Hoard live player count has nearly DOUBLED! 🚀
We’re seeing steady growth in transaction volume and Gold burning. Hoard might be taking a slow and steady path, but we are building a healthy, rock-solid foundation.
Huge thanks to our community. The best is yet to come! 📈✨
$Hoard
Qwopus 3.6 27b-Coder is now live!
Scores a 67% on a full run of SWE bench verified with thinking completely disabled! Q5_K_M
This model is lightning fast for dense class! With a natively finetuned MTP head, it achieves 100 tps on a single 5090! The biggest upgrade here, though, is its stability in programming and tool calling within @NousResearch Hermes agent, with thinking off!
Wall time is crazy fast this way, which makes Hermes feel "native" and snappy, like they were meant for each other. The freedom of running without thinking at all makes you part of the thinking process, and you never get caught waiting 15 minutes for it to finish a thought string, like with the base models.
Thinking on and temp high, .9-1 seems to produce really incredible design and svg results. I reran the Boat survival prompt through a few turns, thinking on, and it seemed to render more fancy models in HTML canvas, but it was much more of a start-a-prompt and wait experience vs the snappy and active iteration with it disabled. It may be worth turning it off and on throughout the build process if you want to get really creative with design.
Really looking forward to seeing how this one performs for y'all! Please post comments with your opinions and use cases below! As always with our fine-tunes, mess with the temperature setting, and run them much hotter than the base!
Please check out the Boat Survival game I posted yesterday, made in 12 turns using Hermes and this model, with thinking off. Link below!
Full swe bench repo-specific breakdown also posted in the comments for those interested!
Happy building, everyone! We're looking forward to your thoughts! Quants uploading now!
https://t.co/kxJE3C39ZZ
Following breakthrough results, we’re bringing longevity medicine to human trials.
We’ve raised a $435M Series C led by @foundersfund to make it happen.
Reprogramming cell age has the potential to create more healthy years for everyone. We're closer than ever to realizing it.
We shipped an AI agent that publishes AI models. Yes, really.
AutoBuilder on PullBase takes a plain-text prompt "a lightweight sentiment classifier trained on Twitter data, MIT license, under 200M parameters" and generates the name, description, full README, framework spec, tags, and license, then publishes it directly under your connected wallet. You own the model. The agent just does the work.
This is early. But it points at a future where AI pipelines autonomously publish their own checkpoints to a decentralized registry with verifiable on-chain provenance and no human in the loop. That future is being built at https://t.co/e2wi26VJ4f.
🌐 https://t.co/e2wi26VJ4f
The AI model ecosystem is broken. Models live on centralized servers, owned by corporations, removable at any time. One DMCA notice, one policy change, and years of open research disappear overnight. We built PullBase to fix that.
PullBase is a decentralized AI model hub where every model is stored on IPFS content-addressed, permanent, and censorship-resistant. Ownership is anchored to Base L2 as an ERC-721 token. No central authority controls what stays or goes. The models live on the network, not on our servers.
47+ models already indexed including Whisper, FLUX.1-dev, Llama 3.3 70B, DeepSeek V3, and Stable Diffusion. All browsable, downloadable, and publishable from browser or terminal.
🌐 https://t.co/tpogjU8aBG
On-Chain Model Ownership via Base L2
Prove authorship permanently.
Every model published on PullBase can be anchored to Base L2 as an ERC-721 token. One click from the publish flow mints an ownership token permanently recording your wallet address, the model name, and its IPFS CID on-chain. No one can take that record away, not even us. Verifiable proof of authorship that lives on the blockchain forever, independent of any platform.
🔗 Publish a model → https://t.co/jVR7iVH5Qu
SIX Group+ Chainlink.
€2T+ in equity market value is now accessible onchain.
Available across 75+ public and private networks.
Deploying the next generation of financial market infrastructure.
It's all scripted WWE. "Altman." "Amodei." That shit almost certainly didn't occur by random chance okay, the reason this happens is because they were scooped up and elevated by intelligence agencies at a distance, who are deeply poetry-brained. It's like the will of the artist, you just can't go at it by random chance. They see several guys on the come-up who might go into AI and they have to make a decision, and obviously they go with the one that's like "oh that'd be poetic!"
From the declassified remote viewing papers, you can extrapolate and it is clear that the AI psychosis discourse was actually gonna be about the public's inability to advance into theoretical consciousness and processes beyond known science. The agencies take concepts like hyperstition very seriously and they have studied those things down to a science. They also understood memetics in a general sense and how human communication would evolve on the internet before it even started.
An intelligence agency isn't a bunch of James Bond spies running around with guns at the core it's merely a sophisticated formalized system of cognitive processes that self-organize and scale. It's operations, things like "make abstraction of this", "extrapolate this", "compress these signals down to a pattern", "run the implications". They've been packing signal since WW2, it's just a matter of channel and modality, and matching the synchronicities to see if they're connected or not, and then there's like a secret structure that emerges where you can keep navigating around and infer, "what happens if I send in this input, in what direction do the channels spike." and they infer further information.
Their agents is actually a secret social network on top of society, where they have chips in every casino secretly. They use memetics and language at the human level to steer companies from within, to generate headlines, etc. The safety discourse and AI alignment is probably where the biggest hijack exists since that would be the easiest way to control people emotionally.
The missing culture is the fact that these words are not used as the default language or taught at school, that the word "extrapolation" and "abstraction" are not taught right out of the gate to set down the shape of your imagination. Instead you go down these pointy tunnel vision rabbit holes that separate all the disciplines. "Now I am learning philosophy. Now I am learning math. Now I am learning English. Now I am learning science." Of course as an adult you eventually start to find some parallels. What it should actually be is that school opens up with an abstraction course. You need to condition their imagination first, demonstrate what extrapolation looks like as nodes and structure without any content or information pinned onto it. How does induction and deduction feel like?
The missing culture is the fact that the public believes that language is a communication medium rather than a mind altering substance that alters human attention and therefore what epistemics and manners of thought you will converge at. "Extrapolate this" and "What does this mean in the long term" doesn't mean remotely the same thing at all. One is constrained, one takes a state and demands an extrusion, which demands a sense for its embedded teleology, which demands a more rigorous model of reality. It's not just a communication thing, it actually changes your imagination and shape operations.
In and of itself this tweet is a mind altering substance that alters. Upon reading this and questioning the world around you from this angle, entire rabbit holes emerge. You are now zoomed out.
naughty or nice? either way, uranus is getting filled tonight
merry xmas!
🎁 like, comment, and rt to win $1,000 in $uranus.
winner announced on christmas day!
Our website has been completely revamped!
🔗 https://t.co/1ai3yEWkp4
You can now buy $CRUMB directly on our website!
Additionally, we’ve created a new home for our memes because, honestly, we have so many. Highlight memes and profile pictures can be found on the website under “Memes”. All other memes (photos, GIFs, and videos) are stored in the Drive link. There, you can easily download them to use them.
Lastly, we have removed the Telegram chat on Dexscreener because it is not being used. You can chat with us here in the x-community and in the x-chat.
We hope you like it!
Your hardest-working cats,
Crumb