> Let’s face it: by the time I manually ship version 1.0 of a product, the AI-assisted version could have been deployed 10x faster. By then, enough real-world feedback would have surfaced to identify the major issues, and tools like Claude Code would make it possible to fix and ship version 2.0 at an incredible pace.

This precisely why I still have mixed reaction towards AI, even AI can produce functional code but might be filled with foot guns. I personally don't use AI (the full automated ones, e.g., Claude code, Codex, Cursor) but also I don't complain about people using AI.

This also reminds me of Jonathan Blow's Software is in Decline[1] talk. Even when the humans coded everything, we gave up on quality a long ago for speed. So people complaining about low quality AI code is ignored.

Simply put software engineering is not as rigorous as other engineering and most of the time when software ultimately fail there isn't major consequences.

[1] https://www.youtube.com/watch?v=FeAMiBKi_EM

Yesterday, I asked it to build a Minecraft clone. After about 20 minutes it gave me a basic voxel game with infinite world generation etc.

Obviously so many people made Minecraft clone long before LLMs so I would train dataset contained several implementation of it, hence it made me a working game in 20 minutes.

But it makes me wonder, everyone whose getting into the field probably used these tools to program projects and put it on their resume as well.

so how do y'all vet resume in 2026?

Comments URL: https://news.ycombinator.com/item?id=48346399

Points: 15

# Comments: 10