@LootRealms & @Starknet βοΈ
interested in fully onchain games, the roman empire, sourdough, raw milk and all sorts of things!
Nothing here is financial advice
LTX 2.3 π₯° Paper Cut Out Style LoRaπ
This LoRA creates a cozy handcrafted paper cut-out cinematic diorama style where every subject and environment looks physically assembled from layered paper pieces
πWorkflow + repo:
https://t.co/ZoQ0EAITKY
@AzFlin@0xCramJam@Framer_X yep, check out what is happening in the LTX community recently, lora's are starting to give open source models insane capability
well the base models are indeed not fully there yet, I tried for 15-20min in a back and forth with grok + gpt image2 and could indeed not produce one that comes close to an actual pixelart which only hints at many of the character's details to activate the viewer's imagination, the best ones where quite blatant copies like this one below.
Nevertheless, I think every cognitive task can be done by an agent once the reasoning path, the available tools + context is good enough.
I think the problem with current AI generated pixel art is that base models are just made for very different purposes and resolutions, what one would need for this to work is to train a lora on about 500+ images that suit the pixelart demand, describe them using key phrases or sentences and the reuse those in a model that has the lora injected.
Lora's are basically just manuals for models on how to properly do xy, when the model is then facing the same phrases that the lora was trained on, it just calls these weights.
@_breim@theo I think it's the fact that users pay more tokens instead of reasoning, just look at your usage, before it was impossible to use up
At this point I think they're really selling the latest Gen Haiku equivalent as Opus 4.7
1890: rinderpest arrives in East Africa. The Serengeti wildebeest collapse from over a million to 200,000.
Ecologists expect the grassland to flourish. Fewer mouths, more grass. Obvious.
The grassland goes backwards.
1960s: vaccination clears rinderpest from cattle. The wild herds recover. Ecologists brace for overgrazing.
The wildebeest rebuild to 1.5 million. Largest herbivore population on earth.
The grassland gets greener.
More soil carbon. Lower fire frequency. More tree cover than in 1900. The landscape gets more complex, not less.
The papers have been sitting in prestigious journals for decades.
The campaigners continue to argue that grazers destroy ecosystems.
The largest natural experiment on earth says the opposite and it cannot be cited, because it does not say the right thing.
It just sits there. Getting greener.
Taalas wants to etch Qwen3.5-27B into silicon. But 27B is near the transistor limit for a single die.
A distilled 4B fits comfortably. Same capability ceiling for most tasks, 7x fewer parameters. When inference moves to ASICs, the model that fits wins.
Distillation is not a compromise. It is the deployment strategy.
> been paying $200/month for cloud AI APIs
> laptop: M2 MacBook, 16GB RAM
> tried running models locally, garbage quality after 4K tokens
> read this TurboQuant breakdown on Tuesday
> applied 3-bit KV cache compression
> same MacBook now runs 100K token conversations
> quality: identical to cloud
> cancelled all API subscriptions Wednesday
> it's been 3 days
> saved $200/month forever
> with a free algorithm from a free paper
> my MacBook didn't change. the math did
Content is only becoming more enticing from here on
this just elevates the baseline of how enticing the average video is going to be
soon taste & ideas will be all that AI can't help you with
Today we're introducing TRIBE v2 (Trimodal Brain Encoder), a foundation model trained to predict how the human brain responds to almost any sight or sound.
Building on our Algonauts 2025 award-winning architecture, TRIBE v2 draws on 500+ hours of fMRI recordings from 700+ people to create a digital twin of neural activity and enable zero-shot predictions for new subjects, languages, and tasks.
Try the demo and learn more here: https://t.co/VkMd1YpQWI
Pattern recognition 101:
People w/ the best health & physiques mostly eat the same simple meals on rotation everyday like this
People who do extravagant cutesy eating with a million 'healthy' supplements & ingredients have questionable skin, health & yo-yoing struggle physiques
When I moved to Poland, I bought S.Pellegrino, thought this was S-tier water.
Started tasting different waters. Found Muszynianka, more minerals than S.Pellegrino, taste not even close. And everyone lowkey knows about it here.
Went once to a sauna, everyone got Muszynianka, so good that I couldn't find my bottle & the guy drank from mine thinking it was his.
Mineral spring water is the way we were designed to get electrolytes.
β‘οΈ KNOW. PREDICT. REVEAL β‘οΈ
The @Armored_MMA app is now live.
Get to know the fighters. Predict on upcoming fights. Reveal their match winning Relics.
Markets for Pittsburgh on March 21st are open. It's time to #makeyourcall
π https://t.co/psXL1NywHq