Bitkom's Softwarewelt 2036 study says the same thing a lot of us have felt while working: writing code is becoming a commodity, and the value moves to judgment, domain knowledge, architecture, and trust. Here is what the study gets right, read through the eyes of someone doing the work, and the parts worth being skeptical about.
Since early August, Claude embeds an invisible watermark in the text it generates, driven by the EU AI Act. It survives copy-paste, not paraphrase, and it proves one thing that most of the reaction gets wrong: that Claude processed the text, never that Claude wrote it. Which is the right frame for writing with AI at all, use it to turn knowledge into prose, and verify every claim it hands you yourself.
Most people run an AI to write code and then review the output by hand. The step almost nobody adds is the feedback loop - let the CI pipeline produce a report, hand it back to the AI, and let it fix its own mistakes on the feature branch before a human ever looks. It cuts the amount of AI code you have to review by a lot, because the model catches its own garbage first.
Big tech has discovered a new way to measure engineering productivity. The metric is the number of AI tokens an engineer consumes per month. It rewards the exact opposite of competent engineering, and the engineers who understand this are being systematically ranked beneath the engineers who do not.
Modern pull request review is built on a clean trust boundary. Code in the PR is untrusted until reviewers approve it. Configuration in the PR is treated as ambient context, more or less along for the ride. The Gemini CLI hook CVE that landed in May 2026 made it
Production-grade public MCP server