Passion
Why we got into this matters as much as what comes out of it. I work to keep the craft and the spark alive in the people doing the work.
For you: a partner who treats your people as the point, not the resource.
Software & SaaS product builder · Hamburg N 53.55° / E 9.99°
Twenty years building software products and the businesses around them. The rare part is the range: I work hands-on across all six domains it takes to build a software company, not just the code.
Selected organisations
The range
Most people who build software go deep in one place. The code, say, or the org, or the numbers. The six below are usually six different people. I have done the work in each, and they inform each other far more than anyone expects.
Where the revenue comes from, and whether the model holds up. I have drawn the canvas more than once, then had to make it pay. Incubators, spin-outs, new revenue lines inside an existing business.
What gets built, for whom, and why it earns their attention. The right two-week feature, shipped at the right moment, has moved a single revenue line into the three-digit millions.
Teams shaped so they can ship without waiting on each other. I have redrawn the org to match the architecture, putting Conway's law to work instead of fighting it.
Treating cost and return as a first-class input to design, not an afterthought. The numbers tell you which complexity earns its keep and which doesn’t.
Twenty years of reactive, vertical, self-contained systems, and now agentic engineering. Writing it, reviewing it, and being accountable when it breaks.
Holding direction when product, organisation, and technology pull in different directions, and one of the three has to bend. Steering is the call about which one, and when.
The constants
Roles and companies changed plenty over twenty years. How I show up to the work did not.
Why we got into this matters as much as what comes out of it. I work to keep the craft and the spark alive in the people doing the work.
For you: a partner who treats your people as the point, not the resource.
Twenty years in, I still go at the hard problem until it gives. A hurdle is the part of the climb where most people turn back.
For you: the work gets pushed past the comfortable answer, not parked at it.
When something is wrong, I name it early, in the room. The hard conversation now is cheaper than the one you delay.
For you: no quiet drift, no late surprises. The risk is on the table while it is still cheap.
The work counts when the bar moved and the team is still the kind of place people want to be on Monday. That is the result I measure against.
For you: an outcome you can see in the team, not just in the deck.
On stage
I have spoken at conferences since 2016. The talks I am proudest of started the same way: something was bugging me at work, and the talk was how I worked it out in public.
Current topics
How teams move from random prompts and inconsistent output to AI-assisted development with real context, feedback loops, quality gates, and human ownership.
For developers, tech leads, and CTOs
AI makes code cheap to produce. That makes architecture, boundaries, testing, review, and maintainability matter more, not less.
For architects, senior engineers, and leaders
Why fulfilled teams move faster, why friction hides in everyday workflows, and how tooling and leadership shape delivery you can sustain.
For CTOs, leaders, and platform teams
What Hacker School, mentoring, and AI-assisted development taught me about how knowledge actually moves between people.
For developers, mentors, and educators
Formats
Speaker bio
Benedikt Stemmildt is a founder, software architect, speaker, educator, and agentic engineering advocate based in Hamburg. He is the founder of hackers&wizards, a founding member and long-time Inspirer at Hacker School since 2014, and part of the alphalist and code.talks ecosystem. With 20+ years in software engineering and leadership roles at OTTO, Breuninger, BLUME 2000, and TalentFormation, he helps teams build better software with more clarity, craft, and human responsibility in an AI-assisted world.
Selected stages
Recent talks
2024
Retrieval-augmented generation over Statista's own data: the cost and latency of each request, staffing a team for a technology this new, and why a model still hallucinates with the right data in front of it.
2024
Building a web application with no JSON, no virtual DOM, and no separate frontend deployment. A working e-commerce shop in Kotlin, Spring Boot, and plain Tailwind.
2023
Efficiency gets heard as cost-cutting, and that reading was always too narrow. Building an IT department that gets stronger under stress, not just cheaper.
The path
Developer, architect, CIO, CTO, founder. The titles changed. The work, building software organisations that move fast and stay sane, mostly did not.
2009 — 2013
Rotated through controlling, SAP HR development, e-commerce, and corporate strategy. Helped get the Liquid Labs incubator off the ground, where Risk.Ident later spun out as its own company.
2013 — 2017
Hired onto otto.de, starting in a SCRUM team building a new backoffice. Co-founded a small team focused on social media features. Two weeks in, a feature shipped from that team moved a single revenue line into the three-digit millions.
2017 — 2020
The online-shop platform, built by 13 teams and 70 developers. I established a vertical architecture so they could move without waiting on each other. Online order revenue grew 35% in 2018.
2020 — 2022
Owned e-commerce IT. The department was not set up for what was coming, so we rebuilt product, organisation, and technology at once, and shipped a new shop on the new foundation.
