hari

Despite all criticisms this was still a free pcc and personally one of my best

@songjunkr Hey Jun sorry for asking again I'm an AI developer and noticed your massive contribution in the field,I have before asked you how you have been able to solve tool calling issues in models,as an AI developer it would really be helpful as less information is available online on it

Qwen3.6-35b 사용자분들에게 질문 모델에서 어떤점이 아쉬운가요? 고치고싶은 특별한 파트가 있나요?

@songjunkr @k3ntosan Hey Jun i wanted to ask you how exactly people are fixing tool calls exactly?I don't see my ch relevant info online and it's a problem i constantly face with local models

@k3ntosan 도구호출은 Mlx-vlm을 사용해야 합니다

SuperGemma4와 같은 로컬모델들은 세팅에 매우 민감합니다. SuperGemma-fast : Text only, 140+tok/s, 매우 빠른모델 SuperGemma-Multimodal : 비전 및 에이전트 툴콜링 작업을 위한 스마트 모델 MLX : 맥 애플실리콘 환경을 위한 모델 (LM Studio는 아직 지원을 하지 않아서 백엔드 수정이 필요합니다.) GGUF : 범용 압축 모델로, 대부분의 환경에서 구동 가능합니다 (맥에서는 MLX보다 50%정도 느립니다.) 가장 쉬운 세팅 방법은 Codex / Claude 에게 최적화된 방향으로 세팅하라고 지시하는것 입니다. 제가 사용하는 하네스는 @NousResearch 의 Hermes Agent입니다. 토큰을 적게 사용하기 때문에 더 빠릅니다.

@songjunkr @waefrebeorn How exactly does one fix tool calling if you don't mind me asking im more interested in the technical details and how generally constrained decoding is fixed in practice?

@waefrebeorn 네 전부 고쳤습니다

Supergemma4 31B 작업이 끝나갑니다. - 0/100 완벽 무검열 - 벤치마크 성능 18% 향상 - 툴콜/토크나이저 수정 - 속도 25% 향상 Gemma4 hackaton 참여작품으로 만들었으나, 그것은 오픈소스 정신에 맞지 않는것 같아서 오늘중으로 모두에게 공개할 예정입니다. 팔로우하고 조금만 기다려주세요.

@bnjmn_marie @marcaruel Btw were there any tests you did that checked tool calling capabilities of gemma?

@marcaruel @hari94809472 Probably not I checked other replies, seems ok. If it's a bot, it's a very good one 😅

Gemma 4 31B vs Qwen3.5 27B, Thinking Enabled I ran multiple benchmarks multiple times. Gemma 4 31B looks better and more stable (smaller accuracy variations between runs, which makes sense since it generates shorter sequences). I'll publish my full results and analysis on my blog later this week (link in profile).

@bnjmn_marie @marcaruel Im not a bit.i used to play this game I just follow ai a lot now I'm a ai developer as a professional

@bnjmn_marie Which versions are you testing and what is the actual model id you are using

@bnjmn_marie Fixed by Google or independent publishers?

@laman_gh @Bhnewman05 @Hesamation Im More interested in the structured output side of things.I don't see many implementations online where you can pair an agent with structured output generation other than through json.Internally they use context free grammar to implement them for json they still fail occasions

@hari94809472 @Bhnewman05 @Hesamation github.com/snrraptopack/a… This is a recent research ai agent I built with it... It uses python binding, you can use js if you want

this thing disappeared like smoke in a thunderstorm.

@laman_gh @Bhnewman05 @Hesamation Link?

@Bhnewman05 @Hesamation I have a language for creating ai agent it doesn't use json mode to get structured output yet it's reliable and allows multi conversation in single turns

@SarikaMaha86540 @ShivaKarka @mindmusclepro Lab tests, brand matters,periodic testing with appropriate measures to not bias the test etc, channels like trustified are doing this

@ShivaKarka @mindmusclepro Today's condition protein powders worth 2 to 3k are containing only 50% of described value of protein then how you are sure that 100 rs of paneer is pure

More popular these products get, the better. Vegetarian Indians desperately need low calorie high quality Protein sources.

@foodpharmer2 @ganeshsonawane Not sure where you are doing your market research but all the brands are less expensive than your product.Take the same tier product (Naturaltein,whole truth) all are cheaper after discount code have similar minimal ingredients and testing, only difference is your reach

Thanks @ganeshsonawane! The reason why our price is lower than other leading protein powders, even though we are high in quality is because we spend nearly 0 on marketing. We are growing entirely on word of mouth, which helps us keep our costs low. Posts like yours really helps us reach more people, so thank you for the continuous support!

@leerob I want to build a cursor replica from scratch, can you talk about AST trees and codebase indexing?Im trying to develop an LLM that can understand custom code so the indexing strategies aren't too obvious to me

What should I teach next? 👀 cursor.com/learn

Introducing Cursor Learn! My first course is a six-part video series on AI foundations. It's designed for beginners to learn concepts like tokens, context, and agents. If you prefer to read, there's quizzes + AI models to try out. You can watch it in ~1 hour, 100% free!

@AndrewHogben1 @alexandrelwt @LFC_PAUL_ @UTDTrey @Redfellas_ Your midfield is so crap relegation teams bypass it, defence and keeper shambolic as well,that attack needs support to create chances because their is no individual in your team to create chances on their own (think yamal/messi/neymar) until you fix this you will be like 10th

Manchester United stealing Sesko from Newcastle to stop Newcastle from selling Isak to Liverpool…

@snaggycfc @InvertTheWing This match is more important than that final fuck that trophy if we miss ucl

@InvertTheWing reece’s whole season has been about managing minutes so do you think that maresca will want to manage him more or save his minutes for the uecl final or nah?

Do you go for Enzo Fernandez or Romeo Lavia against Forest? For me, Lavia’s absence was really felt against United. I think there was a real disconnect between the defensive line and forward midfield. Against compact structures, you need a carrier (in my opinion) to enable this seamless transition between phases. I prefer Enzo against teams that like to play on the front foot a bit more. In an ideal world, they exist together — but I don’t think you can drop James after his performance.

@itscapcapalot @DusanVlahovic @maroonmatters I literally said they have different roles and men and women aren't equal!! Never mentioned men is above women or women above men!We live in a society where each gender are judged differently whether you like it or not

@hari94809472 @DusanVlahovic @maroonmatters ahhh i see. you believe women are beneath men. i can see why no one takes a liking to you

Faze Rug got it like that ?

@itscapcapalot @DusanVlahovic @maroonmatters You the type of guy to wife Bonnie blue and lily Phillips, Absolutely no shame you do realize that men and women aren't really equal and have different roles and benefits? women aren't scorned for not working while men are losers , whoreish behavior is penalized for women not men

@DusanVlahovic @maroonmatters weirdo behavior. when a guy sleeps with women he’s a player, when a woman sleeps with men she’s “damaged goods”

@Blue_Footy I don't think you get it,if forest lose/draw expect the nastiest low block from them against us out of spite as they are out of ucl,if they win they will have to come out and play against us and I think our midfield destroys theirs,it's not straightforward

Beat Nottingham Forest today and all your sins are forgiven, Potter.

@willnotclap @SongFinderBot what is the song

@willnotclap @Shazam what song is this