Open Practice Start free — no signup

Free · no signup · open materials · your work to keep

You have watched enough tutorials.
Today you build something real.

Months of videos, and still not sure you could do it on your own? Pick one of 20 practical AI build paths, work through real examples with real tools, and finish with something you made yourself and can explain. Free, and no account needed to start.

  • 20 build paths
  • 250 activities
  • 39 free resources
  • ~127 hours of practice
  • Works offline
  • 20build paths, each ending in something you can show
  • 250activities, with worked examples and evidence checklists
  • 39free tools that need no account
  • 127hof estimated practice, at your own pace

01 — The build collection

Something you can test, explain, and keep.

Each path is a short sequence of real experiments with real tools, not a lecture. Completion means a self-reported practice record — not a verified qualification, accreditation, or a promise of employment.

FOUNDATIONS

AI, without the guesswork

What a language model actually does, where it stops being useful, and how to check its work instead of trusting it.

OFFICE WORK

The work you already do

Documents, data, email and meetings — the tasks that eat a week, done with a model beside you.

BUILDING

Make a small useful thing

A form, a page, a workflow. Something that runs, that you wrote, and that you can fix when it breaks.

DATA

Read a number honestly

Where the figure came from, what it leaves out, and how to say what you are unsure about.

JUDGEMENT

Know when not to use it

The part most courses skip: the cases where a model makes the work worse and a person should decide.

FOR EVERYONE

AI from zero, in plain words

Large-print safe steps for seniors, a grown-up-and-child path for kids, and a glossary with no jargon in it.

02 — New: 50 guided experiments

Make a little. Learn a lot.

Prompt experiments, source checks, creative briefs, useful data and everyday workflows. Every activity has its own cover, illustrated guide, worked example and evidence checklist.

A guided experiment is twenty minutes long. It has one thing to make, one thing to check, and one question you should be able to answer afterwards.

// a guided experiment, as it opens
Check a claim before you repeat it
time      20 minutes · no account
you make  a two-line correction, in your own words
you check  one primary source, fetched yourself
evidence  [ ] source link   [ ] what it actually says   [ ] what changed
// and the question: where would this have gone wrong if you had trusted it?

03 — 39 free tools

Learn-by-doing workshops, no signup.

A tool field desk, an honest tools directory, side-by-side comparisons, a five-question finder and a library of 60 prompts to practise on.

Field desk

Pick a task, get a practice prompt built for it.

Tools directory

What each tool is good for, and where it is not.

Comparisons

Side-by-side, honest, including the boring differences.

Tool finder

Five questions, one shortlist, no email required.

Prompt library

60 prompts to practise with, each one explained.

LLM guide

Language models explained without a single sales claim.

Glossary

The words, in plain language, in one page.

Practice packs

Handbooks and standalone HTML workspaces for offline work.

04 — How it works

Your pace. Your progress. Your copy.

  1. 01

    Start without an account

    Explore activities and keep local notes. Sign in only when you want your work across devices.

  2. 02

    Work through a path

    Real examples, real tools, one activity at a time. Download the standalone HTML workspace if you want to work offline.

  3. 03

    Keep the evidence

    Each path ends with your own project and a rubric you score honestly. Keep the file — it is the only thing a future employer can actually check.

  4. 04

    Ask when you are stuck

    An optional cloud tutor, using your own provider key if you prefer. Your key is used for that request and is not stored.

05 — Access should not define your potential

Useful practice, wherever providers leave gaps.

For people across underserved and excluded regions. Check country eligibility, language, identity requirements, cost, quotas and data terms before choosing a service. Use legitimate available alternatives — never false residency or restriction bypasses.

Downloaded lessons and fictional fixtures remain useful without a live provider. That is deliberate: a connection should not be the price of learning.

  • ✓
    The learning material is genuinely free Activities, original handbooks and practice packs are free to read and download. No payment required. External AI providers have their own quotas, costs and terms.
  • ✓
    You can learn without an account Local notes, no sign-in. An account is only needed for cross-device saves and connecting your own provider.
  • ✓
    You can work offline Download the standalone HTML workspace and the practice pack before disconnecting. Live cloud inference needs a connection.
  • ✓
    No advertising trackers First-party learning analytics record named progress events — not keystrokes, private prompts or API keys. You can export or delete your account data.

06 — Honest about what this is

Not a qualification. Practice you can show.

"The best way to understand a tool is to make something with it — and know where it stops being useful."

Progress and rubric scores are learner self-reports. Keep your actual project evidence, and do not present the completion indicator as verified certification.

Independent subject-matter review is still pending. Code is MIT, original lessons are CC BY 4.0.

Is the material really free?

Yes. Activities, original handbooks and practice packs are free to read and download. No payment is required. External AI providers may have their own quotas and costs.

Can I learn without an account?

Yes. Explore and keep local notes without signing in. An account is for cross-device saves and the learner-owned provider connection; browser drafts do not sync automatically.

What can I use offline?

Download a standalone HTML workspace and the practice pack before disconnecting. The HTML includes activities, editable notes, review questions and the capstone rubric. Keep edits by downloading an updated copy.

Is this accredited?

No. Open Practice is independent, non-degree and non-accredited. Completion is a self-reported practice record, not verified assessment.

Do I have to give you an API key?

No. All core practice works without one. If you connect a provider, your key is used only for that request through the platform and is not saved in an account, browser storage, downloads or analytics.

Make something useful. Keep learning from it.

Free, no signup, no card, and everything you make stays yours. Start with one twenty-minute experiment today.

Start free — no signup

Quick answers

What Open Practice is, in plain terms.

Open Practice is free, non-accredited AI training built around doing: 20 build paths, 250 activities and 39 tools, with offline downloads and no account required.

What is Open Practice?

Open Practice is free, non-accredited AI training built around doing: 20 practical build paths, 250 activities and 39 free tools.

Is it really free?

Yes. No account, no card, no sign-up form; the materials download for offline use and the work stays with the learner.

Do I get a certificate?

No. Open Practice is non-accredited; what you keep is the work you built and the evidence around it.

Do I need a paid AI subscription?

No. The activities are built around free tools in the directory.

Can I use it offline?

Yes, the packs download and work without a connection.

Who is it for?

People who have watched months of tutorials and still are not sure they could do it alone.

Prices: Free, with no paid tier. Every plan, with its checkout link: