antiobli

I like it because it's a terminal first workspace and has bindings to use it programatically. Tried Superset and Intent, but they seemed like way overkill / unintuitive to use agents like I prefer. I use WezTerm as a terminal, so being able to have herdr exist in that application (not having another app to keep track of) as a tab really helps me stay in flow. Over the last few weeks, Herdr has given me the vibe of "works out of the box, but is highly customizable", and I think that power-user angle really sold me.

Oh, and I didn't know how much I'd love session persistence. I already use tmux for every other terminal process, so it being built in to herdr and a focused feature (server start, detach, stop) is nice to see.

Great tutorial, appreciate the approachable language. A lot of writers bathe their articles deep in Rust lingo that only Rust devs can really comprehend.

Their lines "A meta-harness is the loop that improves the harness automatically" and "the bottleneck is diagnostic context: most optimizers compress prior runs into summary statistics, while meta-harness gives the proposer up to 10M tokens of raw execution traces to grep through," seem good, no?

Have to dig into the code, but it looks like they have sound engineering around a "self-improving" agentic coding harness. Will be fun to take the code for a spin.

Working on something similar but in reverse (Python to JavaScript). A few tips:

- If the scripts are primarily by the same author, it's likely they copy-pasted a lot of their functions throughout their scripts. Do something like a "Find All" in the repository holding the scripts for the function name. Regex will help a lot in this, and help you see if a function ever has variations on number of arguments, slight name changes or misspells, etc.

- Refine an AI prompt as you convert scripts over and over.

- AI (ChatGPT, Copilot, etc.) is going to be the best automation you can get for this, because often times conversions don't match up one-to-one in a language. Especially if your scripts use npm, you might not find one-to-one matches in PyPI. AI refactors will also force you to examine the conversions.

- Understand what tech debt could be introduced in a one-to-one conversion. There may be some language workarounds that make a lot more sense to introduce over exact conversions of workaround functions for the language. I've had to work with several legacy Python scripts that called functions that can be better written in your target language. For example, one of our functions was conditionally_get_keys(level1, level2, level3) to kind of recursively get a nested value. This didn't need to be a function when I rewrote it in JS, rather I just wrote the variable as something like `city = user.location?.city`. No need to one-to-one convert a function that can be better written with a JS language feature. You'll probably encounter this when you can more succinctly write list manipulations with slicing (e.g. `steppedList = fooList[::2]` instead of converting a function necessary in JS to do the same thing).

Sorry if you have a time-crunch with converting things, but I would recommend a more hands-on conversion strategy, leveraging AI to do the basic conversion and then manually testing, debugging the solutions.