Back to blog
Mar 09, 2026
3 min read

Wading Through AI

Casey Muratori and veteran AI researcher Dimitri Spanos dropped a video named Should I Become a Carpenter?. They explored the realistic, unhyped impacts of AI on the current and future job market

I’d encourage you to watch the full conversation, but I’ve dropped the tldr; points and timestamps below.

The Threat to Junior Engineers

  • Hiring is Plummeting: The immediate threat is to junior developers. Consensus estimates among recruiters suggest junior engineering hiring in the US is currently down by 50% to 70% [34:00].
  • The “Pipeline Problem”: If companies stop hiring juniors because AI can do basic “Stack Overflow copy-paste” tasks, the industry won’t produce the senior engineers it will desperately need in the future [37:56].

The Realities of AI Code Generation

  • Current Capabilities: AI is currently reliable for generating around 1,000 to 2,000 lines of standard boilerplate code (e.g., simple web dashboards or CSV parsing scripts) using agentic loops [01:05:48]. However, it completely fails at specialised or obscure tasks like numerical analysis.
  • ”Code Sycophancy”: Because AI models are trained via Reinforcement Learning from Human Feedback (RLHF) to please the user, they will frequently flatter bad ideas and generate poor code if your prompt is leading or incorrectly assumes a bad practice [42:16].
  • Jevons Paradox & “Slop”: Because the cost of generating code has dropped to nearly zero, companies will start generating massive amounts of low-value, automated code that wouldn’t have been worth paying a human $200 to write previously [01:07:10]. This risks flooding the world with massive, unmanageable codebases.

Career Advice: Should You Be A Carpenter?

  • Learn the Boundaries: If you are entering the job market today, you cannot ignore AI. You must familiarise yourself with what it can and can’t do by testing it on typical entry-level tasks in your industry [01:20:43].
  • Shift Away from Syntax: The specific skill of typing out “syntactic constructs” (writing code manually) will likely lose its market value. Programmers will need to pivot their focus toward system design, problem-solving, and understanding adjacent business needs like game design, market opportunities, or client relations [01:16:15].

Impacts Beyond Programming

  • The Legal Industry: A similar shift is happening in law, where tools like ChatGPT are matching the output of first-year lawyers and paralegals when it comes to summarising thousands of case documents, making it harder for law firms to justify entry-level salaries [01:29:32].
  • The Art Industry: While high-end 3D artists and AAA game riggers are safe for the foreseeable future, lower-tier commercial art is in trouble. “Shovelware” apps are expected to transition entirely to AI-generated assets within the next five years [01:33:35].

(Video URL: https://www.youtube.com/watch?v=RJyPVLMyyuA)