Serien
Drei Bögen, erzählt in Atomen.
Jede Serie ist eine zusammenhängende Folge kurzer LinkedIn-Beiträge, geschrieben über zwei Wochen. Gemeinsam tragen sie ein Argument, das kein einzelner Beitrag tragen kann. Die ausführliche Synthese jeder Serie steht unter Schreiben.
SERIE 01 · 4 ATOME · APR. 2026
AI Cognitive Debt EN
Speed is the input. Understanding is still the job. An arc on the 83-point gap between AI adoption and organizational output, and the architecture move that closes it.
SYNTHESE · The Invisible Bill: What Your AI Tools Are Really Building → 01 93% of engineers use AI coding tools.
Org-level productivity moved about 10%.
That 83-point gap is not a tooling problem.
It's an architectural one. THE AI PRODUCTIVITY PARADOX → 02 GitClear analyzed 211 million lines of code.
Refactoring collapsed from 25% to under 10% of all code changes since 2021.
Code duplication grew 4×.
This is not a forecast. This already happened to your codebase. THE EVIDENCE: YOUR CODEBASE RIGHT NOW → 03 By year two, AI-generated code costs 4× more to maintain than traditionally written code.
68% of AI projects exceed their initial budgets — by an average of 42%.
Nobody put that in the business case. THE HIDDEN COST NOBODY PUT IN THE BUSINESS CASE → 04 Thoughtworks just named "cognitive debt" the defining AI risk of 2026.
Technology Radar Vol 34. Published April 15.
Most engineers haven't read it yet.
Here's what it actually means for your team. THOUGHTWORKS NAMED IT. NOW YOU NEED TO EXPLAIN IT TO YOUR ORGANIZATION. →
SERIE 02 · 5 ATOME · APR. 2026
Generative AI Strategy for Leadership EN
$37B spent. 80% no EBIT impact. Where boardroom AI strategy actually fails, and why architecture decisions are strategy decisions in disguise.
SYNTHESE · The $37B Question Nobody's Answering → 01 $37 billion spent on GenAI last year. 80% of companies saw zero EBIT impact.
Not "marginal." Not "hard to measure." Zero. THE $37 BILLION EBIT PARADOX → 02 40% of enterprise agent projects will be dead by 2027.
I've been deep in multi-agent architecture for months now, and I can already see which ones are headed for the graveyard. THE AGENTIC AI GRAVEYARD → 03 Inference spending just overtook training spending for the first time. $9.2B → $20.6B in one year.
Your Kubernetes clusters weren't designed for this. YOUR INFRASTRUCTURE WASN'T BUILT FOR THIS → 04 "We'll add governance later."
That sentence has killed more AI projects than bad data ever did. GOVERNANCE SPEEDS UP DEPLOYMENT → 05 The most important AI decision your company will make this year has nothing to do with model selection.
It has everything to do with what you build around the model. YOUR ARCHITECTURE IS YOUR AI STRATEGY →