A staff engineer's journey with Claude Code

Every one of these discussions boils down to the following:

- LLMs are not good at writing code on their own unless it's extremely simple or boilerplate

- LLMs can be good at helping you debug existing code

- LLMs can be good at brainstorming solutions to new problems

- The code that is written by LLMs always needs to be heavily monitored for correctness, style, and design, and then typically edited down, often to at least half its original size

- LLMs utility is high enough that it is now going to be a standard tool in the toolbox of every software engineer, but it is definitely not replacing anyone at current capability.

- New software engineers are going to suffer the most because they know how to edit the responses the least, but this was true when they wrote their own code with stack overflow.

- At senior level, sometimes using LLMs is going to save you a ton of time and sometimes it's going to waste your time. Net-net, it's probably positive, but there are definitely some horrible days where you spend too long going back and forth, when you should have just tried to solve the problem yourself.

If you have a sophisticated agent system that uses multiple forward and backward passes, the quality improves tremendously.

Based on my set up as of today, I’d imagine by sometime next year that will be normal and then the conversation will be very different; mostly around cost control. I wouldn’t be surprised if there is a break out popular agent control flow language by next year as well.

The net is that unsupervised AI engineering isn’t really cheaper better or faster than human engineering right now. Does that mean in two years it will be? Possibly.

There will be a lot of optimizations in the message traffic, token uses, foundational models, and also just the Moore’s law of the hardware and energy costs.

But really it’s the sophistication of the agent systems that control quality more than anything. Simply following waterfall (I know, right? Yuck… but it worked) increased code quality tremoundously.

I also gave it the SelfDocumentingCode pattern language that I wrote (on WikiWikiWeb) as a code review agent and quality improved tremendously again.