Human in the Middle presents

Arthur's AI Lab

Useful AI, tested in public.

Real tutorials with real material, visible failures, and clear human checkpoints. Start with daily life, build systems, or follow the experiments that wander off the map.

01

Everyday AI

Start usefully.

02

Systems & Agents

Build inspectably.

03

Strange Experiments

Distrust the map.

Not only the parts that work

Arthur's AI Lab is not a collection of perfect software demos. Every episode begins with a real task, shows the build or experiment, and keeps visible the point where a person must inspect, decide, or stop.

Material

Something real, currently in the wrong shape.

Transformation

One bounded job for the AI system.

Checkpoint

The decision that remains human.

Three entry points. Three different questions.

The first three videos are in production. Join free to hear when they are ready.

Coming soon

Pilot 01

One Hour Back

Five everyday jobs AI is actually good at.

Explain letters, compare offers, organize notes, prepare difficult messages, and build plans without surrendering the decision.

Pilot 02

Make or n8n?

The same automation, built twice.

A messy inquiry becomes an inspectable draft brief. Same input, same schema, same human approval boundary.

Pilot 03

Can You Automate Taste?

Arthur tried.

Twenty-four images, one sealed personal ranking, and three AI curators. The conclusion stays open until the experiment is complete.

Color describes the path, not the access

Green, Yellow, and Purple are content tracks. Free and paid membership are handled separately.

Green · 01

Everyday AI

Why and how should I use AI in daily life?

  • Explain things more clearly
  • Turn material into plans
  • Protect privacy and verify results

Yellow · 02

Systems & Agents

How does a useful interaction become a maintainable workflow?

  • Make, n8n, and Dify
  • ComfyUI and creative pipelines
  • Approvals, failures, and ownership

Purple · 03

Strange Experiments

What can we learn from systems that fail, surprise, or lead nowhere?

  • Failed automations
  • Creative edge cases
  • Public research notes

Ask questions. Submit your work. Keep building.

01

Every episode gets a workbench

Chapters, transcript, prompts, corrections, and downloadable material sit beside the video.

02

Questions stay with the episode

Ghost comments keep discussion and replies where the relevant context already lives.

03

Your version is welcome

Members can link their results, describe what failed, and explicitly choose whether anything may be featured publicly.

04

Ask this tutorial

A bounded assistant answers from the transcript, materials, and corrections instead of pretending to be a general AI oracle.

Explore a lesson. Bring the questions to a Lab.

Arthur’s AI Lab and the live HITM Labs support different moments in learning. Video lessons introduce ideas and show examples you can revisit. A live course gives your team space to ask questions, try things together, and work with content designed around your goals.

1. Find a starting point

The first video lessons are in production. As they are released, use them to explore a topic and collect questions for your team.

2. Choose your learning style

Applied Lab is direct and structured; Quest Lab is playful and exploratory. Book an introductory course or bespoke content in either style.

3. Keep practising

Bring what you learn into your work and return to relevant lessons as they become available. Your team’s questions can shape the brief for a further course.

The separate AI Governance Lab connects to a planned customer-onboarding walkthrough: see a workflow become a governance blueprint, then practise the method together in Blueprint Studio.

No advance viewing is required to take a Lab. We agree suitable preparation when we plan your course.

The first three experiments stay free

Join early for the pilot episodes, production notes, and invitations to ask questions or submit your own attempts. Paid material begins only when there is enough applied value to justify it.