baq

27 ms·

841.
by baq 8mo ago
It’s the same deal
842.
by baq 8mo ago
sama did just that a couple years ago
843.
by baq 8mo ago
Foundational model provider manifesto: ‘While there’s value in safety, we value the Pentagon’s dollars more’
844.
by baq 8mo ago
I can’t believe I’m actually writing this: parent is an underrated comment.
845.
by baq 8mo ago
Consider the possibility that “bomb bomb bomb” was the entire and only point of the exercise.
846.
by baq 8mo ago
It should be reminding you of something which happened a few decades earlier and was much, much worse than Serbia.
847.
by baq 8mo ago
> Discord just burned years of goodwill and trust. ...not here, they never had any. it is good tech, but so is the w80 nuclear warhead, the tiger iv (for its time) and the j-35.
848.
by baq 8mo ago
this design at 7 transistors per weight is 99.9% burnt in the silicon forever.
849.
by baq 8mo ago
the radiator wouldn't be though
850.
by baq 8mo ago
Goalposts wooshing by at maglev speed. Of course it needs human supervision, see IBM 1979. Oversight however doesn’t mean the robots wait for approvals doing r&d and that’s where the magic is - the magic being robots overseeing their tr
851.
by baq 8mo ago
The magic of test time inference is the harness can improve even if the model is static. Every task outcome informs the harness.
852.
by baq 8mo ago
I like ‘claw’ because the s in it stands for security
853.
by baq 8mo ago
Absolutely the wrong take. You can teach CS with just pencil and paper, but that doesn’t advance the technology, it might only benefit academia in a narrow sense. CS students should be actively engineering software in addition to doing scie
854.
by baq 8mo ago
one step closer to being able to purchase a box of llms on aliexpress, though 1.7ktok/s would be quite enough
855.
by baq 8mo ago
defer is a stack, scope local and allows cleanup code to be optically close to initialization code.
856.
by baq 8mo ago
Harder: the bike must work Hardest: the pelican must work
857.
by baq 8mo ago
and if you write directly in assembly you don't even need a C++ compiler
858.
by baq 8mo ago
Ralph Wiggum would like a word
859.
by baq 8mo ago
Literally a skill issue.
860.
by baq 8mo ago
It’s been obvious from the start that the $200 point is the free tier
861.
by baq 8mo ago
It’s kinda obvious he’s a well spoken shark. Personally not an issue for me, you have to be at the top of a unicorn, but it isn’t something people in general like.
862.
by baq 8mo ago
Don’t be mad at it, be happy you were able to throw some of that sweet free vc money at your hobbies instead of paying the market rate.
863.
by baq 8mo ago
Home users being on multiple tailnets is serious power user territory
864.
by baq 8mo ago
Turns out people see value in imperfect experiences.
865.
by baq 8mo ago
I wonder if this was found with LLM assistance, if yes, with which one and is it a one-off or does it mark a start of a new era (I assume it does).
866.
by baq 8mo ago
Back when I was in a big org and in meetings 5-7h basically every day (as an engineer IC) this workflow would have absolutely hit it out of the park.
867.
by baq 9mo ago
> The second part is going to be the hard part for complex software and systems. Not going to. Is. Actually, always has been; it isn’t that coding solutions wasn’t hard before, but verification and validation cannot be made arbitrarily c
868.
by baq 9mo ago
China will stop releasing open weights models as soon as they get within striking range; c.f. seedance 2.0.
869.
by baq 9mo ago
Sonnet 4.5 was not worth using at all for coding for a few months now, so not sure what we're comparing here. If Sonnet 4.6 is anywhere near the performance they claim, it's actually a viable alternative.
870.
by baq 9mo ago
I agree completely, except looking at my 2fa app I'd need 20 physical tokens, so we actually need a super-duper-yubikey

More ›