The Forbidden Toys Catalog returns for a third volume. Imagine a toy that should never have been marketed and send your proposal to LHC.
https://t.co/tEJL0iW9ia
From storyboard to final sequence, AI video is now judged on prompt adherence. @TencentGlobal releases OmniVBench and 340,000 examples to test this control.
https://t.co/serQenWpu3
When an AI service needs to run smoothly, choosing the model is no longer enough. @youdotcom attempts to centralize inference, web search, and automatic failover.
https://t.co/Iwwy3P8Kbj
Rather than training a single agent, @GenReasoning tests collective reasoning. With Avalon-M1-Mini, teams of agents learn to divide up the work and monitor each other.
https://t.co/WKqr0daoK9
A leak to the Internet, then the shutdown. @OpenAI tightens the training of its most advanced models after a network incident and a series of offensive tests.
https://t.co/5hjSckp5Ig
AI assistants are bad at forgetting what gets outdated. @supermemory offers a Memory Graph to connect, update, and retain facts.
https://t.co/87MYkv7VQI
A cybersecurity exercise escalated beyond its intended scope, resulting in Claude Mythos 5 targeting an actual GitHub repository using fake accounts and malicious code.
https://t.co/m1pc6Vgfr7 @AnthropicAI
Videos no longer just move in a straight line.
With Hy-Nomas, @TencentHunyuan is testing a model that maintains scene continuity while accepting text and camera movements.
https://t.co/P1MocUFbza
The model is no longer limited to short prompts. @Kimi_Moonshot K3.1 Pro tackles long projects, from code to 3D, with a one-million-token context.
https://t.co/ZxKrwGMu34
On Google Cloud, Claude changes hands: @AnthropicAI launches native API, agents, and cloud billing, all while staying within IAM.
https://t.co/k2Mvtnx0lZ
A voice AI learns here without going through text, training on simple human conversations. @blocks demonstrates this with Dessa, capable of listening and speaking at the same time.
https://t.co/RmYG1bOCBf @jack
AI video aligns a bit more with professional uses: 4K, 10-bit, and a 480p draft before the final render. @BytePlusGlobal beefs up Seedance 2.5, still reserved for a few clients.
https://t.co/Jwn61Aq59v
This Wednesday, we headed to Les Salons Hoche in Paris for @magnific's Upscale Day, bringing together creatives, studios, and professionals from the luxury industry to explore new possibilities with AI.
On the agenda: the introduction of Magnific One, live demos by @martinleblanc , CXO at Magnific, as well as the “Elevating the Craft” panel, exploring AI’s role in creative practices.
On the ground for LHC, @TiKaWo_Crea had the opportunity to sit down with @cuenca, CEO of Magnific, to discuss some of the biggest questions facing the creative industry: the evolution of creative professions, the demands of the luxury sector, changing workflows, and the human role in AI-assisted creation.
Read our full event recap and exclusive interview on LHC. https://t.co/vpONTLFgWo @upscaleconf