The Talent Forge With No Curriculum

The four-year degree was built for a slower machine. Print the curriculum once, sort a population with standardised tests, hand out the queue position, and let the employer finish the training on the job. The last part was the bridge. The entry-level role absorbed a graduate who knew nothing operational and turned them into somebody useful across 24 months of paid, supervised practice. Take the bridge away and the queue position buys much less. The bridge is largely gone. It was eaten from below by automation and from above by the demand for experience years before the first performance review. A graduate is left holding a document that certifies attendance at a system that no longer runs the way it was designed to run. The hiring problem came first. A group building industrial AI could not find the combination it needed, technical depth plus execution instinct in the same person at the numbers the work required. The school came second, as a fix for that gap and not as a position on education.

A curriculum is a photograph of what was true on the day it was written. The half-life of a technical stack is shorter than the length of a degree. So the rule here is blunt: the curriculum is rebuilt weekly. Not because rewriting material is a pleasure. Because anything static is already obsolete by the time a student finishes reading it. That inverts the usual order of authority. The syllabus stops being the contract and becomes a working document. What stays fixed is the standard of output, not the reading list. The subjects that survive the rewrite are the ones tied to a job that has to run: neural operators, multi-spectral vision, autonomous agents, edge hardware sitting in a factory.

No grades. No exams. No lectures. A participant is placed inside a live company as a contributor with real responsibility: shipping features, debugging production systems, presenting to clients who hold budgets and deadlines. The measure of progress is deployed work. A portfolio of shipped products is a claim a hiring manager can test in an afternoon. A transcript is a claim they have to take on trust. That difference decides interviews. The probability that an employer chooses to verify a credential instead of testing the candidate is falling every year, because the verification now costs one link and ten minutes of reading.

Admissions have run past 25,000 applications to fill 25 seats in a cohort, with 210 alumni holding passports from 19 countries. That is a funnel near one in 1,000, and the ratio is the design rather than the marketing. Direct access to the people running the ventures is the product. Access cannot be diluted across a lecture hall without turning into a recording, so every extra admission spends the current cohort's attention. The other model, admit by the thousand and grade on a curve and certify by the thousand, optimises for tuition and produces a document the market discounts. A cohort of 25 optimises for placement. One trajectory from the record: a participant who arrived straight from a Goldman Sachs internship in Singapore built across the waste-to-energy venture and the wider group, and 12 months later was running operations for Spain out of Malaga and Las Palmas.

A transcript is a claim they have to take on trust. Deployed work is a claim they can test.

AI is the operating environment from day one, not a module bolted onto a later semester. The systems in production span multi-model orchestration, edge computer vision on NVIDIA Jetson hardware, neural-network digital twins for process engineering and agentic pipelines serving industrial clients across 9 German federal states. Research partnerships run with TU Zittau and Fraunhofer IKTS. Placements sit inside real ventures with paying clients, a second campus opens in Malaga in September 2026, and executive training runs in Las Palmas. None of that matters as a brochure line. It matters as the difference between a case study written for a classroom and a client who depends on the thing you built.

Three questions separate a programme from a credential mill. What will I ship, and who will use it? Who signs my reference when I leave? What is still running in production 12 months after my last commit? A programme that answers all three is selling a job-shaped education. One that answers with accreditation language is selling a document. A credential was a proxy for capability at a time when capability was expensive to test. Testing is cheap now. The portfolio carries the evidence, and the interview is where somebody finally reads it.