.
The certificate is a demo link
Not a diploma.
The shift

From template-shaped —
to unmistakably yours

Now

You ship fast, and it looks like everything else. One more AI builder, indistinguishable from the rest.

After the lab

You start from the goal. The work looks considered, yours. You ship things people actually use.

The difference isn't the feature count. It's judgment — and a system that makes it repeatable.

Instructor
Portrait of Gleb Kalinin holding a camera

Gleb Kalinin

Ex-CPO of Ozon Travel and Ostrovok, 25 years in IT building products and leading teams across startups and corporations. Solopreneur, AI product developer, and educator — 450+ people have been through my labs.

These days I build AI-first apps and services (Cenno, Cull, a genetic toolkit) and teach the way I work: start from goals and human needs, keep taste the whole way through.

Who it's for

Who this is for

Designers going agentic
You've tried Claude Code or Codex and want your craft to survive the transition.
Product managers, founders, small-business owners
Turning an idea in your head into a shipped product.
Career growers
Staying on the technology frontier — and dreaming of launching something exceptional.

Not for you if a certificate matters more than a shipped product.

Already shipped · Claude Code Lab

This isn't theory —
people ship products

Print shop, 20 years · no code

Payroll went from a full day to minutes; prepress checks without reprints for a month; a new business unit runs on its own lightweight ERP instead of corporate software.

Production studio · investors

"On the first try it produced a prototype we show to clients and investors." Two passes, ≈$50 instead of months of development.

Agency · Google Ads analyst

An agent gathers negative keywords and competitors and writes recommendations; weekly reports run on parallel agents.

HR · candidate scoring

A forty-question survey became an analytics dashboard any hiring manager can read. Built in ninety minutes.

Projects by Claude Code Lab participants — the cohort this course grows out of. Everyone shipped.

What your company gets

An investment that
stays with the team

A work project as the coursework

Your employee takes a real company task through the lab: the output is a working prototype and a DESIGN.md of your brand.

An asset, not a certificate

The pipeline, templates, and recordings stay with the team and transfer to the next task.

Time under control

Sessions in the evening (Mon & Wed, 18:00 CET) — outside core hours. Practice is 3–5h a week.

Risk covered

A full no-questions refund within the first 7 days. Invoice payment available for companies.

Formats

Pick your format

StandardPersonalTeam
€500 €1,200
only 3 seats
€2,700
up to 7 people
10 live sessions + recordings, WhatsApp group, homework, DESIGN.md templates Everything in Standard + 4 private 1:1 sessions, a review of your project Everything in Standard + 4 dedicated team sessions

Invoice payment for companies. Full refund within the first 7 days.

Enroll

Starts July 27

One small cohort, 8–15 builders. Give your apps, decks, and every piece of communication a recognizable style.

ai-design.salient.community