Tour de AIJoin the peloton
Flat poster-style cyclists riding a diagonal yellow route with AI-inspired map details.

Grand Départ · Barcelona

Tour de AI

Master Claude in 21 stages. One lesson a day.

AI splits into two roads: the work it does for you, and the things you build with it. Most people find neither — they open an empty chat box and hunt for a use case. Show it your actual week instead. Over 21 days, Barcelona to Paris, we ride from your first delegated task to an AI system of your own.

Join the pelotonView the route

Built with

Show it your realityPrompt as briefingBuild your brainConnect your toolsDelegate to agentsShip your system

The Route · 23 days · One stage a day

GRAND DÉPART · BARCELONAPARISStage 1 — Sign-on: working with AI vs building with it1Stage 2 — Your use cases2Stage 3 — Basic vs better prompting3Stage 4 — Adding connectors4Stage 5 — Using & creating skills5Stage 6 — Projects6Stage 7 — Cowork: delegating the task7Stage 8 — Artefacts: making it visual8Stage 9 — Scheduled tasks9Rest day — Rest day: the grupettoRStage 10 — Context is everything10Stage 11 — The company brain11Stage 12 — Going advanced: APIs & MCPs12Stage 13 — Top tips & gotchas13Stage 14 — Meet Claude Code14Stage 15 — Build your first tool15Rest day — Rest day: system checkRStage 16 — The solo time trial16Stage 17 — Agents 10117Stage 18 — Orchestration: many hands18Stage 19 — The big climb: assemble your system19Stage 20 — Alpe d'Huez, again: harden it20Stage 21 — The Champs-Élysées: the road ahead21

Join the peloton

One stage a day. Miss a day? The grupetto always rides back on.

Week 1

AI for working
1Barcelona → Barcelona19.6 kmTeam time trialSign-on: working with AI vs building with it

The one thing to grasp on day one: there are two roads — working with AI and building with it. Get yourself a tool today, and see that it's far more than a chat box. We go deeper on all of it as the tour rolls on.

You'll learn

  • The two roads: working with AI vs building with it
  • Why it's more than chat — Cowork, connectors, skills, Code
  • Get set up on one tool today
Ride the stage →
2Tarragona → Barcelona168.5 kmHillyYour use cases

A run-through of real, everyday use cases — the tasks you already do — and why AI helps each one. You leave with a shortlist that's actually yours.

You'll learn

  • Spotting the tasks AI is genuinely good at
  • Why AI helps: speed, volume, first drafts, recall
  • Building a shortlist of your own use cases
Ride the stage →
3Granollers → Les Angles195.9 kmMountainBasic vs better prompting

The gap between a basic prompt and a great one. The move: don't guess the perfect words — get the AI to ask you the questions that build the prompt for you.

You'll learn

  • Why basic prompts get basic results
  • Using questions to build a better prompt
  • Setting context, goal and constraints
Ride the stage →
4Carcassonne → Foix181.9 kmHillyAdding connectors

Connect AI to the tools you already live in — Drive, calendar, inbox, Slack — so it works with your real information, not in a vacuum.

You'll learn

  • What connectors actually do
  • Connecting your email, calendar and docs safely
  • What changes once AI can reach your work
Ride the stage →
5Lannemezan → Pau158.3 kmFlatUsing & creating skills

Skills turn a repeatable process into something the AI runs the same way every time. Learn to use ready-made skills and to capture your own.

You'll learn

  • What a skill is — vs a prompt or a template
  • Using existing skills
  • Capturing one of your processes as a skill
Ride the stage →
6Pau → Gavarnie-Gèdre186.2 kmMountainProjects

A project is a dedicated space for one area of work — its own files, its own standing instructions, its own memory. Set it up once and every conversation inside it starts briefed.

You'll learn

  • What a project is, and when to use one
  • Instructions and context — briefing it once
  • Managed vs local project folders — and why local wins
Ride the stage →
7Hagetmau → Bordeaux175.1 kmFlatCowork: delegating the task

Cowork is where you stop chatting and start delegating: hand a whole task over, run several at once, and let the routine ones repeat on a schedule.

You'll learn

  • Delegating a task vs chatting about it
  • Running several tasks at once — and on a schedule
  • Setting a task up: projects, approvals, trusted folders
Ride the stage →
8Périgueux → Bergerac180.4 kmFlatArtefacts: making it visual

An artefact is just an HTML file — a custom front end for your data. Point your connectors at it to build dashboards and trackers that show only what you care about, and act on them in place.

