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?

> it might lower the need for devs to actually fix the code and they might start just relying on Fil-C.

Well, the program would still halt upon memory flaw, so there would still be a need to fix it

I suppose /some/ performance loss is inevitable. But this could be quite a game changer. As more folks play with it, performing benchmarks, etc -- it should reveal which C idioms incur the most/least performance hits under Fil-C. So with some targetted patching of C code, we may end up with a rather modest price for the memory safety

It seems they are using zstd now for .conda packages, eg, bzip is obsoleted, so that should be faster.

It would be nice indeed if there was a good solution to multi-gigabyte conda directories. Conda has been reproducible in my experience with pinned dependencies in the environment YAML... slow to build, sure, but reproducible.

> because new poetry releases would stall trying to resolve dependencies.

> uv is much faster than both of these tools

conda is also (in)famous for being slow at this, although the new mamba solver is much faster. What does uv do in order to resolve dependencies much faster?

> uv will magically install and use the specified modules.

As long as you have internet access, and whatever repository it's drawing from is online, and you may get different version of python each time, ...

Sounds like a great use for a local LLM to strip ads from the output of the ad-infested LLM.

Well, heat capacity and thermal conductivity are not the same thing

ecause it's a much better experience than copy-pasting into a webapp

I'd like to try it, but enterprise only?

Surely there must be a way to do the joins in software, without doing it by hand, eg a SQL-like library? Pandas or equivalent?

Indeed. KYC has a purpose though -- prevention of fraud, money laundering, etc. Getting rid of KYC without a similarly-effective solution for those things seems unlikely. Ideas?

Things like this juice the anti-landlord sentiment: https://www.propublica.org/article/justice-department-sues-l...

The Gell-Mann amnesia effect applies to LLMs as well!

https://en.m.wikipedia.org/wiki/Gell-Mann_amnesia_effect

> Having a whole other device constantly running just to wake up my main device feels like a waste.

Indeed, my first thought is the best place to run this is on an OpenWRT router. Perhaps as a package, or a builtin feature?