Skip to content

AI Lab

What happens when you stop treating AI like a chatbot and start treating it like part of the team.

This is where I document the tools, agents, workflows, and experiments I am exploring while building software and companies with AI. It is an active workshop, not a finished textbook.

I do not have everything figured out. That is part of the point.

Inside the lab

Tools

3

The AI tools I use

The AI tools I actually use or have meaningfully evaluated, with an honest take on each.

Agents

Specialised AI agents

The specialised AI roles I design and experiment with, and how to make an agent genuinely useful.

Workflows

4

AI workflows

How AI actually fits into real work, step by step, with the human decisions kept in view.

Experiments

2

AI experiments

Questions I run at AI, honestly documented. Some work. Some fail. The lesson is the point.

Tutorials

AI tutorials

Practical, step-by-step guides for using AI for real work, written for more than just developers.

What I'm exploring now

Experiment · Running

Can multiple AI agents produce better software than one?

Splitting engineering work across specialised agents, and watching whether it actually helps.

Developers

My AI-assisted development workflow

How I actually build software with AI in the loop, and where I still make the calls.

Tools I use

See all

Reasoning & writing · Currently using

Claude

The AI I spend the most hours with, for thinking, writing, and research.

Coding · Currently using

Claude Code

The AI coding environment I use to build software across a real codebase.

Coding · Currently using

Codex

A second AI coding tool in my workflow, because one tool does not have to do everything.

Go deeper by topic

Claude

Guides and lessons on using Claude well — for building software, thinking through problems, and doing professional work.

Claude Code

Workflows, setup, and hard-won lessons for building real software with Claude Code — beyond autocomplete.

AI Agents

What AI agents actually are, when they beat a single assistant, and how to build ones that survive real work.

Multi-Agent Systems

Designing systems of AI agents that plan, build, and review together — orchestration, specialization, and quality gates.

MCP

Understanding MCP — the open standard connecting AI models to tools and data — with practical server design lessons.

Vibe Coding

AI-speed building without shipping garbage: what vibe coding is good for, where it breaks, and how to keep quality.

AI for Work

Practical AI for professionals: getting real value from AI tools in everyday knowledge work, whatever your field.

AI for your profession

AI for LawyersAI for HRAI for FoundersAI for MarketersAI for DevelopersAI for Product Managers

Latest AI writing

See all
How I use Claude to build software3 min
What AI agents actually are2 min
How I think about multi-agent systems2 min
How professionals can use Claude (no code required)2 min
From the AI Lab

The AI experiments worth keeping.

I test AI tools, workflows, agents, and systems while building real software. The ones that work well enough to keep go into The Build Log.

Practical notes on Claude, AI agents, multi-agent systems, and building with AI. No hype for the sake of hype.

  • Claude and Claude Code experiments
  • AI agents and multi-agent systems
  • Workflows that actually held up
  • The ones that broke, and why

No spam. No daily noise. Unsubscribe whenever you want.

One email when there is something worth sharing. From Charles, no one else.

Let's create somethingawesome together.