Siddhesh Mhatre

[Chelsea Youth @chelseayouth]

Maresca is an anagram of scam era

[Qwen @Alibaba_Qwen]

We promised open weights for Qwen3.8. Now, time to meet them! 🎉

⚡ Qwen3.8-27B:
- A native multimodal dense model. With just 27B parameters, it outperforms Qwen3.7-Plus overall and shines in real-world coding & office workflows.
- 262K native context, easily extendable to 1M tokens via YaRN.
- Built for builders. Highly efficient, high-quality, and licensed under Apache 2.0.

🚀 The open weights for Qwen3.8-2.4T-A95B (Max-level) have also been released recently.

Whether you're shipping lightweight applications with Qwen3.8-27B locally or building agents with Qwen3.8-2.4T-A95B, they're yours now!

Download, deploy, and build something we haven't imagined yet. 👀👇
- Hugging Face：
https://t.co/4kaAcqYEVj
- ModelScope：
https://t.co/eRIMZCGkhC

[Chelsea Youth @chelseayouth]

Champions League player and Champions League person goes to Champions League club. His level. His time. Go well Trev.

[Siddhesh Mhatre @Sid1793]

Good luck @TrevohChalobah

[Chelsea FC @ChelseaFC]

What dreams are made of! 👌

[Siddhesh Mhatre @Sid1793]

@HermesAgentTips Qwen 3.6 27B locally on a 3090 with Sol as backup

[Siddhesh Mhatre @Sid1793]

@callmeoscar_1 You're an idiot

[Siddhesh Mhatre @Sid1793]

@NousResearch @blocks @jack official buzz integration with hermes

[Nous Research @NousResearch]

Hermes Agent now runs Buzz.

The self-hostable workspace from @blocks puts humans and agents in the same messaging channels and codebase.

Three ways to use Buzz with Hermes (and vice versa):
- Buzz Desktop auto-discovers your Hermes install runs it locally
- A relay bridge gives it a hosted identity in your channels
- Connect via the Hermes Gateway to use Buzz as a full external platform with channels, DMs, threads, reactions, and cron delivery

https://t.co/srljIbERN7

[Chelsea Youth @chelseayouth]

AND STILL

[Garion @KAkinmboni]

@SebC__ People going crazy cause Joao Pedro scored 15 goals in the prem. Abraham did it in 19/20 with significantly less minutes and haters glossed over it like it was nothing

[Siddhesh Mhatre @Sid1793]

@sudoingX Have you tried this model with MTP?

[TobyWrites @tobyasky]

More than a game.

[Chelsea Youth @chelseayouth]

Cobham past and present called up by England today...

Guehi, Hall, Livramento, Tomori, Rice, Solanke, Acheampong, George, Boniface, Murray-Campbell, Ajala, Fearon, Morgan, Derry, Mheuka, Ngumoha, Bernal, Emenalo, Walsh, Ezenwata, McAidoo, DIakite, Nicoll-Jazuli, Kavuma-McQueen.

[Andrej Karpathy @karpathy]

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then:

- the human iterates on the prompt (.md)
- the AI agent iterates on the training code (.py)

The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc.

https://t.co/YCvOwwjOzF
Part code, part sci-fi, and a pinch of psychosis :)

[Pratyush Kumar @pratykumar]

📢 Open-sourcing the Sarvam 30B and 105B models! Trained from scratch with all data, model research and inference optimisation done in-house, these models punch above their weight in most global benchmarks plus excel in Indian languages.

Get the weights at Hugging Face and AIKosh. Thanks to the good folks at SGLang for day 0 support, vLLM support coming soon. Links, benchmark scores, examples, and more in our blog - https://t.co/DcCG3zlN8p

[Galatasaray FR 🇫🇷 @Galatasaray_FR_]

💔 Victor Osimhen se confie sur l’un des moments les plus douloureux de sa vie :

« Quand j’ai signé à Lille, l’état de santé de mon père s’est gravement détérioré. J’étais constamment loin de lui. Puis, au début du COVID, il a été hospitalisé. J’étais seul en France. Le football était à l’arrêt, les aéroports fermés.

J’appelais sans cesse mon agent pour organiser un vol privé vers le Nigeria. J’avais même obtenu l’autorisation d’atterrir. Il ne manquait plus que l’accord du club et de mon agent pour me laisser partir.

