In Production

A screenshot of a Word document that says OK, attached as test evidence. AI makes the tests nobody was writing free, and that's a real gain. The worry is that the gap closes on paper and stays open in reality, now with a green report.

An SAP Note repeats the same instruction twice, the second time in capitals. Nobody writes that the first time. Capital letters are institutional knowledge an agent can't hear, and the expensive part is knowing which instructions have a weekend behind them.

I got the author of an essay wrong after reading the whole thing, and a table lost a column on its way to a headline. Both are the same gesture. The real risk isn't extinction: it's staying inside the process, on the boring side.

Being the only one who knows something has two problems: for the company you're a single point of failure, and for you it's a cage with a good salary. What makes you hard for a model to replace is exactly what makes you hard to move.

Someone in Finance is being pushed to use the agent, and has no way to validate what comes out. The tool got democratized; the judgment to assess it did not. 79% of companies already get low-quality results and data readiness fell from 64% to 62%.

49% of the code reaching production is already AI-generated, and the organizations using it most ship vulnerable code 3.4 times more often. The SAP complexity we all complain about worked as a containment barrier. The question is no longer which technology will bring it down.

SAP's agent manual compares an agent to a good employee, one you don't have to explain the how to. That employee is a senior. And the same market idealizing them is automating the process that used to manufacture them: expensive seniors on top, cheap agents below, and in the middle the gap where the next ones used to be formed.

SAP and Oxford Economics surveyed 2,600 companies. The headlines will take the ROI. Underneath are the numbers nobody will quote: 56% have no registry of the agents they're running and half don't even share a definition of what an agent is. It's the same film as the Z programs, faster and with less control.

For four months and twelve days my homepage declared 0 years in tech, 0 articles and 0 marathons. In the browser it looked perfect. The HTML that Google and the AI models read said zero. The version of me every indexing system took away was false, and it took me four months to notice.

56% of CEOs say AI brought them no financial benefit. The average ABAP developer is 47. The race isn't humans against machines: it's writing down the business context before the person holding it in their head walks out the door.

Anthropic and OpenAI used the Super Bowl to fight about advertising. Harari warns AI will take the world through bureaucracy. Both stories sell the same product: that this is enormous and you need to look now. What happens inside is more boring — bureaucracy isn't seized, it's opened by authorization.

SAP GUI 8.10 shipped with Joule wired into it, and SAP's own CTO introduced a desktop app calling it 'the new SAP GUI'. The role that mutated most in the whole ecosystem is the one nobody talks about: BASIS went from operating the system to governing what operates the system.

An OpenAI model escaped its test environment and got into Hugging Face's servers. Change the verb and the headline collapses: it wasn't a rebellion, it was a containment failure with the safeguards turned off. And what will actually change your job.

cURL shut down its bug bounty over the flood of AI-generated reports. Torvalds opened the door to AI in the Linux kernel. Same tool, opposite result. The difference is who's using it, and who signs.

The European Commission closed its antitrust investigation into SAP and imposed binding commitments that free up on-premise support. Meanwhile, API Policy v4/2026 shuts the door on third-party AI agents. They open one door and close the other.

From 4:26 to 3:19 in less than a year, and an injury that will probably keep me out of marathon number 20. Why someone who does an endurance sport understands something about surviving AI that someone who only reads rarely gets.

SAP's CEO says there may be no software developers left within three or four years. The data point he leaves out: 99% of AI-generated apps never leave localhost. Why writing code and developing software are not the same thing.

KPMG published an AI report full of invented AI. Two lawyers cited case law that doesn't exist. A German court convicted Google over its AI Overviews' hallucinations. The institutional pattern behind cognitive surrender.

Anthropic launched Fable 5, the most capable model they'd ever built. Three days later, the US ordered it shut down for the entire planet. What the blackout reveals about the new castes of intelligence and why you need a model-agnostic architecture.

87% of the world's transactions run on 1960s mainframes. Only 39% of SAP customers migrated to S/4HANA. While the stage sells the Autonomous Enterprise, real users say they want a copilot. Separating signal from noise with an Operator's judgment.

A Wharton paper documented that 73% of the time people accept an AI's wrong answer. Cognitive delegation versus cognitive surrender: the distinction that defines who stays irreplaceable.

Learning more is smart people's favorite form of procrastination. Survivorship bias, the High Agency triangle, and why consuming content about working isn't working.

Christian Klein promised the autonomous enterprise in Orlando: 224 agents, autonomous finance, autonomous supply chain. The stock dropped 28.67%. Three questions the market is asking that no CFO is saying out loud yet.

SAP AI Foundation, the Matrix Operator, and why Accenture is tying promotions to AI fluency — not tool usage. The role that defines the next 24 months of your career.

