ugh, translation is worse than passing a parakeet transcript to a cloud llm. disappointing model.
testing it against parakeet right now for ASR, it's slower, heavier, and has a higher error rate. but i'm doing a pass with the embedded translation to see if that's any good, it might be worth it. e: also i'm testing…
2nding parakeet, had great results with a japanese-tuned variant, beat the pants off whisper-kotoba
To me the most interesting chart on artificial analysis is intelligence index vs cost per task. It's like the third chart down and they graph a pareto line showing where the frontier is and every model on the pareto…
I'm finding the idea of spawning then killing Jellyfin over and over again like Groundhog Day really funny
> I disagree with this. you disagree with freedom of association?
psst, they're talking about movies
I am also starting to take issue with "we've developed a new cutting edge model, and nobody can use it" announcements. Most recently with Google's Argon, at least they were using it internally and they'd be slowly…
Interesting. Chart is on this page: https://artificialanalysis.ai/evaluations/omniscience "AA-Omniscience Hallucination Rate (lower is better) measures how often the model answers incorrectly when it should have refused…
max thinking means benchmaxxing i guess
The gemma 4 models are pretty great
Yeah I've been stuck on 4.8 too, the opus 5 problems seem mostly fixed. It obeys and writes concisely rather than running off, doing its own thing, and emitting a word salad.
We just straight up posting Russian propaganda to HN now? https://www.cbc.ca/news/world/russia-tribunal-ukraine-1.6512...
Funny to see this after I spent the morning troubleshooting and fixing my crowdsec install on my debian 13 vps. Apparently they stopped supplying a community blocklist to my machine because I'm running the old debian…
This looks excellent, I've been wanting cheap IP KVMs for a while for my workbench at the office and for my NAS server at home - at this price I'd get the three pack and still have a spare.
Does it do RCS?
"we cannot rule out that de-identified data derived from their usage of our products helped improve our models ." What a landmine sentence to bury in this report, you can't rule out your models were spying on other…
> Gemma 4 is great, but really, Llama, in 2026? I'd assume the author is just getting confused because of ollama and llama.cpp and all the other ecosystem "llama" that are still in use. Llama really did kick off the…
> Oh. The connection is maybe more tenuous than Strike 3 claims. Yeah they seem to be following the time-honored legal strategy of throwing shit at the wall and seeing what sticks
> This acquisition and the Stripe acquisition of OpenRouter are puzzling. Both HF and OR, from a technological standpoint, don't appear to have much IP that cannot be easily replicated. I was puzzled by the comments on…
To think that humanity only needed to invent hard AI to make printers usable /s
I only have the $20 claude subscription. Last weekend I was doing a fairly heavy task (porting quake to raspberry pi native GLES 1.0) and ran out of claude usage twice, then hooked in with Luna in OpenCode and finished…
> There were some who believed that making the shareware disc easily-crackable was an intentional stroke of genius. I doubt it. I would believe that the id guys weren't too choked about the lost revenue after the fact,…
There are things to learn from the drone conflict in ukraine, eg drone netting, but I don't think neo dazzle camouflage is one of them https://www.businessinsider.com/ukraine-stunned-uk-military-...
> It's weirdly uncomfortable knowing that phrasing has such a big impact on one's emotions. It really shows how vulnerable we are to manipulation. My dad would discuss it in high school english courses: terrorist vs…
ugh, translation is worse than passing a parakeet transcript to a cloud llm. disappointing model.
testing it against parakeet right now for ASR, it's slower, heavier, and has a higher error rate. but i'm doing a pass with the embedded translation to see if that's any good, it might be worth it. e: also i'm testing…
2nding parakeet, had great results with a japanese-tuned variant, beat the pants off whisper-kotoba
To me the most interesting chart on artificial analysis is intelligence index vs cost per task. It's like the third chart down and they graph a pareto line showing where the frontier is and every model on the pareto…
I'm finding the idea of spawning then killing Jellyfin over and over again like Groundhog Day really funny
> I disagree with this. you disagree with freedom of association?
psst, they're talking about movies
I am also starting to take issue with "we've developed a new cutting edge model, and nobody can use it" announcements. Most recently with Google's Argon, at least they were using it internally and they'd be slowly…
Interesting. Chart is on this page: https://artificialanalysis.ai/evaluations/omniscience "AA-Omniscience Hallucination Rate (lower is better) measures how often the model answers incorrectly when it should have refused…
max thinking means benchmaxxing i guess
The gemma 4 models are pretty great
Yeah I've been stuck on 4.8 too, the opus 5 problems seem mostly fixed. It obeys and writes concisely rather than running off, doing its own thing, and emitting a word salad.
We just straight up posting Russian propaganda to HN now? https://www.cbc.ca/news/world/russia-tribunal-ukraine-1.6512...
Funny to see this after I spent the morning troubleshooting and fixing my crowdsec install on my debian 13 vps. Apparently they stopped supplying a community blocklist to my machine because I'm running the old debian…
This looks excellent, I've been wanting cheap IP KVMs for a while for my workbench at the office and for my NAS server at home - at this price I'd get the three pack and still have a spare.
Does it do RCS?
"we cannot rule out that de-identified data derived from their usage of our products helped improve our models ." What a landmine sentence to bury in this report, you can't rule out your models were spying on other…
> Gemma 4 is great, but really, Llama, in 2026? I'd assume the author is just getting confused because of ollama and llama.cpp and all the other ecosystem "llama" that are still in use. Llama really did kick off the…
> Oh. The connection is maybe more tenuous than Strike 3 claims. Yeah they seem to be following the time-honored legal strategy of throwing shit at the wall and seeing what sticks
> This acquisition and the Stripe acquisition of OpenRouter are puzzling. Both HF and OR, from a technological standpoint, don't appear to have much IP that cannot be easily replicated. I was puzzled by the comments on…
To think that humanity only needed to invent hard AI to make printers usable /s
I only have the $20 claude subscription. Last weekend I was doing a fairly heavy task (porting quake to raspberry pi native GLES 1.0) and ran out of claude usage twice, then hooked in with Luna in OpenCode and finished…
> There were some who believed that making the shareware disc easily-crackable was an intentional stroke of genius. I doubt it. I would believe that the id guys weren't too choked about the lost revenue after the fact,…
There are things to learn from the drone conflict in ukraine, eg drone netting, but I don't think neo dazzle camouflage is one of them https://www.businessinsider.com/ukraine-stunned-uk-military-...
> It's weirdly uncomfortable knowing that phrasing has such a big impact on one's emotions. It really shows how vulnerable we are to manipulation. My dad would discuss it in high school english courses: terrorist vs…