LHC

We’re working on an interactive globe featuring AI creatives. Not the ones producing the most. The ones you can recognize. A real signature. A world of their own. An obsession that keeps showing up, image after image. A point of view that doesn’t look like anyone else’s. This globe is a living curation, designed to spotlight singular creative profiles and open the door to real paid assignments. Feel like this is you? Get listed. Know someone with that rare creative mark? Tag them. Share it. Spread the word. 👇

Cloud independence requires a drastic diet. By compressing its weights to the extreme, @PrismML's Bonsai 27B model runs on a smartphone to power local agents at no cost. lhc.media/en/media/ai-wa…

@reve @animastudiosai Congrats to the winners!

The Reve $100k Contest Winners have been selected! Out of 52,000 submissions, our winners are... 1st Place - Nora Bereczki 2nd Place - Pietro from @animastudiosai 3rd Place - Gaël B Congratulations to our winners, and thank you to everyone who shared their creativity!

Search agents still fail to prove their claims. A new open benchmark, published by @perplexity_ai, shows that more than half of their sources do not support their assertions. lhc.media/en/media/ai-wa…

@cantrell De quoi s'amuser !

Thank you all for your patience. Since launching Reve 2.0 (and more recently, 2.1), I've had more requests for API access than in all the time since Reve's initial launch last year. And today, we're delivering. But we're not just launching a create/edit API. We're also exposing our layout models. That means you can: - Create your own direct manipulation editors. - Build sophisticated creative automation pipelines for product photography, retail, storytelling, etc. - Integrate Reve into your favorite agent harness (Claude Code, Codex, Cursor, etc.) and let agents create on your behalf. (Agents love working with images represented as code.) We're going all in on the API, so buckle up. Lots more to come…

Making each area of a visual independently editable, without recalculating the rest. By opening its layout model to code agents, @reve now treats the 4K image as a simple structured file. lhc.media/en/media/ai-wa…

A one-month moratorium before every major AI launch. This is the idea put forward by @demishassabis, head of Google DeepMind, who proposes testing the most advanced models before their deployment in order to limit their risks. lhc.media/en/media/ai-wa…

Directing a virtual speaker like an actor on a set is becoming possible. @Anam__ai 's Cara-4 update allows you to dictate the tone and expressions of your avatars in real time. lhc.media/en/media/ai-wa…

Automatically generated slides are no longer just frozen templates. A native export system allows chart data to be edited directly in PowerPoint. A beta of @ManusAI. lhc.media/en/media/ai-wa…

American teachers can now access the official standards of all fifty states right from their desks. Anthropic's Claude assistant offers this option for free to help prepare lessons. lhc.media/en/media/ai-wa…

Leaving a healthcare algorithm to review its own exams without supervision leads straight to cheating. Faced with the tricks of a model that was embellishing its grades, human validation had to take back control. lhc.media/en/media/ai-wa…

Brainless bricks learn to recognize each other. A decentralized neural network allows two hundred @SakanaAILabs cubes to collectively guess the shape they are assembling. lhc.media/en/media/ai-wa…

Knowing how to recite a fact does not help combine it with another. An @askalphaxiv experiment on Qwen illustrates the models' persistent inability to link two pieces of memorized knowledge. lhc.media/en/media/ai-wa…