ThouYS

whohoo, fable will (have to) stay

same, precisely the reason I haven't signed up yet. GLM can be used without any account fwiw

I am still on a very old version of logseq. Why update?

Why did this get the coveted "5"? I want an Opus that can compete with GPT 5.5

why are people using these products exactly?

signing away their rights to their photos? making psychopaths filthy rich?

if the surveillance glasses are coming, these people will also have signed away the commons, which are not theirs to give away

This is the correct take

3.6 is an absolute beast! makes you wonder why the big heavy models are even needed?!

Qwen is all you need

ironically, coding is the area that small models excel most at. qwen 3.6 27b is an insane agentic model. but Opus can be much more than a programmer: language tutor, knowledgeable "friend", therapist

I feel like the car usecase demonstrates that these models are not really useful for the cutting edge: They produce exactly the kind of in-domain data that already exists in droves. What is needed, and what tesla collects, are the edge cases!

(Now for a startup with zero data, this is of course still useful)

Just to clarify: I mean VRAM, RAM and runtime performance, not the numerical outputs (even though those also vary to some degree, haha)

I feel like there is no portable advice for performance. A torch model exported as onnx is a different model.

That onnx model run using onnxruntime with cuda ep is a different model than the one run with TRT ep.

And even among the same runtime, depending on the target hardware and the memory available during tuning, the model behaves differently. It is a humongous mess

don't know about this guy, but qwen3.6:27b with the UD 4bit quant and little-coder/pi has been amazing. the first local LLM experience that can do actual meaningful work

Maybe most interesting about the piece is, that we'll likely see more large scale interviews like this (even if this one is a bit bland)

what!

very nice. has a claude touch to it

this is.. a nothing burger? they don't exclude working for autonomous weapons, nor do they exclude mass surveillance. so what gives?

the headlines these days

for me gpt-oss:20b was that. glm 4.7 flash was not better, but much slower on a 16GB card

wow, everything is exactly unfolding as some AI doomers have projected

Anki is and was truly a blessing. Not sure I would have gotten through my studies without it. Thank you dae!

it gives you flexibility, options

lost 4% battery while reading the blog post

I think systems like this will turn drones (or at least, drone swarms) into nothingburgers. We're just one layer deeper into rock paper scissors now

I got deep into the stormlight archive by brandon sanderson. great escapism!

it's quite the bummer. some models you simply can't export with dynamo. for the time being the jit exporter is the only good option.

in particular selective function scripting is essential!

P is stored in the computer

what on earth is a DX?

the beauty of pure functions :)

100% my beef with it. same style as c++ where you never know where something comes from, when clangd starts throwing one of its fits