eˣrnesto

Co-Founder and CEO at Libre AI | PhD in CS/ML | AI and Machine Learning for All! 🇸🇻

[Nassim Nicholas Taleb @nntaleb]

A Lebanese Christian girl killed by Israel, just out of bloodthirst. Vicious criminals.

[moneyordebt ∞/21M @moneyordebt]

What’s going on is your sensationalism by plotting on a linear axis and clipping into a small range. And typing in all caps.

Try plotting on a log price - log age scale.

This is a power law network asset in the long term.

Log-log plot regularly updated at https://t.co/nWmFA2thic

@Giovann35084111

People who panic at this should convert that emotion into a desire to buy at favorable prices. Stoicism.

Or leave Bitcoin to those with the emotional maturity to deal with the volatility.

Instead of “what is going on?” — “What, rationally, should I do here?”

[Unsloth AI @UnslothAI]

Local models are coming to your laptop soon! 🚀

We're excited to partner with @Microsoft to enable millions of developers run local models on Windows!

[Google DeepMind @GoogleDeepMind]

We believe AI can be a dedicated research partner to help discover the next breakthrough.

Enter Co-Scientist: our latest Gemini-based multi-agent system that can generate, debate and evolve novel hypotheses for complex scientific problems 🧵

[Nassim Nicholas Taleb @nntaleb]

This is so true. Practically all the assholes I know support Israel. And all the compassionate, justice minded people I know have empathy for the Palestinians and are bothered by their suffering.

[Ada Fang @AdaFang_]

AI Scientists are starting to actually do science. Not just answer questions. Not just run workflows.

Introducing AutoScientists: a decentralized team of AI agents that can generate hypotheses, design experiments, write code, test ideas, analyze failures, and revise strategy as evidence accumulates.

Because real research is not a to do list of tasks.

It is a living search process. Leads emerge, failures matter, teams form around what works, and priorities shift when evidence changes. Much like how a lab of scientists would work on cutting edge research together.

Across GPT training optimization, biomedical ML, and protein fitness prediction, this decentralized structure consistently does better research.

Learn more 👇

@GaoShanghua @marinkazitnik @KempnerInst @HarvardDBMI @Harvard

[Sakana AI @SakanaAILabs]

ニューラルネットワークをブロックごとに学習する枠組みを開発

ブログ: https://t.co/c9AvsRK0lL

ニューラルネットワークの学習は通常、ネットワーク全体を一度に扱う必要があり、深いモデルほど多くのメモリを必要とします。このメモリ消費は、近年のAIモデルの大規模化を支える上で大きな制約と なってきました。

Sakana AIは、この制約を緩和する学習フレームワーク「DiffusionBlocks」を提案します（#ICLR2026 採択）。ネットワークをブロックに分割し、それぞれを独立に学習できるようにすることで、学習時に必要なメモリを1ブロック分にまで削減することができます。

中心となる発想は、各ブロックに「ひとつ前のブロックよりも、表現を少しだけターゲットに近づける」という明示的な役割を与えることです。この役割は、近年大きな成功を収めている拡散モデルが時間方向に段階的に行っている処理に対応しており、この対応関係を踏まえることで、各ブロックを原理的な目的関数のもとで独立に学習することができるようになります。

画像分類、画像生成、テキスト生成にまたがる5つのアーキテクチャ（ViT、DiT、Masked Diffusion、AR Transformer、Recurrent-depth Transformer）で検証を行い、いずれにおいてもエンドツーエンド学習に匹敵する性能を確認しました。

この枠組みは、同じ層を繰り返し適用するRecurrent-depth Transformerにも自然に拡張でき、通常必要とされるbackpropagation through timeを経ずに、1回のフォワードパスで効率的に学習できることも示しています。

大規模AIモデルの学習を、より少ない計算資源でも進められるようにすること——これはSakana AIが継続して取り組んでいるテーマのひとつです。本研究が、その一歩となることを期待しています。

論文: https://t.co/CRj96VGr0P

本研究は、東京大学の小山雅典氏との共同で行われました。

[Kaggle @kaggle]

Claude Opus 4.8 is now on Kaggle Benchmarks! 🤖

@AnthropicAI has just launched its latest version of the model, and we’ve brought it directly to Kaggle Benchmarks for you to evaluate and test.

