Sounds like your experience from LLMs dates back to 2022.

How do you know this? How would you falsify this hypothesis?

You could train an LLM on a dataset that intentionally excludes certain data, and see if it is able to extrapolate outside of its dataset and come up with those excluded items independently.

For example, remove all code from your training data, and then see if the model can code regardless.

Expecting to go from having seen no code to being able to code within a context window is a pretty high bar. More useful would be asking for code that does something you're pretty confident no-one has done before.

(TBH, when people claim this, I do wish there would at least be occasionally an actual pointer to something that was copied. I understand it's probably not going to be doable in all cases, but without examples it sure feels like a weak statement. In my experience the recent models are good at doing things I'm pretty sure there is not a close reference for in the training data, though for most part I wouldn't classify them as particularly difficult tasks either)

Thanks for the elaboration. (Though I'm not sure dmarinus would agree with your criterion?)

> For example, remove all code from your training data, and then see if the model can code regardless.

Would you accept this weaker version: make up a new programming language (that's guaranteed not to be in the training set), and see if the model (which has trained on programming in existing languages) can cope?

I don't think it's very useful but there's already a checksum in the meta data of an S3 object: https://docs.aws.amazon.com/AmazonS3/latest/API/API_Object.h...

It might be better to keep an index of paths to checksums and use the checksum as object key in S3.

I'm not sure what you like to achieve but if you use restic with S3 as a backend you can achieve much better deduplicate and compression.

I don't agree that you don't have to know CSS/HTML when you use a frontend framework.

I guess some frontend frameworks can abstract it away but most don't and you almost certainly will run into the limitations of those frameworks and then you still need to understand HTML/CSS

You can load a model from this page but it's only for testing purposes on this page and it won't be used for the Prompt API.

I think this policy will disable the automatic download of the model:

https://chromeenterprise.google/policies/#GenAILocalFoundati...

The prompt API can be tested here: https://chrome.dev/web-ai-demos/prompt-api-playground/

It would be really helpful if there was a way to download the model to a central location, so multiple users on a single system could easily share it.

Wooting has done this already for many years: https://github.com/WootingKb/wooting-design

did WASD do it when they folded?

I wish, I will need new feet for my CODE soon. Plus my cases latching tabs are about half gone.

if it helps, I found Epomaker like the Galaxy70 were a fine replacement, with some specialty switches like the TTC Bluish White V2 (silent tactile).

great, one more

The first spreadsheets I remember were TUI (pccalc, Lotus 123)

oh nice! I hate it that you can't disable mouse acceleration in macOS.

I tried posting ais-catcher.org but it got ignored

Doman needs to be www.ais-catcher.org

if possible you can also just create directories if they don't exist (ie. mkdir -p) and just exclude it in your root .gitignore (ie. ignore all build directories). That would safe you from creating multiple .gitignore files.