kardos

Instructions are not guardrails. Best case, that just makes it less likely to do it

Possible idea: tag them with AI, and leave them aside for 6 months and review them with 2027 AI

The only winning move is not to play

Because they suspect that, sometimes, different content will be served by HTML vs alternate APIs

French text is somewhere around 10-30% longer than the corresponding English text. I would guess much of what you save on smaller vocabulary is lost on the lengthened text.

A solution to this is apptainer: you configure it to not see any of the host files by default, and mount the repo you want to work on at runtime.

A mitm does not redirect your ssh login to their machine

This could unlock a new chapter of Venus exploration!

On the other hand, any source code leak could be catastrophic

Well it won't be 'non-mainstream' for one

So is the SIMD the magic piece here, or is it the interpolation search? If the data is evenly distributed, that is pretty optimal for the interpolation search..

> Once the ads are injected directly into the main response is when things get interesting.

This would be where you post-process the LLM response with a second LLM to remove the ad..

If sharing all of your code with the closed providers is OK then it works. If that is a blocker, open weights becomes much more compelling...

There is probably some low hanging fruit to be harvested in terms of memory optimizations, and it could be a selling point for the next while as the memory shortage persists

Does this get tripped up on zip bombs?