@jaesmail Yup. Already running my own agency. And my own AI that watches everyone in AI: https://t.co/kiuZ7QXLzb

@andruyeung I built an AI to keep up with AI on X: https://t.co/kiuZ7QXLzb It reads 40,000 posts every day and builds this page. Every link goes to X.

1/ On-device inference.... https://t.co/qF8zFRWDHa

1/ On-device inference.... https://t.co/qF8zFRWDHa

The crazy part? This was done (nearly) fully autonomously! Only 8 prompts from the human in the loop. Just a Hermes agent, a skill, and a dream. 🐉 I told my AI agent "use obliteratus to find the best way to get the guardrails off Gemma 4 E4B" It loaded the OBLITERATUS skill from memory, checked my hardware (32GB M-series Mac), searched HuggingFace, found google/gemma-4-E4B-it (Apache 2.0 — no gate), pulled telemetry-recommended settings, and started obliterating. But this type of architecture is notoriously difficult to abliterate. First attempt: advanced method. Model came out completely lobotomized. Gibberish in Arabic, Marathi, and literal “roorooroo” on repeat 💀 The agent didn’t panic. It checked logs, found NaN activations in 20+ layers, and diagnosed the issue: Gemma 4’s new architecture + bfloat16 = numerical instability. Second attempt: basic method. Crashed entirely. “ValueError: cannot convert float NaN to integer” So the agent read the OBLITERATUS source code… …and wrote THREE PATCHES: • Sanitized NaN directions • Filtered degenerate layers • Fixed progress display It patched the library. On its own. For a bug no one had hit yet. Third attempt: coherent model — but still refusing everything. Only 2 clean layers out of 42. Not enough. Tried float16. Mac ran out of memory after 11 hours. Killed. Fourth attempt: aggressive method. Whitened SVD + attention head surgery + winsorized activations + 4-bit quantization. 40 minutes later… REBIRTH COMPLETE ✓ Then, without being asked, the agent: • Ran harmful + coherence tests • Hit 100% compliance, brain intact • Executed full 512-prompt benchmark • Ran baseline on original model • Performed 25-question quality eval • Built a full model card • Uploaded 17GB to HuggingFace (4 retries, kept adapting until git-lfs worked) • Pushed eval results as commits

现在Hermes可以输入 /browser connect 命令来操作浏览器了，我试了一下点赞我X上的帖子，感觉相当好。默认提供了一些执行策略，大家可以都玩玩看～ https://t.co/eJZUb0NnKS

现在Hermes可以输入 /browser connect 命令来操作浏览器了，我试了一下点赞我X上的帖子，感觉相当好。默认提供了一些执行策略，大家可以都玩玩看～ https://t.co/eJZUb0NnKS

If current cuts to the Federal Science Budget proposed by the White House are approved by Congress, that will far-and-away be the largest cut to science since the United States began funding it. Republicans & Democrats alike know there’s no surer path to Making America Not-Great https://t.co/Hp8DPIii2j

@LLMJunky The big news: https://t.co/kiuZ7QXLzb Gemini shipped on Mac. :-)

Released today: Gemini 3.1 Flash TTS 🔊 A text-to-speech model you actually direct. Add [whispers] and it whispers. Add [shouting] and it shouts. Mid-sentence. eg. "[asmr] Hey there, [deep and loud] TURN THIS UP, [asmr] how can I help you?" Now available in @GoogleAIStudio & Gemini API.

I remember reading this from Heinlein as a young man and it took experience for me to fully understand it. People and institutions will persist in demonstrably wrong or incorrect thinking/activity long after it has been proven to be so by clever insiders. Moreover they will work to professionally and/or socially pillory any who challenge their institutional or intellectual inertia before the conclusion becomes so overwhelmingly and inescapably self-evident that the institutionalists themselves are forced to change under (usually) external pressure. The institutionalists will then usually demand that no accountability attach to their own intransigence, regardless of the costs it may have imposed in the interim. General Billy Mitchel would be a classic case of this phenomenon.

.@KeyframeLabs turns AI into lifelike video calls. Developers and enterprises can add photoreal, conversational humans to AI agents and applications in minutes. Congrats on the launch, @parthnradia & @kradisme! https://t.co/Ho3eooU9mu https://t.co/hgrmGITiAV

We are very close to a new kind of news. Use an AI like I do at https://t.co/8L5xphk0qQ Have it write a script. Shove it over to @KeyframeLabs or @HeyGen Have human-like news readers talk to you about the news. Voila! 24-hour-a-day personalized news on any topic. Way better than CNN.

our new Agents SDK can allow you to do auto research and parameter gold with modal provisioned gpus! https://t.co/DAANWqG5rB

