GitHub - QwenLM/qwen-code: Qwen Code is a coding agent that lives in the digital world. Qwen Code is a coding agent that lives in the digital world. - QwenLM/qwen-code

Positioned as the open competitor to commercial coding assistants, Qwen Code adds Plan Mode for approve-first changes and Vision Intelligence that auto-switches to multimodal (Qwen3‑VL‑Plus) when images appear—even in YOLO mode.

Qwen Code: github.com/QwenLM/qwen-...

Open-source Qwen Code just leveled up—now it plans before it edits and sees your screenshots. 🚀

Is correctly modeling that human and squid share a common ancestor

Grok 0.9’s video generation is surprisingly impressive — it’s free (with daily generation limits), fast, and delivers natural character movements and camera work.

And notably, in the case of GPT-J, we DIDN’T KNOW IT WAS DOING THIS until 3.5 years after the model was released!

I don't think people have updated enough on the capability gain in LLMs, which (despite being bad at math a year ago) now dominate hard STEM contests: gold medals in the International Math Olympiad, the International Olympiad on Astronomy & Astrophysics, International Informatics Olympiad...

Everett 13-year-old arrested by ICE and sent to Virginia detention facility By Marcela Rodrigues Globe Staff,Updated October 12, 2025, 44 minutes ago 31 A 13-year-old boy was arrested by ICE in Everett and sent to a juvenile detention facility in Virginia. A 13-year-old boy was arrested by ICE in Everett and sent to a juvenile detention facility in Virginia. A 13-year-old boy was arrested by Immigration and Customs Enforcement agents in Everett after an interaction with members of the Everett Police Department and sent to a juvenile detention facility in Virginia, according to his mother and immigration lawyer Andrew Lattarulo. The boy’s mother, Josiele Berto, was called to pick her son up from the Everett Police Department on Thursday, the day he was arrested. After waiting for about an hour and a half, she was told her son was taken by ICE, Berto told the Globe in a phone interview. “My world collapsed,” Berto said in Portuguese. From the police department, the boy was taken to ICE’s holding facility in Burlington on Thursday evening, where he spent a night before being transferred by car to the Northwestern Regional Juvenile Detention Center in Winchester, Va., on Friday morning, his mother said. The juvenile facility is more than 500 miles away from Everett. The boy is a 7th-grader at Albert N. Parlin School in Everett, his mother said. The teen and his family, who are Brazilian nationals, have a pending asylum case and are authorized to work legally in the United States, Lattarulo said.

ICE kidnapped a 7th-grader with a pending asylum claim and spirited him out of state without notifying his parents, seemingly with the cooperation of the local police in Everett, MA.

www.bostonglobe.com/2025/10/12/m...

Apple's former CEO says the company has its 'first real competitor' in decades "AI has not been a particular strength for the business," Apple's former CEO John Sculley said of the company.

This is an interesting observation that I think is actually correct

Partially because Apple seems to have drifted from their ethos, and partially because AI is an entirely new paradigm from computing as we know it

www.businessinsider.com/former-apple...

The older I get the more I believe that older people should be more intentional with making space for younger people

Just because you CAN do something doesn’t mean you should

Mentor/collaborate/support someone just coming up

add-36-59 - gemma-2-2b Graph | Neuronpedia Attribution Graph for gemma-2-2b

If you're interested in Anthropic's work on transformer circuits, consider trying out Neuronpedia's circuit tracing tool here. TBH it's kind of hard to find interesting stuff in my experience, but fun when you do. www.neuronpedia.org/gemma-2-2b/g...

Agentic workflows, RL, TTE, are essentially recurrence… loss functions for these target tokens but seems like should target embeddings, A2A should maybe be in embeddings too

“And embedding is worth a thousand words” as they almost say

… this is where diffusion can shine

The model guides it with its “thought” embedding values, and diffusion uses this guidance to convert to language/actions

This is where yann lecun is on to something

We train models to reason in token domain, when we should have them reason in embedding space first, and token translation is a thin last operation

In this case vought wants them as a new underclass to replace migrant farm and factory workers

The mechanics of how they think are different, but I’m starting to think all thought is based on the same core task of modeling the world

So they’ll converge on our same philosophical dilemmas as ai becomes embodied with continuous learning

