A veteran software engineer who shifted from general hacking (2007‑2014) to AI work expresses strong enthusiasm for large language models, autonomous driving, video generation and coding assistants. He reports running a local GLM‑5.2 instance on a Linux box and using it to install tmux with a custom configuration, calling it a sign of the "Year of the Linux Desktop."

He criticizes two recurring narratives: hype that paints AI progress as a personal failure and sensational claims that LLMs will cause sudden, uncontrollable disruption. He links these narratives to market players who, he argues, fear open‑source commodification and seek large valuations. The author cites a 2016 superintelligence presentation and a 1991 sci‑fi film to illustrate the long‑standing nature of such fears.

He notes that programming productivity is shifting, comparing agent‑driven assistance (10× gains) to compiler advances (potentially 1,000×). He acknowledges cognitive fatigue from model use and the current lack of "magical" software that fully realizes promised gains, but treats LLMs as practical tools comparable to find‑replace, Stack Overflow searches, or regex shortcuts.

The piece frames AI as a continuation of the broader computing revolution, driven by Moore's law rather than exclusive corporate breakthroughs.