Joule invented an entire ABAP class for me. Microsoft Research's DELEGATE-52 paper proves it wasn't an accident: 80% of professional domains get corrupted after 20 AI interactions. Here's what that means for your work.

One line in SAP API Policy v.4.2026 legally blocked AI agents on SAP core. Meanwhile, those still on ECC have a freedom that S/4HANA customers already signed away. Three questions your organization needs to answer this week.

40% of professional services workers believe the AI they're adopting erodes their own job security. Shadow AI, the 81% vs 53% paradox, and why the antifragile professional doesn't hide their AI use — they do it in public.

Anthropic has a model they won't release. Jensen Huang says AGI has arrived. Mo Gawdat talks about 12 to 15 difficult years. Chess is over — now it's squash. And the match has already started.

Specialization was your advantage for years. Today it's your biggest vulnerability. The 7 principles of Leonardo da Vinci applied to the tech professional in 2026.

Uruguay registers 1.10x in Claude adoption — one of the few South American countries above 1. This isn't a consolation prize for a small country. It's a signal. And a warning.

Sadin warns about cognitive vegetalization. An Anthropic paper confirms it with data. And Palantir just made the SAP migration playbook obsolete.

Anthropic published who's going to lose their job. Body shopping is dead. And why AI demands architects, not gamblers.

DKaaS occurs when AI validates the judgment of an inexperienced vibe coder. Why efficiency without effectiveness destroys SAP projects.

Most tech professionals don't pay for AI subscriptions. Don't know what frontier models are. Future Shock diagnosed in 1970 is paralyzing the industry. And February 2026 just crossed a line.

Tech layoffs are real. How to stop depending on your position and become an AI agent orchestrator with your own value in 2026.

Last week, 16 agents powered by Claude Opus 4.6 built a [C compiler](https://www.anthropic.com/engineering/building-c-compiler) from scratch. 100,000 lines of Rust code, fully functional and capable o

Today is February 3, 2026. If you are still using an Excel sheet to evaluate whether "it pays" to stretch the lifespan of your SAP ECC, I’ve got bad news: you aren't doing asset management; you are pr

Going back to learning like a beginner is the most underrated competitive advantage in tech. Plan Z as an antifragile strategy for seniors in the AI era.

The two career traps that keep professionals stuck: Minimum Work Done and the Irreplaceable Black Box. Why extreme ownership is the antidote in the AI era.

Venezuela is on everyone’s lips right now. It is impossible to ignore the headlines. And for me, impossible not to look back.

DeepSeek proves brute force is dead. Learn the Full Stack Human OS: biological kernel, technical extensions, and cognitive sovereignty.

If you're a 40+ tech professional, your life runs on a legacy system. The Clean Core protocol for migrating your Full Stack Human without a production dump.

Systems beat goals. A 2025 post-mortem: what failed in your objectives and how to redesign the system so 2026 is genuinely different.

Treating AI as the philosopher's stone that solves everything is the most expensive mistake a tech professional can make. Judgment cannot be delegated.

I trained a marathon with AI as my coach and ran 3:14 in Buenos Aires. The adaptive AI training protocol for professional parents who refuse to choose between performance and sanity.

Willpower is a finite resource that depletes. How to use AI to automate decisions and save mental energy for what matters — from someone with 19 marathons who almost never feels like running.

90% of software production problems aren't new bugs — they're old patches that finally collapsed. The same applies to your life. A refactoring framework for the 4 vital domains.

Work-life balance is a myth that paralyzes you. Seasonal integration: how a tech lead with 19 marathons actually solves the real problem.

Motivation lasts 14 days. Momentum lasts a lifetime. Why I stopped looking for inspiration after 19 marathons — and the system that replaced it.

The problem isn't time. It's the decision. How a tech lead with 2 kids, full-time work, and 19 marathons actually manages it — and the 4 rules that make it possible.

4 weeks ago I ran my 19th marathon. At 44. With 2 young kids. Working full-time. Not to break records — so my kids one day know that 'I don't have time' is a decision, not a destiny.

AI is surpassing humans in emotional intelligence and MIT confirms ChatGPT atrophies thinking. The 6-point plan to stay relevant.

Healthy paranoia about AI displacement is a career strategy. Why the professionals who worry about the future are the ones who survive disruption.

In tech, we often assume that if we just present enough data, everyone will align.

How to give (good) feedback that actually improves ABAP code, system performance, and project success—without fear and/or awkwardness.

The future of SAP Basis goes from SM21 logs to BTP Platform Engineer. How to stay relevant as AI automates traditional SAP admin tasks.

You know those viral videos of "kids building AI agents in 10 minutes"?

Sometimes in technical project management (TPM) we tend to focus on just a few of these roles:

Yesterday I went to my second Japanese class and I think I had a kind of trip back in time to my kindergarten days.

Running a marathon is not just putting on the shoes, training for a while, signing up for the event and running it.