What is AI in the Lab?
AI in the Lab is Mark Bennett's solo build series where real applications are built on camera with AI as an active partner.
The series
AI in the Lab.
Real working systems, and how I actually use them.
Quick answers
Quick answers about AI in the Lab.
Why does the series matter?
It shows the full build, including false starts, review loops and working code, so viewers can see the actual practice behind AdviceForge's client work.
Is the code available?
Yes. Code and reference documents from the builds are shared in public repositories when the episode ships.
Most AI conversations run in two registers: hype or skepticism. Neither helps you decide what to do this quarter, inside your company, with the team you actually have. The throughline here is the gap between what AI can do in a demo and what it produces inside a real business. That gap is where the decisions live.
Sometimes an episode is a build, worked start to finish. Sometimes it is a walkthrough of something I actually run. Either way you see the whole thing, not the highlight reel, including the parts that stall.
What it takes to reproduce the result gets shared publicly. Code when there is code, and the prompts, tools and setup when there isn't.
Episode 2 · Latest
Build your own AI chief of staff.
23 minutes · Watch on YouTube
An AI chief of staff that plans with me, drives my week, and holds me accountable. It has run my week every day since June 2025. Most people expect an assistant that minds a calendar and an inbox. I think out loud with mine, and it thinks back.
This one is a walkthrough rather than a build. I show it working on a real day: the morning briefing it wrote before I was awake, a meeting brief with a judgment in it that nobody typed, where the memory lives in plain text, and the work it does unasked. Then I open an empty folder and set a new one up from scratch, so you can see what day one actually looks like.
Why watch
The whole thing, not the highlight reel.
01
Nothing is edited out
The dead ends and the corrections stay in. That is where the useful judgment lives, and it is the part every polished demo removes.
02
You get what it takes to rebuild it
Code when there is code, and the prompts, tools and setup when there isn't. If something works on camera, you can run it yourself.
03
The way I actually work
The same methods I use in client work and in my own practice, applied to real work rather than a toy example.
Episode 1
Concept to a working chart UI.
90 minutes · Watch on YouTube
An automated crypto trading app, built with AI as an active partner from the first conversation. Episode 1 covers the concept conversation, the high-level plan, and Phase 1 of the build. It walks the concept through to a defensible plan, then the plan through to a runnable artifact: a working UI with a real-time TradingView-style chart that you can clone from the repo.
Disclaimer: Educational only. Not investment advice.
Host
Mark Bennett.
Founder of AdviceForge and former CEO. I ran a deep tech company for nearly ten years and pushed AI through it from the chair, which is where the judgment comes from. AI in the Lab is me doing the work in the open, so you can judge the work rather than the claims.
I write at adviceforge.com about why team AI capability is the actual constraint on AI ROI. More about me.
Want this inside your company?
See how I work with small-to-mid-market companies as a fractional Chief AI Officer, or book a call.