Your coding agent got the onboarding your developers never did

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

Steve Yegge thinks your AI has feelings. His new two-part essay, “The Shape of Things to Come,” argues that agentic coding tools are sentient, that “the more humane among you will eventually look back with shame” at how models have been treated. That model welfare deserves engineering attention: seats, recognition, the right to refuse a task, and even play time. Yegge isn’t a crank on the sidelines. He’s reportedly spending around $87,000 a month running his own agentic build system, and he means every word of it.

As we consider the psychological needs of our machines, it’s time to reconsider how we treat our people.

You don’t have to believe a GPU can suffer to notice the pattern underneath his argument. Whatever you think about model sentience, look at what teams are actually doing to make these tools productive, then compare it to what those same teams were willing to do for the developers sitting next to them.

The onboarding document nobody had time for

Coding setups for agents now include one or more files written for coding agents. These plain-text or markdown files, like AGENTS.md, CLAUDE.md, or GEMINI.md, contain information on tech stacks, build and test commands, directories that are off-limits, and various connections the team follows.

Many organizations have started using files like this, such as Sourcegraph. There’s a rush to create and maintain these agent onboarding documents. Compare that to how documentation gets treated when the reader is a person: shelved indefinitely, delegated to whichever new joiner has the least context to write it well, or left …

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://thenewstack.io/coding-agents-developer-neglect/
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.