Or you can be like P.G. Wodehouse and natter on winningly about shapes in fog and how shapeless they are but THIS shape could beat out a prize pig for shapeless except perhaps for the part where it is obviously rather human.

Oops. Step 3.5, implicit is: commit. :)

1. make a new branch
2. check out the end target commit contents (git checkout end-state --)
3. revert a bunch of code, sometimes then making additional changes to make everything work
4. goto 2

This is easier than rebase because there are no merge conflicts.

No idea how an agent will handle it.

Hah. Keep us posted about how you get it working in the end! I love seeing worked examples, and I always have something to learn from you.

FWIW, the way I usually do work like this myself as a human is:

Ah. Right, well in that case, carry on.

Have you tried insisting that you ARE her?

That assumes the left must choose one or the other, as opposed to allowing both.

This looms larger in the US where (a) coalitions are formed before the election, not after it and (b) representation is significantly based on place, not just population.

TypeScript something something? It does do a fair amount of code searching, which is local.

FWIW, I’ve been getting consistently better results out of Codex CLI (with GPT5-high model) for a while. You might see if it helps.

Addendum: Zoe lives in Switzerland; Matthew in India.

Also the nickel is made of wood. And the dime is counterfeit.

Hi!! How long you here? Been ages, maybe we can meet up.

FWIW, it landed for me.

I already checked ECC

Helpful decryption failure errors

There’s an important comma missing in that sentence.

youtu.be/T-Wc8vluyp0?...

It’d be great if you quit Bluesky: you’re the only real reason I’m here, and I could use less social media in my life.

If by “got this” you mean “t.Skip”, yep, totally, 100%.

And answer: yes