Statistical watermarking of AI text looks like a tidy answer. I think it is the wrong one. It can make the writing worse, it muddies who owns the draft, and anyone who wants to cheat can strip it.
https://t.co/TTAMjiI5Ew
China already ran the short-drama experiment Hollywood is still debating. In Zhengzhou and Wuxi, AI flipped live-action crews into prompt factories. The numbers are worth sitting with.
https://t.co/PtC5WICIuB
While we're waiting for the announcements around the #FutureVisionXPRIZE -- here's a trailer I created for 'Space Cadets' while delayed 10 hours in Orlando by @AlaskaAir last month! Turning lemons into lemonade...
https://t.co/gSyyJqpF8M
Another entry for the #FutureVisionXPrize. Got my process of trailer creation down to be able to put this one together in about 2 days! AI Filmmaking has its quirks, but I love it!
https://t.co/fLQsvNETkb
My fabulous friend @lmoroney has been busy putting his passion for sci-fi and Stargate to work on a couple of show concepts for @xprize contest ...you may recognize a snarkastic voice on Glass Age...please watch rewatch and share into reality! https://t.co/iV00CueX0m #FutureVisionXPrize
Which model can you actually run where you already work? On phones and Pi boards, Gemma 4 is still the practical default. On a single consumer GPU, Muse Glimmer and Qwen3.8-27B just raised the local-agent ceiling.
https://t.co/GIQlZwwrD2
Erik Brynjolfsson on why an AI job apocalypse is unlikely. I work with him through the AI Fund and I agree. Some jobs change. Precedent says net gain if we get ahead of the trend. Learn the landscape. Less to fear.
https://t.co/SNTa7934Kj
AI film infrastructure is forming in public. Frame Rate's https://t.co/8OGG8UbiYs sits above the models (waitlist Sept 10, IBC next). COL AI Studios already claims 300+ AI productions on a 5,000+ title network. Moving fast.
https://t.co/pL9hsrspYm
The models I actually teach and ship sit in the 2 to 12 billion parameter range. They run on a phone or a laptop, and they can still call a larger cloud model when a query earns it.
Gemma 4 and Qwen3.5 are Apache 2.0. SmolLM3 still publishes a full recipe. Llama 4 is open-weight, so read the license before you make it the default local stack.
Here are the product defaults I would load first.
https://t.co/Iao84zpbse
A 110-minute film, made in four weeks for $2 million. An MPA deal covering the model stack that shot it. And a Hollywood working group that finally published language for AI and authorship.
If you write or teach, the useful work is in the production notes, the licenses, and those three authorship buckets.
https://t.co/Srty6kzncs
The models I actually teach and ship sit in the 2 to 12 billion parameter range. They run on a phone or a laptop, and they can still call a larger cloud model when a query earns it.
Gemma 4 and Qwen3.5 are Apache 2.0. SmolLM3 still publishes a full recipe. Llama 4 is open-weight, so read the license before you make it the default local stack.
Here are the product defaults I would load first.
https://t.co/Iao84zpbse
I want to share a little project I've been working on with @dhewlett of #Stargate fame. It's called 'The Glass Age,' and it's an optimistic sci-fi movie/series for the #FutureVisionXPrize
Please check out the trailer and like/comment on YouTube. Views are important!
https://t.co/a60cUW42iQ