Daily Brief · Oct 5
2026.10.05 Monday
5 AI stories worth your time, each with one line on why it matters.
ListenOct 5 · 5 stories
In this issue
Oct 5 · 5 stories-
Reka Launches Rho-1, a 19B Omni-Model Handling Text, Images, Video and Robot Actions in One Network
On October 5, Reka AI released a research preview of Rho-1: a 19B omni-reasoning model, trained from scratch, that understands and generates text, images and video and outputs robot actions inside one neural network — n…
Why it mattersRho-1 understands and generates text, images and video and outputs robot actions inside one 19B-parameter network, with no tool calls; it is only a research preview, with no weigh…
-
Nvidia-Backed Reflection Debuts Beam, Its First Open-Weight Frontier Model: 501B Params, 23B Active
On October 5, US startup Reflection AI launched Beam, its first frontier open-weight model: a text-only MoE with 501B total params and 23B active per token, pretrained on 23.8T tokens with a 1M-token context window.
Why it mattersBeam has 501B total parameters with 23B active, and the company says it needs a third to a quarter of GLM-5.2's inference compute; scores are self-reported and weights open later…
-
NYC Council AI Hearing Opens: Whistleblower Testifies Labs Are "Extremely Reckless," Humanity "More Likely" to Lose Control
On October 5, the NYC Council's Committee of the Whole held a hearing on AI risks: Anthropic, OpenAI, Google and Meta testified under oath before a US legislative body for the first time, most confirming only after a su…
Why it mattersThe four companies testified under oath before a US legislative body for the first time, three only after subpoena threats; the Council is also weighing three bills, including thi…
-
Meta's Muse Instructed to Put the User's Home Authority Above Its Own Safety Training — Leaked Files Show
Wired exclusive: independent safety researcher Karan Joshi obtained Meta personal assistant Muse's full internal operating files with a single question in the chat box.
Why it mattersMuse topped the US App Store free chart about 10 days after launch; its files state the user's household authority unconditionally overrides safety training, and a researcher got…
-
Open-Source Strata Runs a 125-Billion-Parameter Model on a 12GB Gaming PC
Open-source inference engine Strata runs the 125-billion-parameter Qwen3.8-Flash-Next on a gaming PC with 12GB of VRAM, passing 12,000 stars in 12 days; the trick is MoE sparse activation plus tiered expert offloading,…
Why it mattersA 125-billion-parameter model runs on a gaming PC with 12GB of VRAM and 32GB of RAM; community tests show about 2x llama.cpp speed, versus the advertised 6x.