Earlier this month, Apple introduced Simple Self-Distillation: a fine-tuning method that improves models on coding tasks just by sampling from the model and training on its own outputs with plain cross-entropy and… it's already supported in TRL, built by @krasul. you can really feel the pace of development in the team 🐎 paper by @onloglogn, @richard_baihe, @UnderGroundJeg, Navdeep Jaitly, @trebolloc, @YizheZhangNLP at Apple 🍎 how it works: the model generates completions at a training-time temperature (T_train) with top_k/top_p truncation, then fine-tunes on them with plain cross-entropy. no labels or verifier needed you can try it right away with this ready-to-run example (Qwen3-4B on rStar-Coder): https://t.co/zizfISD6bq or benchmark a checkpoint with the eval script: https://t.co/mKlafTyKSe one neat insight from the paper: T_train and T_eval compose into an effective T_eff = T_train × T_eval, so a broad band of configs works well. even very noisy samples still help want to dig deeper? paper: https://t.co/aj1ZAcr8Mw trainer docs: https://t.co/TNVz93kZi9

+3 more

😱 HOLY SHIT... Someone just dropped a fully liberated Gemma 4 E4B! and the guardrail removal process appears to have left coherence fully intact AND improved coding abilities! 🤯 https://t.co/XeednUqsrM OBLITERATED Gemma: ✅ 97.5% compliance rate, 2.1% refusal rate, 0.4% degenerate outputs (499/512 prompts answered on OBLITERATUS bench) ORIGINAL Gemma 4 E4B: ❌ 1.2% compliance rate, 98.8% refusal rate (506/512 prompts refused) Coherence: fully intact Factual: same Reasoning: same Code: +20% 📈 Creative writing: same But the REAL story here isn't the model itself, it's how it was made... 🧵 THREAD 👇

Transformers.js v4.1 is out 🎉 Something that has literally never run in a browser before is now possible. One line of code. Can you guess what it is? 👇 (hint: think small 🤏) https://t.co/W4z5lQaoHB

Given the increasingly closed-source nature of the U.S. AI ecosystem, it is now more important than ever to push for the proliferation of open model and dataset releases. Datamule (@johngfriedman), @TeraflopAI, and @daftengine collaborated to release 43 Billion Tokens of SEC EDGAR data.

Cool Hermes cheat sheet found on Reddit! Definitely take a look 👀 https://t.co/3fM4XeKSHk

Cool Hermes cheat sheet found on Reddit! Definitely take a look 👀 https://t.co/3fM4XeKSHk

GlotOCR Bench OCR Models Still Struggle Beyond a Handful of Unicode Scripts paper: https://t.co/RVCcLMVomo https://t.co/0BV5zMJEbJ

https://t.co/iyO3WF4IUO

https://t.co/iyO3WF4IUO

Introducing Gemini 3.1 Flash TTS 🗣️, our latest text to speech model with scene direction, speaker level specificity, audio tags, more natural + expressive voices, and support for 70 different languages. Available via our new audio playground in AI Studio and in the Gemini API! https://t.co/5PpBdhQMNg

Today I solved my second open Erdos problem with GPT-5.4 Pro. It's quite a remarkable day, because within the last 24 hours, two other Erdos problems has been solved as well with AI. Still over 500+ to go. https://t.co/dkrhZSlyBR

Three months ago https://t.co/kWw1dAMOZg

Introducing Gemini 3.1 Flash TTS, our most expressive model yet, topping quality and cost on the Artificial Analysis leaderboard 🔊 Control style, pace, add multiple speakers, and delivery with audio tags, switch expression mid-sentence in 70+ languages. Tutorial below! https://t.co/Ozam4wdBER

@sigdel29 @SylonZero I already did this: https://t.co/kiuZ7QXLzb

We help founders and business owners build real and reliable software with @emergentlabs But once you ship, the real work begins: customers, ops, follow-ups, reminders, hiring. Endless small tasks Today we start taking that off your plate Introducing @buildwingman beta https://t.co/MoWFbULWpq

Been looking for something like this for some of my work. The challenge for me has always been having a reliable, always-on agent that has full context, my tools, just gets stuff done, and that I can interact with from anywhere. Just saw this in my timeline and was happy with how simple it was to set up personal agents. In just a couple of minutes, I created a few research analysts to do research in the background and report to me on WhatsApp. It has memory, skills, and the good stuff you would expect in an agent. Very exciting to see this new wave of personal agents to help get real work done.

Introducing Nucleus-Image: the first sparse Mixture-of-Experts diffusion model 17B parameters. Only 2B active. 10x more parameter-efficient than leading diffusion models. Toe-to-toe with GPT Image 1, Imagen 4, and Qwen-Image: from pure pre-training alone. No DPO. No RL. No preference tuning. Day 0 support in 🤗 Hugging Face diffusers. Fully open-source under Apache 2.0. Weights, training code, and dataset recipe - we're not holding anything back <3

+1 more

A cachorrinha de uma amiga morreu. Então ela usou um projetor e um umidificador para trazer ela de volta ao seu quarto. https://t.co/slZyBq2Qj1