This is very true and an underrated comment.

> Any decently skilled human blows them out of the water

No, by far no. I’m by all accounts “decently skilled human”, at least if we go by our org, and it blows anyone out of the water with some slight guidance.

And the most important part: it doesn’t get tired, it doesn’t have any mood swings, its performance isn’t affected by poor sleep, party yesterday or their SO having a bad day.

The thing is, LLM's produce better quality one-shots than any of the products that get returned from overseas ultra-budget contractors in India or SEA. I don't know what that means for Western devs, but I can tell you that the fortune 500 I work for is dialing back on contracting and outsourcing because domestic teams can do higher-quality work faster.

I have 20 years of experience and I don't handwrite any code anymore. Opus does everything, and it only needs a bit of steering occasionally. If you can give it guardrails (ie a pre-existing design system) and ways to verify its output (ie enforce TDD and use Chrome to visually verify) then it gets it right basically every time.

Turns out there are whole categories of software where 'extremely fast and good enough' is what matters, even for skilled software developers.

With the models I've been working with lately, providing them with small, actionable units of work that can easily fit within their context window (before compaction) seems to work well. If you can hit that sweet spot, you can get excellent output.

I don't tell the agents to "just go do it", as that tends to go off the rails for complex tasks. Emulating real world software development processes in meat space with your AI "team" seems to approximate similar outcomes.

I usually start by having the agents construct a plan document which I iterate on and build up well before writing code. This is a living document, not a final design (yet.) If I run into context window issues I just shut them down and rebuild from the document. I farm out research and data gathering tasks to build it up. Once all the findings are in I have the architect take a stab at the technical system design before the break down and delegation work begins. By then the units of work are small and manageable.

Am I an untrained human if I believe that Claude Opus 4.6 produces generally better code than I do in most circumstances?

Even with years as a principal engineer at a company with high coding standards and engineering processes?

I’ve been a full stack developer for 10+ years now and I completely disagree.

Modern models like Opus / Gemini 3 are great coding companions; they are perfectly capable of building clean code given the right context and prompt.

At the end of the day it’s the same rule of garbage in -> garbage out, if you don’t have the right context / skills / guidance you can easily end up with bad code as you could with good code.