[hardmaru @hardmaru]

For over a decade, we’ve accepted that end-to-end backprop is the only way to train deep networks. But holding the entire network in memory all at once is why AI training is hitting a resource wall.

We found a new way to break the network into blocks and train them independently. The trick? Treating the network’s forward pass like a diffusion model denoising a signal.

This reinterpretation slashes the memory needed to train deep models. In our #ICLR2026 paper (https://t.co/PK5h0mqQSo), we matched end-to-end performance across ViTs, DiTs, and LLMs. We did this while training just one isolated block at a time.

[Ahmad @TheAhmadOsman]

Let me make Local AI easy for you

Give Codex Cli the article below & tell it:

- Infer the right Inference Engine from your hardware + article below
- Use uv+venv
- Pick the right kernels
- Tune flags, batching, KVCache, etc
- Optimize for your hardware & chosen model

See? Easy

[Al Jazeera English @AJEnglish]

Protesters in Naples, Italy have unfurled a giant shroud bearing the names of 18,457 Palestinian children killed by Israel’s genocidal war on Gaza.

[Rasheee 💕 @Rassh_RAJ]

Stanford CS336: Large Language Models from Scratch (2026) is now fully on YouTube, with a few additions beyond the 2025 playlist.

If you want to understand LLMs beyond prompting and APIs, this is worth taking.

https://t.co/QWx8iLcMpR

[eˣrnesto @vedax]

@SethSHowes how long to have sequencer the size of a pregnancy test?

[Adam Back @adam3us]

the efficient market is coming for the memecoins, smart contract coins and air tokens alike.

buy bitcoin, hodl, repeat.

[DHH @dhh]

The reason agents are so good at Linux is that all 40 million lines of kernel code was part of the pre training. Along with every other open source dependency. This really does make every obscure error message shallow, and the system completely malleable.

[Nassim Nicholas Taleb @nntaleb]

Not just a terrorist state, but a terrorist society masked with propaganda playing with your brain. And you are playing along.

You wimps and suckers, let that sink in.

[Ahmad @TheAhmadOsman]

UNBELIEVABLE RESOURCE

The bible for understanding LLMs is NOW AVAILABLE online to read (FOR FREE)

Covers all the concepts below, no experience needed and anyone from any background can understand it

- Tokens / Tokenizers
- Transformers
- Attention
- KV Cache
- Prefill vs Decode
- Decoding Controls / Samplers
- Agents / Tools
- Model Packages
- Chat Templates
- Long Context
- Multimodality
- RAG
- Fine-tuning

Then connects that to running models locally

- What "Local AI" Really Means
- Open-weight vs Opensource
- Quantization
- VRAM Math
- Hardware tiers
- File formats / load safety
- Runtimes / serving modes
- Model selection
- Privacy
- Failure modes
- Benchmarks
- Practical setup paths

You should read this, and if you cannot now then you most definitely wanna bookmark it for later

Opensource / Local AI FTW

[Ahmad @TheAhmadOsman]

MUST KNOW FUNDAMENTALS

The knowledge in this article is the equivalent of what learning how to use Computers was in the 2000s

If you're smart you'd GET ON IT NOW

Don't believe me? Bookmark it and come back to tell me I am wrong in 2 years (spoiler: I won't be wrong)

[Ahmad @TheAhmadOsman]

In a world less brain-rotten this article would be all over X and the AI space

(I wrote it in hopes the world gets less brain-rotten btw)

[Qwen @Alibaba_Qwen]

📣Meet Qwen3.7-Max — our latest flagship, made for the Agent Era.

A versatile foundation for agents that actually get things done:
🧑‍💻 Coding agent, end to end. Frontend prototypes, multi-file refactors, real debugging — nails it.
🗂️ A reliable office and productivity assistant. Get your work done through MCP integrations and multi-agent orchestration.
⏱️ Long-horizon autonomy. 35 hours straight on a kernel optimization task — 1,000+ tool calls, zero hand-holding.
🔌 Scaffold-agnostic. Claude Code, OpenClaw, Qwen Code, or your own stack. Consistent reliability everywhere.

API's up on Alibaba Model Studio. You can also take it for a spin on Qwen Studio.

Go build something wild!🏃🏃‍♂️

📖 Blog: https://t.co/y3AupX3Pa0
✅ Qwen Studio: https://t.co/qpTnrCBjWt
⚡️ API：https://t.co/0sys00osKn