AI coding assistants promise to write your code, speed up your sprint, and maybe even make engineers obsolete. But what if the people building with them every day see something very different?
In this special Halloween edition of CRAFTED. ā which also marks the showās third anniversary! ā a masked CTO shares what he canāt say publicly: that these tools are powerful, but insidious. In his view, coding assistants are great for auto-complete, but they canāt do what a human engineer does. He says theyāre terrible at starting from scratch and will often suggest code that āworks in vacuumā, but not in context. And because AI can write so much code, so quickly, itās hard to catch errors. In short, he sees an increase in short term velocity, at the expense of increased defects and an increasing dependency on systems that are untrustworthy.Ā
I want to emphasize that this episode features the experience of one very experienced person. There are obviously others who disagree, who say AI coding agents are incredible, so long as theyāre managed well.Ā
However, there are also an increasing number of people questioning the sustainability of coding agents ā they're incredibly expensive to run ā and also how good they are in the first place.
For example Andrej Karpathy, the guy who literally coined the phrase "vibe coding" and was early at OpenAI and Tesla, just said publicly on Dwarkesh Podcast that the path to AI agents is going to be a lot slower than people in the industry think it will be. He said coding agents are "not that good at writing code that's never been written before" and that there is too much hype right now about where AI really is, with people in the industry, quote "trying to pretend like this is amazing, when it's not."Ā
And he said: "My Claude Code or Codex still feels like this elementary-grade student."Ā
Today's guest agrees with Karpathy on a lot of this. Our guest has worked at startups, scale-ups, and big tech companies you've definitely heard of and today he's at a very AI-forward company and using AI coding tools every day.Ā
Enjoy this special episode of CRAFTED.!Ā
---
And pretty please...!
- Share with a friend! Word of mouth is how podcasts grow!
- Subscribe to the newsletter at https://www.crafted.fm
- Share your feedback! Iām experimenting with new episode formats and would love your feedback on this and other episodes. DM me on LinkedIn or contact me email, via https://www.crafted.fm
- Sponsor the show? Iām actively speaking to potential sponsors for 2026 episodes. Letās talk!
- Get psyched!⦠There are some big updates to the show in 2026!
---
Key Quotes
- 03:16 The myth of AI replacement: āThe idea that AI can actually supplant a software engineer in their current role is basically nonsense.ā
- 06:29 Why AI struggles without human input: āIf you remove the human engineer from the equation, thereās no place to start from. The AI does not do well when youāre starting from scratch because it doesnāt have the real-world context or the continuous learning required to make that system better.ā
- 12:21: The illusion of speed: āCoding assistants help you generate code very quickly. Thereās an illusion that your velocity increases. What actually happens is youāre just shipping more bugs to production.ā
- 13:30 More code than humans can review: āAI generates so much code that no human can keep that context in their head and review it in a meaningful way. At some point you just have to trust ā but who are you trusting? Youāre trusting the AI, and the AI cannot be trusted.ā
- 14:02 AI & Junior Engineer Hiring: āThe narrative that hiring trends have anything to do with AI is absurd. Itās not that AI is replacing junior engineers ā itās that companies are running lean and donāt have the bandwidth to train them.ā
- 15:42: Where the AI Bulls and Bears Differ: āWhereas we see flawed systems that aren't ready for primetime [...] they view this as āoh, that's, that's insignificant. They will get better almost immediately. It's not a big deal.ā But we've been repeating this cycle for years at this point.ā
- 19:50 Where AI Excels: āWhere review and revise are part of the process already, that's a really good place for generative AI because you already have a human in the loop.ā
- 21:02: What builders need to unlearn āTo the extent that people think these things are thinking or reasoning or on any path to AGI at all ā they should discard that. These models donāt think. Theyāre very sophisticated pattern-matching machines, and thatās really it.ā