Video summary

There are only two types of programmers left

Main summary

Key takeaways

News and Commentary

Overview

The video argues that today’s programming culture has polarized into “extremes,” leaving two main archetypes of programmers:

  1. The “AI hype beast” extremist
  2. The “anti-AI Andy” extremist

It then proposes a third, “realist” archetype as a better model, and later transitions into a sponsor segment about GitHub Actions performance.

Archetype 1: The “AI hype beast” extremist

  • Embraces AI coding as effectively already solved.
  • Often stacks multiple AI/self-improvement agents and productivity loops (e.g., “Atomic Habits”-style repetition and automation).
  • Relies heavily on AI-generated outputs, often paired with music, workflows, and other automation.
  • Responds to common AI-coding critiques (implied: feasibility, reliability, and requirements like citizenship) by treating them as delusions rather than practical engineering concerns.

Archetype 2: The “anti-AI Andy” extremist

  • Rejects AI on principle, often driven by claims that the “AI bubble” will burst.
  • The speaker argues this stance isn’t automatically invalid—some top developers genuinely don’t care about AI.
  • Examples include Jonathan Blow and other elite developers who can afford to ignore AI because they aren’t pressured by corporate performance metrics (e.g., “token maxing,” deadlines).

Horseshoe Theory Analogy (and the core critique)

The video claims these two groups are extremists—not positions on a spectrum. It compares the dynamic to horseshoe theory:

  • Despite outward disagreement, both camps lack room for nuance and good-faith discussion.
  • Software engineering already contains dogma and opinion, but AI debates amplify it rather than improving the craft.

The “realist” middle archetype (the better model)

As an alternative to both extremes, the speaker presents a third archetype:

  • Mitchell Hashimoto (the core example)
  • Portrayed as a true software craftsman who uses AI pragmatically, with skepticism:
    • Uses AI agents in limited ways (e.g., “two agents at a time”), emphasizing review/verification instead of blind trust.
    • Performs side-by-side comparisons (e.g., “poison-test” style evaluation of AI outputs; comparing AI tools to human-written code).
    • Avoids chasing hype, anxiety, or maximum profitability—prioritizes shipping high-quality software and maintaining skill.

Conclusion (the video’s claim)

There is a middle ground: use AI tools to get very good while still improving your own programming fundamentals—avoiding both:

  • hype-driven overreliance, and
  • purist anti-AI ideology.

Sponsor Segment: GitHub Actions Performance

The video shifts into a sponsor segment focused on GitHub Actions performance.

  • Sponsor: Blacksmith
  • Claims:
    • Makes GitHub Actions faster
    • About 60% cheaper than GitHub-hosted runners

What the sponsor demonstrates

  • Benchmark-style examples, including:
    • Android repository build time/cost reductions
    • Postto Docker build time reductions

How it works (as described)

  • Faster bare-metal hardware CPUs
  • Better cache locality
  • NVMe to speed up Docker builds
  • An observability dashboard to inspect failures and logs, helping improve tests and pull requests

Presenters / Contributors (as mentioned)

  • Mitchell Hashimoto
  • Jonathan Blow
  • Gatuso (referred to as “Gatuso” in the subtitles; likely a quote attribution)
  • “Wet hair dry beard guy” (named as a comedian reference)
  • Bigboxweet.dev / Blacksmith (sponsor; no individual person named)

Original video