Agent-led development asks you to have a business-owner mindset: What's the outcome we're trying to achieve? The less you're required the dig into the details, the easier it is to see the big picture.

I understand why that's a difficult transition for individual contributors who've perhaps never had to command large teams. There's a deep satisfaction in reviewing every line of code. I did that for decades! But also throttles your ambition.

When I first started using agents, I felt 'this is what it must feel like to run a business/company'

I definitely tried to resist for a while, but the big picture benefits do outweigh the benefit of knowing every line of code. When you see how quickly these things can rewrite an entire code base I just can't justify slowing it down just so I can keep up.

AI handles more of the implementation, but that makes product judgment even more valuable.

Exactly. Agent-led development shifts the mindset from “How do I build this?” to “What outcome am I responsible for?” The less we obsess over implementation details, the more clearly we can optimize for the actual result.

不抠细节的人定方向 细节谁兜 夹在中间那层人最容易先没了

🫡🫡 you are right dhhh

Depends what the agent is a transducer for. I keep coming back to that. If it's compressing the details, great. If it's dropping them, the big picture you're seeing is a lossy compression of reality. Someone still has to know where the signal degrades.

how do you keep the big picture when the agent starts overengineering?

outcome first until something smells. then i'm in the diffs on that one path

Are the people crying the same ones who insisted they write their own getters and setters?

Exactly. Agent-led development shifts the scarce skill from typing code to defining outcomes, constraints, and proof that the result works. The zoomed-out view matters—but only if the feedback loop stays tight.

but isn't digging into details important to have a good, finished product?

Details are important when they have bearing on the outcome. Think of yourself as Steve Jobs running the iPhone development. Endlessly sweating the details, but not writing any of the device drivers directly.

that analogy helped! so, if we sweat on making the best damn product, no matter who or what writes, that's what matters in the end

Half agree. Outcomes first, yes. But someone still has to read the diff. The details come back fast the day an agent quietly rewrites a migration and nobody looked.

so are you reading less of the code yourself these days?

Exactly. Agent-led development shifts the scarce skill from typing code to defining outcomes, constraints, and what “done” means. The better the judgment, the more useful the leverage.

To bestow self-reliance.

The less the understanding of the details the more pixelated the ‘big picture’?

>The less you're required the dig into the details, the easier it is to see the big picture. I feel like this is exactly false. Certainly it's not the lesson of how Tesla is run.

Owner mindset works only if the outcome is executable. With Claude Code, 'make checkout faster' drifts; 'p95 under 300ms, bin/bench passes, no schema changes' converges. The owner's job moves from reading diffs to writing the acceptance check.

I didn’t use to like this guy earlier. But I am starting to like him now, simply because he is speaking the truth or atleast providing a direction for how future engineers should prepare themselves

the less you dig into the details the more you're trusting a machine to get them right, which is a bold stance to take as a lifestyle

Do you actually stop digging into the details, or just tell yourself you should?

was thinking about the same for past few months. I am already thinking in terms of product engineering at a higher level to understand the product outcomes. need to learn and talk in terms of business outcomes as well with my agents to orchestrate them well

has this made you kill more ideas before anyone starts? that's been the bigger shift for me than delegating the code.

This is the reality now that most people do not want to accept.

code artisans will never be able to evolve, their entire personality is built around code... those who see the value in AI will succeed while those who are stuck ego tripping over who wrote the best function will ultimately fail

does this change what you look for when hiring engineers?

I think some people suffer from the fact that they can’t think big picture. Normally those are the loudest doomers.

When using agent-led development, try focusing on the outcome you want to achieve. Does digging less into the details make it easier for you to see the big picture?

Own the outcome, sure. But who calls it done?

The person who owns the outcome?

so -- no more OMARCHY ?

One line I really liked from your talk was; > Pure game theory, there's only one play here and it's total fucking optimism! AI doomers (or doomers of any sort really) will only bring each other down, there's no bottom level to their sadness - don't get dragged into it folks!

the outcome is the job