[peterbell_nyc]

I'm generally in agreement with everyone here. - Some code is ephemeral - it's generated to do the thing, thrown away end of session and the csv was imported successfully (or whatever). Make sure you have at least some testing of the output or you may find the email is in the last name field for some rows. If possible, have an API your agent uses with rich domain types and validations that force it to do things right or do them again (and that it' can't rewrite to relax the constraints!) - You can one or few shot a real app - for a few users, for a small set of use cases. Scope of this will improve with models, but at least today it's spelling bee app for my kids" not "salesforce replacement for millions of workers". - You can add rich validation steps for all types of quality that you care about which (assuming they converge) can deliver high performance, well designed and functionally correct code mostly autonomously.

I'm building an orchestrator (who isn't). Haven't looked at the code yet, but it appears to work. But man have I spent hours in loops between Claude, Codex and myself all on the highest thinking levels to figure out what interface portability means for the employee, how best to handle "remote" sessions and the appropriate semantics for pipelines/recipes.

I've also been very opinionated about who does what. I'll let the agent write a script to sync with github and reload workers, but I decided to "waste" the 5 minutes to manually do all of the config steps on render for my server when claude told me that I couldn't just give it read only scope to pull the logs. Bad news, I'm cutting and pasting for my computer overlord. Good news? Claude can't blow away the prod db if it happens to get in the way of whatever interpretation is makes of the instructions I give it.

A chainsaw requires very different skills that an axe. It has different failure modes. Some experience as a lumberjack probably helps using either/both.

No difference (at least now) with agents.

[tombert]

I have found that for low-stakes stuff, where "good enough" really is ok, Claude and Codex have been pretty great. I don't particularly care if the code is optimal, just enough to do that job.

For example, I had Claude generate a language server for TLA+ so I could have nice keystrokes in Neovim. For things like this, I really do think there is such thing as "good enough"; a language server doesn't have to be perfect, and the stakes are pretty low, where I think the worst case scenario is that it screws up my code, but that should be relatively easy to catch in Git.

I have been trying to mostly have Claude generate code from specifications; either a Mermaid diagram for simpler stuff, and TLA+ for more complicated stuff. I usually supply a lot of surrounding context about how I want these specs to be implemented, and it will usually get me about 90% of the way there, but I've found that I still need to hack against it to get over the hump.

It makes me feel a little valuable; I finally have an excuse to use formal methods for things.