We're all CTO now

AI as Coding Tool: Promise vs. Reality

• Many commenters say modern models (e.g., frontier LLMs, Claude Code, Cursor) can dramatically boost productivity when used well: as pair programmers, translators from pseudocode, or for tedious edits (e.g., transforming print statements into severity-appropriate logging).
• Others report deeply negative experiences: rapid progress at first, followed by total loss of code understanding, incoherent architecture, heavy global state, and “cargo cult” patterns worse than any human codebase they’d seen.
• There’s disagreement on why: some argue poor prompts and lack of examples are to blame; others say users control only a small slice of behavior and hidden system prompts and training data dominate.

Code Quality, Comprehension, and Maintainability

• Several developers note that when they let AI write most of the logic, they no longer understand the system, making debugging and evolution painful.
• Suggested mitigations: keep AI for translation and boilerplate, but own the logic; refactor AI output aggressively; provide rich architectural context and examples.
• In some teams, obvious “AI slop” and meaningless commit messages are accumulating while product managers resist refactors, trading long-term health for short-term velocity.

Skills and Atrophy

• One side rejects the “skills as muscles” metaphor, arguing coding speed isn’t the bottleneck and rarely-used details can be quickly relearned.
• Others insist unused skills do atrophy and foresee a generation dependent on autocomplete and agents, with interview expectations (e.g., algorithm trivia) clashing with that reality.
• There’s broad criticism of hiring processes that reward memorized algorithms instead of real problem-solving.

CTO/Manager Roles and Motivation

• Commenters describe CTO roles ranging from hands-on principal engineer to pure C‑suite politician; title is often seen as mostly about signature authority.
• Some agree with the article’s lack of dopamine from management; others say they genuinely enjoy mentoring, protecting teams, and solving user problems, and feel the article erases that perspective.

Industry Trajectory and Workforce Effects

• Some anticipate “we’re all CTOs of agents,” doing high-level orchestration while AI writes most code.
• Skeptics predict instead a flood of low-effort “script kiddie” work, with leadership implicitly betting most systems are disposable rockets, not airplanes that must never fail.