Neat. Less radically, have you considered sketches in the margins, linked to characters? Or sketches that appear in boxes, with text wrapping around them? I ask cuz “sketch directly on top of text” doesn’t seem as useful as those. Aside from, say, editing marks. Maybe increasing line spacing would make sketches literally anchored to characters more useful?

(Or if you do think sketches directly on top of text could be useful or fun, I’d love to see some examples like that.)

It's a good question. This particular configuration came up in conversation, I think, as something that looks and feels like paper. I'm not sure either that it's useful as is.

As a heavy user of pencil-on-tablet annotations of scientific papers, my vote is for annotations attached to a place in the text but shown/drawn on the margins. Plus manual text markup: highlight/underline, strike through.

Hm was the "indicate I'm not going to type further into the old blank" interaction for switching direction intentional? It's kind of surprising that going the first direction the blank updates for every keystroke but it would require additional interactions to get it to recompute the other way..

There's basically 2 scenarios here:

• I'm clearing a field before typing something else in.
• I'm clearing a field so the computer can fill it in.

If I optimized for 2, then I end up with a situation where a blank has say 10, I click into it to change it to 20, but when I backspace to nothing it immediately goes back to 10.

So currently the update as you type only triggers for fields that are blank when cursor is outside them.

It does seem sub-optimal, so suggestions appreciated.

Ah I think I see what you're saying - they're kind of "linked" to each other and always want to stay in linked even during edit, which gets you that first interaction where celsius updates instantly, but you needed to add that additional interaction to go back the other way...

It may not be what you're going for, but in my mind I'd kind of expect more of a leader/follower model for these links - if I start editing field A, anything linked to it should be updated until I unfocus that field

but nothing should really trigger an update on A until I unfocus that field

That's certainly an approach. The drawback there is that you might change a complex sheet in unanticipated ways because you forgot about the presence of some formulas. Things change outside your field of view. Right now that only happens for fields you've not typed anything into. So no data loss. I tend to prefer UIs with a sense of stability that aren't constantly moving what I put into them..

Ahh right...yeah wasn't thinking about the more complex sheets with things out of view

I imagine I'm probably going to run into a similar problem pretty soon with these event-based agents haha

Do the prompts aim in the direction of some rough goal or domain, or are they fairly untargeted, getting the AIs to add anything (that compiles?) to the codebase?

For the daily run its pretty untargeted...just "here's what Sublayer Actions are used for, here's a repository of ones we've found useful in the past, come up with some new ideas for things we might find useful and generate them"

but once something is merged it becomes a bit more targeted: "we just added an action to do X on this service, what other things might people want to do on that service?"

for this example its kind of anticipatory code gen

I see. So the act of humans merging something is guiding the computer to push more in that direction?

This is cool. It reminds me of Doug Lenat's "Automated Mathematician" which would go off and search for interesting relationships between numbers and try to generalize from them. Once it defined some predicate or function over numbers, a human could go in and rename it (say if it was a well-known concept, like 'odd'). One of the heuristics it used was to pay more attention to definitions if a human went in and renamed them.

Yeah exactly! The main thing keeping it from growing more is that since these are external services I need to do a lot of work to verify them by hand, test API keys and stuff....working to come up with something with a little less manual review...

But that Automated Mathematician looks really interesting, I hadn't heard of that before! It seems like I've been pulling on similar threads to that for a while now though with programming concepts instead of mathematical ones...