30 years shipping. A serial founder — previous ventures with up to three co-founders and up to 40 people on payroll, two exits — now running a standing crew of Claude Code agents across four codebases solo. I sit on the other side of the table for founders, CTOs and technical co-founders: the architecture calls, the agent operating model, and the co-founder and fundraising decisions nobody can Google.
Skill transfer for freelancers, solopreneurs, developers and teams: Claude Code the way I actually use it, self-hosted open-source LLMs, and spec-driven pipelines that ship. Power Hour, solo day, or a team workshop.
What you'll learn ↓I listen more than I talk — and I bring thirty years of scar tissue. Expect a lot of questions first, then a straight “here's how I'd do it” and “here's what usually breaks.” Technical or not.
Architecture and setup decisions from someone who has shipped across many stacks — Go, Python, Node, PHP, Flutter and the frameworks around them, Postgres and vector databases like Qdrant, self-hosted LLMs, agent infrastructure and observability, the lot. Not "fix my code": we work on how I'd do it, what breaks at scale, and what you'll regret in six months.
How to actually run agents in production — not a demo. The patterns from my own floor: isolate each agent in its own worktree, gate every task through independent review, tier models so orchestrators think and cheap workers execute, and keep a numbered ledger with no fake-green.
The non-technical calls that decide whether a company survives — co-founder friction, splitting equity, taking money from a co-founder, when to push and when to fold. Lived experience, not theory: previous ventures with up to three co-founders and up to 40 people on payroll. And because I run technical due diligence for investors and have pitched my own rounds, I can tell you how a VC or angel will actually read your deck, your tech, and your team.
For freelancers, solopreneurs, developers and teams. Solo days and Power Hours pick the modules you need; team workshops compress them into a half- or full-day sprint.
Context window hygiene, CLAUDE.md patterns that scale, subagents for verification, and the small habits that turn a toy into a teammate.
Running GLM / Gemma / Qwen on your own hardware or a VPS. Cost math, latency targets, fallback routing, and when to stop.
The workflow I use: write the spec, let the agent draft, run tests, review, merge. How to keep humans in the loop without bottlenecking.
Using models to review PRs, catch regressions, and ship without the usual "AI wrote 80% — now I rewrite 80%" trap. Real merge discipline.
Conductor / worker, waves, supervisor loops — and when multi-agent is genuinely better than one well-prompted model. Separating hype from utility.
Token budgets for real teams, data leakage to avoid, when to route to OSS vs. commercial, and the EU legal bits nobody else covers.
One call, one concrete problem. You bring a repo, an architecture question, a gnarly setup decision — or a co-founder call you're stuck on — and we work it live. Recording included.
Six weekly sessions plus async Slack between them. Ongoing sparring on the technical and non-technical calls — architecture, agent operating model, hiring, co-founder and fundraising decisions — in the context of your real company.
1:1, hands-on. For freelancers, solopreneurs and single founders who want to master agentic coding in their own codebase. Half-day to get moving, full day to go deep.
Onsite or remote, up to 12 engineers. By end of day your team shares one agentic baseline — the same setup, the same review discipline, the same shipping flow.
I run a near-autonomous software factory — a standing crew of Claude Code agents fixing, building, researching and reviewing across four codebases while I watch the floor. When we spar, that's the experience you're borrowing.
moved in one coordinated wave session.
a Flutter app built end-to-end, zero babysitting.
running in parallel in isolated worktrees.
apps, platform, research, error-triage — at once.
“I joined a hands-on GenAI workshop with Hagen Hübel, where he guided me through the entire process, backed by his hands-on experience from multiple complex, production-grade projects. […] We'd build and deploy a full application — from requirements to production — without writing a single line of code ourselves.”
“I rarely come across true tech talents who stand out like Hagen. I had the pleasure to co-found and grow the education technology start-up Zizzle with Hagen. I admire him not only for his exceptional proficiency in solving technical challenges but also for his ability to communicate complex problems in a clear and easy-to-understand manner. He has all it takes to take on technical leadership roles at any company.”
This is how the 6-Week Sparring and the workshops run. A Power Hour skips all of it — it's one call, booked directly, no intake or follow-up.
Free 20 min. We check fit, look at your situation, and agree on outcomes.
I review your repo, stack, and team shape — or the decision on the table. You get a written plan.
Weekly sparring for the 6-week program, one hands-on day for a workshop. A conclusion is the KPI.
You own the workflow — or the decision. Playbook, template repo, and a 30-day follow-up.
Want a build, not coaching? Book a build call →
Technical: architecture, database and setup decisions across many stacks, agent infrastructure, the operating model for running agents in production. And non-technical: co-founder friction, equity, fundraising, the calls a founder can't Google. What I don't do is line-by-line code debugging — that's not where the leverage is.
Yes — it's some of the most valued time I sell. I've built ventures with up to three co-founders and up to 40 people on payroll, raised, and exited twice. I've been through the arguments, the equity splits, and taking money from a co-founder. I also run technical due diligence for investors and have pitched my own rounds, so I can show you how the other side of the table reads your deck, your tech, and your team. I listen, ask a lot, and give you straight, honest counsel.
That's half of what I actually do. The agentic-coding track covers self-hosting GLM / Gemma / Qwen on your own infra, with fallback routing and eval loops.
Claude Code is my daily driver and the default — but the patterns transfer to Cursor, Aider, and your own in-house orchestrators. The workflow is the point, not the brand.
After session 1, if you feel it's the wrong fit, I refund the remaining sessions pro rata. No drama.
English or German. I take on a handful of 6-week programs per quarter and 1–2 workshops per month. Reach out below — I reply within 48 hours.
Tell me the situation — an architecture decision, your agent setup, a co-founder knot, or a team you want to level up. I'll reply within 48 hours with a fit or a kind redirect.