Developers warn: AI coding assistants promote cognitive decline and create 'Cognitive Debt' in organizations
What it really says
On June 18, 2026, Golem.de reported on a growing problem in software development: programmers who work daily with AI assistants like GitHub Copilot or Claude Code are experiencing a gradual loss of their skills. Cybersecurity expert Michael Taggart puts it bluntly: 'It quite clearly makes me dumber.' On platforms like Reddit and Hacker News, reports are mounting from developers who describe being unable to handle basic programming tasks as well as before without AI support. The article identifies two levels of the problem: at the individual level, problem-solving abilities and deep technical understanding atrophy when developers routinely delegate tasks to AI. At the organizational level, 'Cognitive Debt' emerges - a concept analogous to 'Technical Debt': companies accumulate cognitive debt when knowledge about their own systems increasingly resides in AI tools rather than in employees' heads. There is also a motivation problem: many developers find AI-assisted work boring because the creative and intellectually challenging parts of programming are handed off to the machine.
Our assessment
This report warrants a yellow rating - the concern is legitimate but there are ways to address it. What is true: the described effects are real and supported by cognitive science research. Skills that are not regularly practiced deteriorate - this applies to programming just as it does to mathematics or foreign languages. The 'Cognitive Debt' analogy is apt: like technical debt, you only notice the decay when it is too late - for instance when a critical system fails and nobody understands the code anymore. What needs qualification: similar fears arose with every new level of abstraction in IT - from assembly to high-level languages, from manual to automated operations. The solution lies not in rejecting the tools but in using them deliberately: AI as assistance, not as a replacement for one's own thinking. Companies should specifically schedule time for developers to work without AI to maintain their core competencies.
Relevance for Germany
This topic is particularly pressing for Germany because the IT skills shortage is already severe. If existing developers lose their abilities through excessive AI use, the problem worsens. According to PwC's 2026 Global AI Jobs Barometer, only 21 percent of workers in Germany feel competent enough for AI tools - indicating broad uncertainty that could be amplified by the described skills loss. At the same time, the PwC study shows that companies using AI to augment rather than replace employees are more successful - confirming the approach of treating AI as a tool, not a substitute for human competence. German companies and training providers should embrace 'AI hygiene' as a concept: deliberate periods without AI support, regular code reviews without AI, and mentoring programs that keep knowledge person-bound.
Fact check
The Golem.de article from June 18, 2026 is the primary source and quotes Michael Taggart by name. The descriptions of developer experiences on Reddit and Hacker News are community reports, not scientific studies. Golem.de itself had reported on cognitive effects of AI use in April 2026, referencing scientific studies on weakened problem-solving abilities. The 21 percent figure for AI-competent workers in Germany comes from PwC's 2026 Global AI Jobs Barometer, which analyzed over one billion job advertisements across 27 countries.
Source
- • https://www.golem.de/news/gesellschaftliche-folgen-von-ki-wie-ki-in-der-it-den-kognitiven-verfall-foerdert-2606-209867.html
- • https://www.golem.de/news/kognitive-auswirkungen-ki-nutzung-schwaecht-ausdauer-und-problemloesungskompetenz-2604-207621.html
- • https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-jobs-barometer.html