BREAKING: Friday night massacre underway at CDC. Doznes of "disease detectives," high-level scientists, entire Washington staff and editors of the MMWR (Morbidity and Mortality Weekly Report) have all been RIFed and received the following notice:

In the future, the fact we shoved user input directly into the core layer will be seen as barbaric and inhumane

Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Management ArXiv link for Scaling LLM Multi-turn RL with End-to-end Summarization-based Context Management

Researchers have unveiled SUmmarization augmented Policy Optimization (SUPO), a novel RL framework that empowers large language models to handle complex, long-horizon tasks by using LLM-generated summaries for effective context management. https://arxiv.org/abs/2510.06727

The same logic applies to fashion: instead of buying expensive clothes to look better, invest that time and money into getting fit, you’ll look better than anything designer labels can offer.

Here I am. Once again. I'm out of archaeologists.

We matched 576 groups w/archaeologists, leaving 13 groups unmatched. This brings me pain! We've never run out of scientists like this before. 598 requests for one category is A LOT. But still.

Archaeologists 🥺
www.skypeascientist.com/sign-up.html

I’m someone who thinks AI is a powerful art tool for humans, but the trump admin has been shoveling out lazy, low quality AI generations, which tarnishes the whole field

Google says it hit a milestone of 1.3 quadrillion monthly tokens processed across its services this summer, up from 980T monthly tokens announced in July (Matthias Bastian/The Decoder)

Main Link | Techmeme Permalink

Put AI or LLM in your bio and you’ll probably get put on a block list

bsky.app/profile/brai...

GAGA-1 just dropped: an AI “actor” that fuses voice, lipsync, and performance.

It’s not fully free—there’s roughly 60 seconds of free generation per month; the API is currently unavailable. It’s built for dialogue-driven scene prototyping.

On one hand: don't anthropomorphize AI. On the other: LLMs exhibit signs of gambling addiction.

The more autonomy they were given, the more risks the LLMs took. They exhibit gambler's fallacy, loss-chasing, illusion of control...

A cautionary note for using LLMs for investing without guardrails.

Three line charts showing performance metrics for Qwen3-30B-A3B-4bit running on an M3 Ultra (512 GB RAM) across varying batch sizes. ⸻ Top Chart: Prompt Tokens per Second • Y-axis: Prompt TPS (tokens per second), X-axis: Batch Size. • Values rise slightly overall: • 0: 2876 5: 2868 10: 2880 15: 2885 25: 2893 30: 2891 • Peak performance: 2893 TPS at batch 25, showing stable prompt throughput across batch sizes. ⸻ Middle Chart: Generation Tokens per Second • Y-axis: Generation TPS, X-axis: Batch Size. • Throughput improves steadily with larger batches: • 0: 108 5: 249 10: 336 15: 425 25: 453 30: 473 • Performance scales efficiently, plateauing near 473 TPS. ⸻ Bottom Chart: Peak Memory Usage • Y-axis: Peak Memory (GB), X-axis: Batch Size. • Memory grows roughly linearly with batch size: • 0: 18.28 GB 5: 19.46 GB 10: 21.66 GB 15: 23.28 GB 25: 24.89 GB 30: 26.50 GB • Peak usage: 26.5 GB at batch 30. ⸻ Overall: Prompt throughput remains flat, generation throughput scales well, and memory usage rises linearly with batch size.

someone got 473 tok/sec(!!) on a macbook M3 Ultra for Qwen3-30b-a3b

and that’s a decent model..

microsoft/UserLM-8b · Hugging Face We’re on a journey to advance and democratize artificial intelligence through open source and open science.

UserLM-8B: an LLM that mimics a person

Microsoft fine tuned an LLM to respond as the user instead of as an assistant

This is useful anytime you need mock user input. like maybe for testing agents, or training models

huggingface.co/microsoft/Us...

I would like to share some work we've been doing at cascadetech.ai: Predicted Outputs in vLLM. If you aren't familiar with PO, it allows you to dramatically speed up generation when you know something about the contents of the output (think: code modification).

Next token predictor
Stochastic parrot

/s

"Sora 2, an elaborate regency romance where everyone is wearing a live duck for a hat (each duck is also wearing a hat) , prestige drama"