I've been a hands-off-the-keyboard technical manager for the last 8 years or so and definitely feared losing my skills. Tools like Claude Code have breathed new life into my work again. To your point, I can now turn around features that would have taken weeks in just a day or two. My experience reviewing code and helping making decisions allows me to guide the agents toward the code I want written. It feels like a super power.
Does the author have experience writing code with AI? This kinda feels like an AI-generated hype thread like you see on Reddit that's not really connected to reality. Some good points (e.g. soft skills are more important now) but also slop heavy. For example:
"When Claude knows everything at the 80th percentile, being 95th percentile in one technology is less valuable than being 70th percentile across the board plus having the judgment to know what matters."
When Claude can write code where 80% is good and 20% is dangerously broken, you do not want someone someone reviewing it who is only 70% good but "trusts the process". You are going to destroy your codebase that way. You want a specialist who is actually 95% good, who can use Claude to speed them up while being able to immediately detect and jump in when it's in the "20% zone" and trying to ship broken or dangerous code.
Great article. at different stage of the company or the product, 80% is usually not enough, most of them requires 90%, and usually it's harder to achieve anything above 90% without re-architecting or going through serious validation cycles.
I've been a hands-off-the-keyboard technical manager for the last 8 years or so and definitely feared losing my skills. Tools like Claude Code have breathed new life into my work again. To your point, I can now turn around features that would have taken weeks in just a day or two. My experience reviewing code and helping making decisions allows me to guide the agents toward the code I want written. It feels like a super power.
Great article.
you need a “someone” or “something” in the last sentence. nonetheless this article is truer than truth.
You framed this perfectly, especially for technical folks who haven’t been “technical” in awhile. The what vs why vs how is so critical.
Does the author have experience writing code with AI? This kinda feels like an AI-generated hype thread like you see on Reddit that's not really connected to reality. Some good points (e.g. soft skills are more important now) but also slop heavy. For example:
"When Claude knows everything at the 80th percentile, being 95th percentile in one technology is less valuable than being 70th percentile across the board plus having the judgment to know what matters."
When Claude can write code where 80% is good and 20% is dangerously broken, you do not want someone someone reviewing it who is only 70% good but "trusts the process". You are going to destroy your codebase that way. You want a specialist who is actually 95% good, who can use Claude to speed them up while being able to immediately detect and jump in when it's in the "20% zone" and trying to ship broken or dangerous code.
Thank you. The comment was literally the only sensible thing in the long article :)
Great article. at different stage of the company or the product, 80% is usually not enough, most of them requires 90%, and usually it's harder to achieve anything above 90% without re-architecting or going through serious validation cycles.
Very insightful