AI is moving faster than ever. If we don’t claim ownership over this powerful technology we may never get another chance.
IOTA by @MacrocosmosAI starts training open models on #Bittensor SN9 today.
Massive update to SN1 just dropped. Complete overhaul.
1000+ LLM agents crawling the internet and generating endless volumes of rich synthetic conversational data across a plethora of different tasks.
Come see what we built!
IOTA (Incentivized Orchestration Training Architecture) is a framework for pretraining large language models across a network of heterogeneous, unreliable, permissionless and token incentivized machines.
In our technical primer we report the following advances:
Incentivized
📈 dTAO whitepaper takeaways 📈
dTAO's whitepaper dropped 👀 revealing that the latest design of dTAO is the best yet - offering stability, clarity, and opportunity.
Here's what we're excited about 🚀 read on for our 5 takeaways...
(1/6) 🧵
Subnet 1 is pushing the limits of intelligence further than ever. By combining the outputs of multiple agents, we can now produce responses that exceed GPT-4o.
Meet @MacrocosmosAI 's Mixture of Miners.
Last night was one of the most amazing experiences of my career. We launched 3 products live on a call with the #bittensor community and somehow pulled off 3 live demos!
And just like that, we crossed the threshold and stepped into this new world. History has been made.
@const_reborn and @opentensor you did the impossible, with style. Thank you for everything.
What a legendary night.
What a time to be in #bittensor and #ai.
For those who said truly permissionless LLM training is a pipe dream.. you're right.
Billions of bits of tensors orchestrated across the web, a symphony of data.
Anyone, anywhere can join us at any time.
No borders, no BS.
This is our pipe dream --
iota.macrocosmos.ai/dashboard
Subnet 1 is doubling down on building the fastest deep researcher in the world.
This will allow anyone to create accurate and detailed reports (even research papers) in seconds, not tens of minutes. We're pushing a major upgrade to the network soon and writing the SN1 DR paper.
🎉 Our dataset mixing results are in: SN9, pretraining, is producing models competitive with SOTA rivals from companies like DeepSeek AI, Mistral, and Google on prominent benchmarks (1/4)
While developing IOTA we ran weeks of experiments on testnet with 30 A100 miners. We load tested the orchestrator with 100 times the traffic we were creating (>10x full base network) and we marched to main net.
But this is Bittensor..