2026-09-22

小米发布 MiMo-V2.6 Pro 与 Flash 全模态开源模型 / 56 signals

Daily Edition

2026-09-2256 signals

Lead Story

3 items

小米发布 MiMo-V2.6 Pro 与 Flash 全模态开源模型

小米 MiMo 发布 MiMo-V2.6 Pro 与 Flash 两个全模态模型,通过规模化强化学习训练。Pro 在多数 Agent 基准上与 Claude Opus 5 和 GPT-5.6 Sol 表现相当,在 Artificial Analysis Intelligence Index 得分 46,为开源模型中最高;能力覆盖编码、computer use、3D 推理和创作。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmubqsb9l03vbrociqf22sr07

Read original ↗

OpenRouter 解读 NVIDIA Nemotron 3.5 Lightning 如何承担 Agent 高频执行调用

OpenRouter 发文解读 NVIDIA 的 Nemotron 3.5 Lightning,这是一款 30B 总参数、约 3B 激活参数的混合专家开源权重模型,定位于工具调用、编码等高频、边界清晰的 Agent 执行步骤,与负责复杂推理的 Nemotron 3 Ultra(550B 总参数、55B 激活)形成分工。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmubyl2o004y7ro9i986nyhfy

Read original ↗

AI & Builders

6 items

Nathan Lambert 撰文分析开源模型的中美实力格局

Nathan Lambert 发布其向美国国会汇报的开源模型现状综述,指出中国开源权重模型已在下载量、基准成绩和学术采用上领先,自 2025 年 7 月起在 Hugging Face 下载量上领先约 1.6B(总 3.2B,为美国两倍)。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmub7xfbe0putrolnfkkepe2c

Read original ↗

Jev introduces a new shape of LLM - System One, aka Decision Models

Last week TypeSafe AI unveiled Jev , their first example of a new category of model that they are calling "System One models" (I'm with Maggie Appleton, I think "decision models" is a better name for these). Jev is an interesting variant on the usual LLM format: it still accepts text inputs, but instead of text output it returns floating point numbers corresponding to categories, yes/no questions, ratings, and associated confidence scores. TypeSafe describe Jev like this: Think of Jev as a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out. It's also very fast, and really cheap . Regular LLMs are priced in terms of input and output tokens, with output generally charged at significantly higher rates. Jev charges only for input - output is free - an

Read original ↗

Cloudflare Python Workers are now generally available

Cloudflare Python Workers are now generally available After a two year preview, Cloudflare's support for running Python code in their server-side Workers platform is now stable: "Python is now a first-class, fully supported language on the Cloudflare Developer Platform". A neat thing about this is how it works. Cloudflare are running Python compiled to WebAssembly via Pyodide in their V8-based workerd runtime. This comes with some limitations, documented here - most notably both multiprocessing and threading are non-functional in the WebAssembly VM. One particularly interesting detail of this is the local development environment story - their pywrangler development tool (confusingly packaged as workers-py on PyPI) runs a full local simulation of their stack, including executing code with P

Read original ↗

Tech & Industry

15 items

GM can’t ‘bring back’ Apple CarPlay because it never left

Last week, GM announced a new software experience for its upcoming Chevy Silverado and GMC Sierra trucks, including a new look for Apple CarPlay and Android Auto, the popular phone mirroring systems that the automaker made a big deal about removing from its electric vehicles in 2023. Naturally, this led to some confusion about whether […]

Read original ↗

Grab an Xbox controller and a $50 Xbox gift card together for just $80

Newegg is hosting a bundle that includes Microsoft’s wireless Xbox controller (in black) and a $50 digital Xbox gift card together for $80. Given that this model typically sits around $50 to $60 depending on the color you want, this is a fantastic value since it essentially knocks its price down to $30. Another way […]

Read original ↗

A cut cable disrupted hundreds of flights across the US

Hundreds of flights were cancelled or delayed on Monday after construction crews in New Jersey accidentally cut a Verizon fiber cable used for air traffic control. FAA Administrator Bryan Bedford said that a circuit failure led to the discovery of the severed fiber cable, ABC News reports. The outage hammered airports across the Northeast, causing […]

Read original ↗

AI Travel & Travel Tech

10 items

World Affairs

15 items

Design & Culture

1 items

Outside the Bubble

1 items

Sound guardians

As Indonesia builds a new capital city, the race is on to preserve the sounds of the rainforest before they vanish forever - by Aeon Video Watch on Aeon

Read original ↗

Builders on X

5 items

If you enjoyed this, sign up for free to my newsletter to get my best AI and product guides. Join 110,000+ subscribers here: https://t.co/yz9xcsfrhF

Open on X ↗

Jev pod tomorrow subscribe on apple /youtube @latentspacepod thanks to @allenpark and Ke for making this happen! https://t.co/5kDekuLbuS https://t.co/vgkesXELrf

Open on X ↗

6. I forgot to mention Apple and Siri. They have all the devices and a great reputation on privacy. They can easily win this race but what holds them back imo is their annual release cycle for major updates and not being on @X and all the other platforms seeking user feedback. The iteration speed for Siri (from the outside) just feels far too slow to compete.

Open on X ↗

My “armchair” thoughts on the personal agent race: 1. Muse is poised to become the leader. The app is intuitive, and Meta is aggressively promoting it across all its properties. Once you use the Muse app, you realize that it makes little sense for an agent to live in an existing messaging app like iMessage. Unless Meta runs out of compute or something, I think this will become the company’s most successful homegrown app after Facebook. It will be much bigger than Threads. 2. ChatGPT is probably still the personal agent leader because it already has 1B+ users. OpenAI’s models, computer use, and voice are arguably better than Muse feature by feature. But it’s hard to build one product for both work and personal use, and harder still to serve enterprises and consumers with the same UX. Even for work, I started using Grok Bot because the Work/Codex split remains too confusing. I’m sure the team is cooking, and we’ll see something good in the next few weeks. 3. Grok Bot isn’t really a Muse competitor because it’s focused on work. Multiple bots naturally fit team-based knowledge work IMO. My bull case for Grok Bot is that it's a few features away from becoming multiplayer agentic Slack for both bots and humans. 4. Google is obviously the dark horse candidate. All these personal agents live off Gmail, Google Calendar, and your Google app's data. Google has their own personal agent in Spark but the last I checked it's a secondary tab in Gemini. I think Google needs to be far more aggressive in making Spark the primary experience and making it intuitive and competitive with Muse. Google however doesn't have a great track record in multiplayer AI... 5. Speaking of multiplayer, nobody has figured this out yet. I can’t easily add my spouse to a Muse chat to plan a vacation. I also can’t loop coworkers into ChatGPT or Grok Bot threads. Multiplayer AI today mostly means tagging bots in Slack. Every major provider is probably working on this, and I think it will be the next big unlock. My personal AI stack right now is Muse for personal, Grok Bot for cloud tasks, Claude for specific use cases, and ChatGPT for everything else. But I think we’re going to see some major updates this week. Everyone should design their personal skills and files so they’re easy to port between harnesses and agents. Otherwise, you’ll spend half your time moving from one tool to another.

Open on X ↗

Built my son his first computer https://t.co/9tQWZFSkku

Open on X ↗
Daily Feed · 2026-09-22 | Tian Peng