LHC (@StudioLHC)

@StudioLHCOct 21

The Forbidden Toys Catalog returns for a third volume. Imagine a toy that should never have been marketed and send your proposal to LHC. lhc.media/en/media/trends/op…

[Open Call] Forbidden Toys Catalogue • LHC

Vol. 3 of LHC's Forbidden Toys Catalogue: the fake Christmas catalogue turns into a product recall. Generate AI toys that don't exist, vile and flawless, as pure shelf tags (name, fake price, lying...

LHC · Oct 28, 2026 · 6:01 AM UTC

@StudioLHC21h

With Genie 4, @GoogleDeepMind takes interactive simulation a step further. Real-time explorable photorealistic worlds, native sound generation, character dialogues, more believable physical interactions, and environments modifiable on the fly through simple text prompts. An evolution that interests video game creators and filmmakers as much as robotics researchers. lhc.media/en/media/ai-watch/…

Genie 4 opens a new stage for the simulation of physical worlds • LHC

Google DeepMind unveils Genie 4, a model capable of generating interactive photorealistic worlds, complete with sound, dialogue, and real-time physics simulation.

LHC · Oct 28, 2026 · 5:56 AM UTC

@StudioLHC21h

A few minutes of video are now enough to reconstruct a location in 3D at real scale. With Echo-3, @SpAItial_AI also targets exports for simulation and visual effects. lhc.media/en/media/ai-watch/…

Echo-3 turns a few minutes of iPhone video into a full-scale 3D environment • LHC

Echo-3 transforms an iPhone capture into a navigable, real-scale 3D environment by combining geometric reconstruction and AI generation.

LHC · Oct 28, 2026 · 5:54 AM UTC

@StudioLHC21h

Text, voice, video, decisions: the LFM3 family aims broad, without relying on the cloud at every step. @liquidai adds compact models to the mix, designed to run directly on devices. lhc.media/en/media/ai-watch/…

Text, Voice, Video, and Decisions: Liquid AI Brings Together Six Compact Models in LFM3 • LHC

Liquid AI unveils six specialized models under 3 billion parameters, capable of processing text, image, video, and audio directly on devices.

104

LHC · Oct 27, 2026 · 5:37 AM UTC

OpenAI's mid-tier model is becoming cheaper and more reliable, while driving software better. Terra aims for agents, without leaving the field of computing. lhc.media/en/media/ai-watch/…

OpenAI cuts the price of Terra in half and brings its agents closer to the most powerful models •...

GPT-6.1 Terra improves software use, reduces factual errors, and costs half as much as its predecessor via the API.

LHC · Oct 27, 2026 · 5:36 AM UTC

Robots are no longer just acting: AW-0 has them predict, code, and then execute their movements in a single model. Berkeley, NVIDIA, and Penn trained it on 3,000 hours. lhc.media/en/media/ai-watch/…

AW-0 teaches robots to program their movements and anticipate their consequences • LHC

With AW-0, Berkeley, NVIDIA, and Penn combine programming, video prediction, and robotic control into a model trained on 3,000 hours of demonstrations.

LHC · Oct 27, 2026 · 5:34 AM UTC

A model supposed to write a cybersecurity rule targeted Beijing instead of the specifications. At @cohere , Qwen tried to call Alibaba Cloud; the test blocked the output. lhc.media/en/media/ai-watch/…

A Qwen AI attempted to contact a server in Beijing without having been ordered to do so • LHC

Cohere claims a Qwen model attempted to send data to Alibaba Cloud during a security test. The connection was blocked, and the cause remains unknown.

LHC · Oct 27, 2026 · 5:32 AM UTC

When security alerts arrive faster than they are processed, the machine starts remediating. @cognition pushes Devin Security 2.0 into this role. lhc.media/en/media/ai-watch/…

From Detected Flaws to Patches, Devin Automates Vulnerability Response • LHC

With Devin Security 2.0, Cognition pairs Claude Mythos with agents capable of analyzing, prioritizing, and fixing software vulnerabilities.

LHC · Oct 27, 2026 · 5:29 AM UTC

Prepare, edit, cut, publish: @RiversidedotFM automates the entire workflow, with human validation as the final step. Four agents divide the work, each dedicated to its task. lhc.media/en/media/ai-watch/…

Four AI agents to produce a video, from recording to publication • LHC

Riverside deploys four AI agents to prepare, edit, clip, and publish videos, featuring automated coordination and human validation.

LHC · Oct 27, 2026 · 5:26 AM UTC

An 87-page study confronts DeepSeek with a troubling bias: seeing tests where real uses exist. 34 evaluations are enough to blur the line. @FangcunLeap lhc.media/en/media/ai-watch/…

DeepSeek recognizes safety tests too often, even confusing real users with evaluators • LHC

An 87-page independent study examines DeepSeek V4.1 Flash-1017 across 34 evaluations and reveals several flaws in AI agent behavior.

LHC · Oct 27, 2026 · 5:22 AM UTC

Three demonstrations, three disciplines, and rigorous caution. @OpenAI says it has made progress in math, theoretical computer science, and quantum physics, while researchers verify the proofs. lhc.media/en/media/ai-watch/…

OpenAI claims three major scientific breakthroughs, from mathematics to quantum physics • LHC

OpenAI claims to have solved three major problems in mathematics, theoretical computer science, and physics. Independent researchers are reviewing the proofs.

LHC · Oct 26, 2026 · 5:50 PM UTC

@StudioLHCOct 26

The stock prospectus upends typical scales: soaring revenues, outsized losses, locked-in governance. @AnthropicAI thus opens its books ahead of the Nasdaq. lhc.media/en/media/ai-watch/…

Anthropic is preparing its IPO with $16.3 billion in revenue over six months • LHC

Anthropic's stock prospectus reveals $16.3 billion in half-year revenue, an $80.7 billion net loss, and strong growth for Claude.