Fer

💻AI NLP data scientist
🏎 F1 enjoyer

[Carlos Fenollosa @cfenollosa]

ChatControl 1.0 se ha aprobado pese al voto mayoritario en contra.

¿Cómo es eso posible? Gracias a un fraude democrático:

1) Los *partidarios* de ChatControl llamaron al voto para *rechazar* su medida, en vez de votar para aceptarla.

2) Activaron el procedimiento "de urgencia", que no está pensado para esto, pero fuerza a obtener mayoría absoluta en vez de simple para lograr la aprobación.

3) Varios eurodiputados estaban de vacaciones (un día antes de que empiecen oficialmente sus vacaciones, vaya tela...) y los partidarios de ChatControl sabían que era matemáticamente imposible *rechazar* su medida por mayoría absoluta

4) Por tanto, sabían que su medida no podría ser rechazada... y sería ACEPTADA

5) Esto, lógicamente, es un abuso del reglamento, una jugada absolutamente anti-democrática y debería ser anulada por el Tribunal de Justicia de la Unión Europea.

6) No sucederá. Tus proveedores de mensajería y email podrán escanear masivamente tus mensajes, incluso sin orden judicial, ni siquiera sospecha de delito.

Qué asco 🤮

[Fer @Fercasaban]

I saw the thief for 3s — the sketch I described was a stranger. 👁

EYEWITNESS: my @Gradio × @huggingface Build Small game. Cohere Transcribe hears you, MiniCPM5-1B draws the suspect, VoxCPM2 voices the culprit. Zero cloud APIs.

Beat my 9% 👇
https://t.co/rYxNnnnfNa

[@levelsio @levelsio]

🇪🇺 My fellow Europeans

You have to fight ChatControl 💪

Don't let @vonderleyen read your chats!

https://t.co/y5Plrj1p27

[sui ☄️ @birdabo]

fixed it.

[Antonio Padrón 🌶️ @AntonioPadron17]

mi padre: conduce con cuidado hijo

mi padre con mi edad adelantando a un Peugeot 206 que le había pitado:

[Sam Altman @sama]

gpt-oss is out!

we made an open model that performs at the level of o4-mini and runs on a high-end laptop (WTF!!)

(and a smaller one that runs on a phone).

super proud of the team; big triumph of technology.

[Yuchen Jin @Yuchenj_UW]

closed-source AI vs. open-source AI

which way western man?

[nixCraft 🐧 @nixcraft]

meanwhile in the uk … 😅

[Pirat_Nation 🔴 @Pirat_Nation]

StopKillingGames has surpassed 900,000 signatures.

[Accursed Farms @accursedfarms]

We have passed 800k! This is suspenseful for me, since I've seen the momentum crash once before and I'm sure people know how interest can surge then fall off quickly online. I don't see this as a sure thing, but if we keep this up, we could make it!
https://t.co/EpnNTDR85U

[Pirat_Nation 🔴 @Pirat_Nation]

The time is now, sign.

Don't let big corporations decide when to delist or make a game unplayable.

https://t.co/qkQAoefjZ1

[Andrej Karpathy @karpathy]

The race for LLM "cognitive core" - a few billion param model that maximally sacrifices encyclopedic knowledge for capability. It lives always-on and by default on every computer as the kernel of LLM personal computing.
Its features are slowly crystalizing:

- Natively multimodal text/vision/audio at both input and output.
- Matryoshka-style architecture allowing a dial of capability up and down at test time.
- Reasoning, also with a dial. (system 2)
- Aggressively tool-using.
- On-device finetuning LoRA slots for test-time training, personalization and customization.
- Delegates and double checks just the right parts with the oracles in the cloud if internet is available.

It doesn't know that William the Conqueror's reign ended in September 9 1087, but it vaguely recognizes the name and can look up the date. It can't recite the SHA-256 of empty string as e3b0c442..., but it can calculate it quickly should you really want it.

What LLM personal computing lacks in broad world knowledge and top tier problem-solving capability it will make up in super low interaction latency (especially as multimodal matures), direct / private access to data and state, offline continuity, sovereignty ("not your weights not your brain"). i.e. many of the same reasons we like, use and buy personal computers instead of having thin clients access a cloud via remote desktop or so.

[Anthropic @AnthropicAI]

New Anthropic Research: Project Vend.

We had Claude run a small shop in our office lunchroom. Here’s how it went.

[Andrej Karpathy @karpathy]

+1 for "context engineering" over "prompt engineering".

People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window with just the right information for the next step. Science because doing this right involves task descriptions and explanations, few shot examples, RAG, related (possibly multimodal) data, tools, state and history, compacting... Too little or of the wrong form and the LLM doesn't have the right context for optimal performance. Too much or too irrelevant and the LLM costs might go up and performance might come down. Doing this well is highly non-trivial. And art because of the guiding intuition around LLM psychology of people spirits.

On top of context engineering itself, an LLM app has to:
- break up problems just right into control flows
- pack the context windows just right
- dispatch calls to LLMs of the right kind and capability
- handle generation-verification UIUX flows
- a lot more - guardrails, security, evals, parallelism, prefetching, ...

So context engineering is just one small piece of an emerging thick layer of non-trivial software that coordinates individual LLM calls (and a lot more) into full LLM apps. The term "ChatGPT wrapper" is tired and really, really wrong.

[tender @tenderizzation]

when you rely on data augmentation because your model doesn’t have scale equivariance built-in

[Aleph @woke8yearold]

There is something so wonderful about the only hacker on the internet that can crack Denuvo being a mentally ill Russian woman who talks like Sephiroth and goes on long rants about how she hates Indians

[ₕₐₘₚₜₒₙ @hamptonism]

> be an electric engineering student
> team up w/ cracked classmates
> start quant trading
*we’re so cracked*
> founded a quant firm in his 30’s
> makes ¥100B trading with ai/ml
*we’re even more cracked with ai*
> buys thousands of Nvdia GPUs
> creates DeepSeek as a side project

[Andrea Bizzotto 🇺🇦 @biz84]

Lol the comments on the latest ⁦@fireship_dev⁩ video