Practical insights on AI transformation, case studies, and lessons from working with teams and developers.
Pick a path based on your interest
/replan v1.3 and /recheck get three things AI-assisted development needed — project rules as a shared config, memory through lessons.md, an audit trail on disk — and /recheck adds a full /code-review high pass on real code changes. How the pieces compose and what to do this week.
The new /goal slash command in Claude Code 2.1.139 turns Claude into an autonomous agent that keeps working across many turns until a verifiable condition is met. A practical guide — how to write a good goal, where it fails, and where it's worth using this week.
Cowork is out of preview, Opus 4.7 changes how you write prompts, and Claude for Small Business packs AI into ready-made workflows. A practical filter on Anthropic's latest — without launching a big pilot.
Basic plan mode in Claude Code gives you a decent plan — but /feature-dev and /writing-plans give you a dramatically better one. Here's why the upgrade is worth it.
Free Claude Code cheat sheet
Commands, prompts, plugins and workflows from €3,000/day workshops. Download free.
Get the cheat sheet →The first plan from Claude Code is always flawed — invented functions, missed edge cases. How I use the /replan plugin to get a plan I can actually trust.
AI tools let teams ship 40% faster, but 'cognitive debt' means velocity without understanding. Here's why cutting junior hires backfires.
83% of talent leaders say upskilling matters more than hiring. Yet organizations are 3.1x more likely to hire AI talent than retrain. Meanwhile, 52% can't find the specialists they need. Here's the math.
Structured prompts or autonomous AI agents? A practical guide across the spectrum from simple prompt to multi-agent system — with real business examples.
Agentic AI is the 2026 buzzword, but reality is sobering. Where agents actually work, where they don't, and how to decide whether to experiment or wait.
AI-generated code has 1.7x more major issues than human-written code. Over 40% of junior developers deploy AI code they don't fully understand. The difference between vibe coding and disciplined AI-assisted development — and what inaction is costing you.
58% of workers spend 3+ hours/week fixing AI output. 24.7% of AI-generated code has security flaws. How to spot workslop and what to do about it.
One terminal command turns an AI assistant into an agent that plans, implements, and cleans up. A detailed walkthrough of my /improve-gitlab setup.
Practical checklist for teams before an AI workshop. What to install, what to prepare, what to expect.
The workshop's over, the team is excited. But how do you know it actually worked? 7 concrete metrics that tell the truth — with a before/after template.
A senior developer costs your company €400–500/day. A team of five? €2,000–2,500. If AI saves just 50% of their time, every day without it burns over €1,000. We did the math.
A lecture, a few aha moments, back to work. A week later? Nothing. Why typical AI trainings fail and what actually works.
Security restrictions, slow approvals, Copilot as an alibi — the organizational barriers blocking AI adoption. As a CTO, you decide whether your team grows or watches the competition pull ahead. 5 concrete steps.
DropShot is a menu bar utility for developers. Built in an afternoon from a phone with Claude Code — research, plan, Swift code, CI pipeline. Another 2 hours at the computer for polish.
When AI doesn't work out of the box, you build your own tools. Garbage collector, Telegram notifications, documentation optimizer, remote server setup, and DropShot.
Developers aren't afraid AI will be bad — they're afraid it'll be good. Three psychological barriers that explain why your team isn't using AI, even when they have access to every tool.
Invoicing in 1 day, double-entry accounting in 2 more days. AI generated a specification against legislation – 3,430 lines, 142 sources. The system was built overnight.
A project planned for one developer for 7 weeks. Result: 1.5 hours of my work + Claude Code running autonomously. A functional web application.
More than three-quarters of AI pilots never make it to production. Here's why — and how to beat the odds.
Capacity: max 4 teams per month (each workshop = 1–2 days on-site with you).