You'll learn

  • Artefacts as a front end for your data
  • Building a personal dashboard from your connectors
  • Interactive views — update a record, share a link
Ride the stage →
9Malemort → Ussel185.5 kmHillyScheduled tasks

Put routine work on a timer: tasks the agent runs on a schedule — a morning brief, hourly triage — landing where you'll see them, done before you ask.

You'll learn

  • What scheduled tasks are good for
  • Setting one up — and where the result should land
  • Making it last: cheaper models, skills on a schedule
Ride the stage →
RRest dayRest day: the grupetto

Behind? Ride back on. Week 1 recap and catch-up.

Week 2

The crossover
10Aurillac → Le Lioran166.6 kmHillyContext is everything

Everything you give an AI — files, connectors, skills, instructions, the whole conversation — is context, and it shares one finite window. Brief it like a junior: the relevant slice, not the whole handbook. Less, well-chosen context beats more.

You'll learn

  • What counts as context — and the finite window
  • Briefing like a junior: the relevant slice, not the handbook
  • Context rot — why less, well-chosen, beats more
Ride the stage →
11Vichy → Nevers161.3 kmFlatThe company brain

Build a shared 'brain' on a drive you already use — Google Drive or SharePoint: a root folder split into context (durable reference) and work (live projects), with an agents.md map. Any agent, any model, any teammate works from one source of truth.

You'll learn

  • Why a shared brain beats re-loading context
  • The shape: context vs work, plus an agents.md map
  • Build it on a shared drive — portable, governed, familiar
Ride the stage →
12Magny-Cours → Châlon-sur-Saône179.1 kmFlatGoing advanced: APIs & MCPs

When a tool has no connector: add its MCP server URL as a custom connector, or allowlist its API domain (Settings → Capabilities) and let Cowork write the code — and a skill — to call it. Keep the API key in a .env file, never the chat.

You'll learn

  • MCP vs API — and adding a custom MCP by URL
  • Allowlisting an API domain so the agent can call out
  • Keeping API keys in a .env file, not the chat
Ride the stage →
13Dole → Belfort205.8 kmHillyTop tips & gotchas

The last stop before building. Treat AI like a brilliant intern who never says 'I'm not sure': spot how it slips (plausible-but-wrong, stale threads, sycophancy), verify the facts, and build the habits — draft-don't-send, cite sources, feed gotchas back — that let it earn your trust.

You'll learn

  • Common failure modes — plausible-but-wrong, stale threads, sycophancy
  • What to always verify — facts, figures, freshness
  • Trust habits — draft-don't-send, cite sources, feed gotchas back
Ride the stage →
14Mulhouse → Le Markstein155.3 kmMountainMeet Claude Code

The crossover into building. What Claude Code is, why it's different from a chat, and how to get comfortable in it — no developer badge required.

You'll learn

  • What Claude Code is, and when to use it
  • Getting set up and finding your way around
  • Your first small task in Code
Ride the stage →
15Champagnole → Plateau de Solaison183.9 kmMountainBuild your first tool

Build your first small tool — a simple app or interface over work you already do. Proof that building is now within reach.

You'll learn

  • Scoping a first build you'll actually use
  • Vibe-coding a simple tool
  • Shipping it and using it for real
Ride the stage →
RRest day coming soonRest day: system check

Week 2 recap. How's your system holding up?

Week 3

AI for building
16Evian-les-Bains → Thonon-les-Bains26.1 kmIndividual time trial coming soonThe solo time trial

One real task, end to end, alone against the clock. The exam.

17Chambéry → Voiron174.7 kmFlat coming soonAgents 101

Delegation, not conversation. What an agent actually is.

18Voiron → Orcières-Merlette185.2 kmMountain coming soonOrchestration: many hands

Divvy the work out like you would to three or four people — and know what to trust each with.

19Gap → Alpe d'Huez127.9 kmMountain coming soonThe big climb: assemble your system

Context, brain, skills, connectors, tools — together at last. Capstone, part one.

20Bourg d'Oisans → Alpe d'Huez170.9 kmMountain coming soonAlpe d'Huez, again: harden it

The Tour climbs Alpe d'Huez twice in two days. You harden the system until it holds.

21Thoiry → Paris133 kmFlat coming soonThe Champs-Élysées: the road ahead

How this compounds. Your three-month roadmap. Champagne.

Ride the whole tour

One stage a day. Miss a day? The grupetto always rides back on.

Powered by

Community · Happy Operators Hackathon

Built at the Happy Operators Hackathon

Tour de AI is one entry in the Happy Operators Hackathon — a community build-along where operators ship real projects alongside the 2026 Tour de France. More builds are rolling in through the race; here's what's on the road so far.

See the whole Happy Operators community →