LHC

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…

AIs monitoring other AIs, detecting discrepancies in as little as fifteen minutes. @OpenAI is testing this system following an incident, but it still lets flaws slip through. lhc.media/en/media/ai-watch/…

Antibodies designed to activate or block receptors: that is the bet behind Pythia-1. Researchers say they have achieved cellular effects on targets known to be difficult. lhc.media/en/media/ai-watch/…

Generative video is speeding up even further, but the real issue lies elsewhere: maintaining control over references and sound. @pika_labs details H3 Flex, its separate exports, and its open weights. lhc.media/en/media/ai-watch/…

In twenty milliseconds, joey-1 cuts where wordy models write. @baseten bets on a compact decision model, at 0.01 dollars per million tokens. lhc.media/en/media/ai-watch/…

A single image is now all it takes to draft a 3D object. @speridlabs presents Idol, an open model claiming 7.2 GB of VRAM and multiple levels of detail. lhc.media/en/media/ai-watch/…

Claude self-evaluates after its responses. @AnthropicAI sees it as a discreet filter, but the method remains vulnerable to attackers. lhc.media/en/media/ai-watch/…

Tools are becoming commodities; the real scarcity now lies in method. With Fieldwork, @floraai focuses on training for designers, filmmakers, and fashion pros, centered on real-world cases. lhc.media/en/media/ai-watch/…

From living neurons to graphics cards: @BioComputingCo claims to have extracted an accelerator for AI video from a laboratory experiment. Mischief 1 relies on Wan 2.2. lhc.media/en/media/ai-watch/…

Delegating one's emails, research, and bookings to agents working overnight: the concept is well-known. @heyrobynai takes it a step further through orchestration, featuring memory and access to over 300 services. lhc.media/en/media/ai-watch/…

An open-source contest around creative tools turns into litigation. @higgsfield_ai demands the halt of the operation launched by @ComfyUI, which maintains its schedule. lhc.media/en/media/ai-watch/…

Brain signals to adjust an AI to an individual’s tastes: the idea is anything but ordinary. Sabi Research is testing RLNF on faces, with results that are still highly experimental. lhc.media/en/media/ai-watch/…

Changing a title without disrupting the rest is now Brush's business. @canva sees it as an AI capable of reading a graphic document before tweaking its text, layers, and layout. lhc.media/en/media/ai-watch/…