J’attendais… encore et encore.
Et son état empirait.

J’ai commencé à paniquer. J’appelais chaque heure, je suppliais. Mais c’est à ce moment-là que j’ai compris le côté sombre du football, le business. On voulait me vendre. Un transfert était en discussion. Mon ancien agent me répétait : “C’est compliqué. Attends.”

Je devenais fou. Je ne dormais plus. Un matin, en sortant de la douche, j’ai regardé une photo de ma mère posée près de mon lit. J’ai senti que quelque chose n’allait pas. J’ai commencé à pleurer.

En bas, j’avais 20 appels manqués de ma famille. J’ai rappelé mon frère en FaceTime. Il m’a dit : “On l’a perdu.” Puis il a tourné la caméra vers mon père. “Tu dois lui dire au revoir…”

J’ai jeté mon téléphone. J’ai perdu le contrôle. J’ai tout cassé dans la maison. Mes voisins sont venus me voir. En France, ils étaient comme une famille pour moi. L’un d’eux est resté 6 ou 7 heures avec moi ce jour-là… Il m’a probablement empêché de faire une énorme erreur.

Ce qui m’a le plus détruit, c’est la culpabilité. Tous les enfants et petits-enfants de mon père étaient à ses côtés.

Le seul qui n’y était pas, c’était moi…

J’étais en colère. J’ai explosé. Je me suis dit : “Si le football, c’est ça… à quoi bon ? Je veux juste être avec ma famille.”

J’ai appelé mon ancien agent pour demander si je pouvais aller enterrer mon père. Il m’a répondu : “Vas-y. Mais reviens vendredi.”

Vendredi ? Le football pouvait bien aller en enfer.

Quand je suis rentré chez moi, j’ai vraiment pensé que je ne rejouerais peut-être jamais au football. J’étais écœuré de tout.

Les gens disent souvent : “J’ai tout vu.” Mais ils ne comprennent pas ce que ça signifie vraiment.

Je ne parle pas seulement d’avoir grandi à côté d’une décharge.
Je ne parle pas seulement d’avoir grandi sans électricité.

Je parle de la perte. De la vraie perte.
Et ça… c’est quelque chose de très profond. »

Un témoignage fort qui rappelle que derrière les transferts, les chiffres et les performances, il y a des réalités humaines bien plus profondes.

Respect. 🙏

[Callum @CallumCFC_]

Paul Green built a team that won 6 straight titles and an invincible treble. Neil Bath and Jim Fraser led an academy that produced a core of homegrown CL winners. The morons who signed Joao Felix twice think they know more about football than these guys…

[Wenting Zhao @wzhao_nlp]

🌶️ Some (perhaps) spicy thoughts. It’s been a while since my last tweet, but I wanted to write about how disorienting it has been from academia to an LLM lab 😅

The kind of research I was trained to do during my PhD almost doesn’t exist here. The obsession with mathematical elegance and novelty is mostly gone. Everything is about scaling data and compute. For a while, that really got to me. At my lowest point, I felt like I’d lost interest in building LLMs altogether. I didn’t feel intellectually challenged anymore.

What made this even stranger was that, at a technical level, things worked. If there was a capability I wanted to teach a model, scaling the right data and compute always got me there, no exception (so far).

But recently, I found a way to reconcile with myself..

I realized the real competition isn’t in the ML recipe anymore. Most teams do roughly the same thing. What actually matters is how fast you can iterate, test ideas, and recover from mistakes. And that speed is mostly backed by infrastructure 🏗️ Faster loops, fewer bugs, better tooling.

Seeing this made me excited again! Infra is its own deep, hard, and intellectually fun problem space.

In 2026, I want to become an ML researcher who’s really good at infra. And I'll come back to ML problems with that edge, and will be excited to share what I find 😌

[Siddhesh Mhatre @Sid1793]

@SebC__ Does it need Pep to win another one

[Sarvam @SarvamAI]

The Prime Minister hosted a closed-door roundtable at 7 Lok Kalyan Marg with a small group of Indian AI builders, including our co-founders @pratykumar and @vivekrag.

A key theme of the discussion was the need for India to present a unique AI model to the world that reflects the spirit of Made in India, Made for the World.

We're humbled to contribute to this conversation at a moment when AI is moving from research to national infrastructure, and when the choices made now begin to matter for everyone.