2022 — 2024
Company rebuilding, architecture, and delivery. Where agentic engineering moved from an interest to the centre of the work.
2024 — NOW
Turning agentic engineering into something teams can actually use. Training, workshops, and AI-assisted development that keeps the craft intact.
Working notes
These are working notes, not finished thinking. The arc I keep coming back to is how teams move from one-off prompts to a working system that builds software with them, and what stays a human's job along the way.
Crafting the right prompt for one task. The skill is real and the output can be sharp. But the win lives in that prompt, not in the team, so it does not compound.
Durable, reusable context, so people and AI work to the same standards. Decisions, conventions, and review criteria, written down once and used by everyone.
Workflows where AI carries more of the work while humans keep judgment, direction, and accountability. Not replacing developers. Raising the level they work at.
A chain of specialised agents across the full development cycle, owned by the team that uses it. The work shifts from writing every line to setting the rules the chain runs on.
You can outsource your thinking, but you cannot outsource your understanding.
What I believe so far
A clever prompt helps once. Shared context compounds. Decisions, standards, and review criteria are worth treating as reusable assets, not things you re-explain every time.
Useful AI work looks more like mentoring than commanding. You give context, examples, boundaries, and a clear sense of what good looks like. The same things a new teammate needs.
Past a point, the team is not shipping features. The team is shipping the chain of agents that ships the features. The work moves up a level, and so does the skill.
AI-generated code is still code. It needs tests, review, a fit with the architecture, operational thinking, and a human who owns the result.
Lines I keep
Any sufficiently advanced technology is indistinguishable from magic.
Arthur C. Clarke
The amount of energy needed to refute bullshit is an order of magnitude bigger than that needed to produce it.
Alberto Brandolini
Statistics only works when the person doing the statistical evaluation is, on average, more intelligent than chance.
Peter Kruse
The person
At ten I took my father's computer apart to see the part that made it work. I got it back together, mostly. The curiosity stayed.
That thread runs through all of it since. Writing software, drawing system boundaries, leading teams, teaching kids and students, and now working out what AI means for the craft.
I studied business informatics as a dual student, university and OTTO at the same time, and learned one thing early: the interesting problems are rarely only technical. Developer, architect, CIO, CTO, and founder all followed from there.
Read the full storyHow do we use AI to make software teams more capable without losing craft, judgment, or ownership?
Borrowed thinking
The people below solved problems I keep running into, often in another decade or another field. When I am stuck between a clever solution and a clear one, I am usually borrowing one of their ideas.
Heroes
Software engineering is an experimental practice.
David Farley
Continuous delivery
As independent as possible, as integrated as necessary.
Stefan Tilkov
Self-Contained Systems
Refactoring is how good software gets built, not cleanup work.
Martin Fowler
Software design
Make it work, make it right, make it fast.
Kent Beck
Test-driven development
Quality enables speed.
Nicole Forsgren
DORA research
Give control away instead of giving orders.
L. David Marquet
Leadership
Happy engineers build better products.
Gene Kim
DevOps
What happens to human purpose when machines get good at almost everything?
Max Tegmark
AI and society
Fast, automatic thinking and slow, effortful thinking.
Daniel Kahneman
Decision-making
You learn to paint mostly by doing it. Ditto for hacking.
Paul Graham
The hackers&wizards name
Every solution creates new problems.
Liu Cixin
Second-order effects
From the reading list
Off the clock
Most of this site is about software. This part is the rest.
I live in Hamburg with my family, and the rest gets arranged around that. Being a dad reset what is actually worth being stressed about. Most of it turns out not to be.
I think better after I have moved. Calisthenics and running are the steady part. Surfskating is the part I do because it is fun. Tennis is competitive without it mattering.
I play long-form RPGs and strategy games. A well-built one is systems design told from the inside. You notice when the designers respected your time and when they did not.
I still tinker. A functional programming idea, a domain model redrawn a few times to see if it gets cleaner. No deadline, no point beyond wanting to understand the thing.
Connect
Tell me what you are working on, what you have already tried, and what kind of answer would help. I would rather start there than with a vague "let's connect at some point".
The best invitations include the audience, the format, the date, the language, and the one question you want people to leave thinking about.
Invite me to speakhackers&wizards
Commercial work: agentic engineering, team training, and AI-assisted development that changes how an organisation builds.
Visit hackers&wizardsEducation and Hacker School
Digital education, teaching programming, and getting young people curious enough to build with technology, not just consume it.
Visit Hacker SchoolCommunity and peers
alphalist, code.talks, CTO conversations, software architecture exchange, or a book worth arguing about.
Send a note