The process, in the open
Lab posts: real decisions, mistakes and lessons learned.
How you know when change has truly worked
The difference between a project that has ended and a change that has truly taken root. Three signs that tell you if change is real: autonomous use, ability to repair, and self-extension.
Read →When it's too soon to automate
Automating a chaotic process doesn't fix it — it leaves it chaotic at machine speed. The fifteen-minute rule as a criterion for knowing when you're truly ready to scale.
Read →The mistake that makes you better
How diagnostic errors, handled honestly, build real trust with the client. Because trust is built in the moments when something fails.
Read →When clients don't know what they want until they see it
How we accompany clients who can't quite articulate what they need, and why the exploration step is half the work.
Read →Why we don't work for everyone
It's not exclusivity. It's coherence. Why DyMagoo is selective about the projects it takes on — and why that's a guarantee for clients who do fit.
Read →The difference between delegating and trusting
It's not a technology change. It's a visibility change. How monitoring shifts from control to operational trust.
Read →How we measure what has changed
Because standard KPIs don't capture everything that transforms. The two measures we use to evaluate real change in processes and people.
Read →Why the first month is usually the hardest
And why we say so from day one. The J curve and how to prepare for change when implementing a new system.
Read →When automation isn't enough
The AI does the work. But no one uses it. The invisible problem of adoption in automation projects.
Read →Publishing without losing control. The editorial circuit we built.
At DyMagoo, web content is generated by an AI agent. But not without oversight: there are roles, states and human approval at every step. Here's the real circuit we use.
PròximamentAutomating without changing your tools. Where we start.
The first question SMEs ask us is: 'Will we have to change all our tools?' The short answer is: no. The long answer is here.
PròximamentDyMagoo's coordination runs on AI agents. What we've learned.
We built our own operational coordination system on autonomous AI agents. Not to experiment — out of necessity. Here's what we've learned in the first months.
PròximamentDyMagoo: born in a lab
The story of a worker cooperative that was born building its own AI infrastructure. Because the best way to understand technology is to live it.
PròximamentAI agents in production: what we've learned
Twelve milestones and three mistakes from operating autonomous agents in a real environment. The lessons no documentation explains.
PròximamentOpenClaw: an AI agent built from the lab
How we designed the architecture of an agent with Spring AI, Claude API and Ollama on OKD.
PròximamentArchitecture and hardware for a local AI lab
Why the Mac mini M4 Pro and the trade-offs we made between cost, performance and scalability.
PròximamentBuilding an AI lab with OKD and Ollama
Step by step: how we set up the local inference infrastructure integrated with the DevOps stack.
Pròximament