They are deploying optimizations weekly (if not daily) with various AB tests. They don't manipulate model performance, but they do actively perform tests.

Steam still does this! For your local network only and not using the torrent protocol.

I've always taken the example that at any given moment openai and anthropic have models that are "far better and exceed any expectations", yet their coding tools suck ass and are outperformed by random people on the internet.

  items.filter(|(i)| cond(i)).map(Pointer::idx).collect()

you can probably get away with this if you implement a Trait, not sure how but I know for a fact this is possible, idk why there's an enumerate there when you aren't even using it.

items is already an iteratible so you can do direct .filter on it as well

tl;dr if the code looks ugly you're probably not taking advantage of a language feature that allows it to look pretty.

This is in fact more readable, thank you.

I still don’t know that I find it more intuitive or readable than the simple loop, but it’s much less awful than before.

I am sure there's a way to filter directly on cond variable and collect isn't needed most of the time since iterators are way more useful in general (unless you want to print that data)

   items.filter(cond).map(Pointer::idx)

I know this is possible, but you would have to consult some rust wizard for this.

This has existed for a long time and has been abused by quite a few people. I've seen some cc nodes using a random known cloudflare site and spoofing hostname to a temporary cloudflare site. IMO this should require a login at bare minimum.

if you think people won't abuse it because you're making them log in with a free email address...

This requires zero effort and you can deploy hundreds to thousands while maintaining 1000 gmail, proton etc takes resources and money.

They already do this. the email accounts already exist

It's the same reason that exists for PoW captcha, it's just way more annoying. Also accounts offer traceability versus a new identity every time you want to do something unless you want to go through hundreds of accounts a day.

There are n services and n tasks, the chances of overlap are minimal. In most cases people take a much needed break during downtime.

Actually I think it's better if everything fails at once and everyone can take the day off (thanks, AWS!). Having your CI fail one day and your package repository the next might well cost you two days of productivity.

you can run 10 ds-f agents at the same time for the same cost exploring a much wider surface area.

They never show exactly how they use it? Only a bunch of animations of it 'working'. Would like to see the actual code used for the demos!

It's a bit hastily put together, but I made a dspy fork where you can add a decorator to automatically use TypeSafe where possible on Signatures. It shows a fair bit of what actual, hands on usage looks like.

https://github.com/typesafeainate/dspy-typesafeify

DSPy seems like the right comparison and this is the first comment I've seen mentioning it.

Thanks for putting this together. I'm surprised the cost saving is so little though. I expected much more based on the post.

The doom demo shows the program state / query.

What people don't understand that programming is very much an art. You iteratively work on it ripping parts out, rewriting and rewriting and rewriting, while also rewriting and then rewriting every time a new feature, bug fix or scaling changes are needed.

Personally no amount of social outlook would want me want children. I simply don't want to have them I don't like raising or interacting with children it's just not in my nature and I really feel like a lot of that has to do with how people are nurtured.