Few days ago, many doubted the pace of development on SN102… now the dashboard tells a different story.
📈 3,108 scored models
⚡ 5/5 validators live
🔥 256/256 miners active
🧠 Continuous training, scoring & weight distribution in motion.
This isn’t hype — it’s visible progress. Building never sleeps. SN102 is getting stronger by the day. 🚀
@connitoai #ConnitoAI #SN102 #Bittensor #TAO #DecentralizedAI #AI
🧠 AMA Recap — Isabella Liu $TAO subnet @ConnitoAI SN102
The Nerds hosted Isabella Liu, subnet owner of @ConnitoAI and founding engineer at OTF since 2021.
She was there before subnets existed. The first OTF network was a distributed training task. Five years later, she is back to finish what she started.
This is not an agent subnet. Not inference. Not data scraping.
This is distributed training, the hardest problem in the Bittensor stack.
The core issue is simple. Traditional distributed training tops out around 40B to 80B parameters. SOTA models are now 600B to 1T. A single miner cannot afford to train at that scale. If Bittensor relies on traditional architectures, it will always be playing catch-up with centralized labs.
Isabella’s answer is to stop forcing distributed training into architectures built for centralized compute. Reinvent the architecture for distributed from the ground up.
That is where Mixture of Experts comes in.
Each miner trains different experts locally with no communication during training. At step X, the pieces merge back into one model through weight-based merging in a DiLoCo-style setup.
Real example from the AMA: take a 16B model with 64 experts per layer. A miner grabs 8 experts most relevant to a math task, finetunes locally, then passes weight updates back to the validator for loss evaluation.
Cost per miner drops compared to data parallel training. Communication overhead drops compared to pipeline parallel training.
To prevent drift between independently trained experts, there is a shared expert acting as a communication channel. Attention layers are frozen to serve a similar stabilizing role.
This builds on Meta’s Branch-Train-Mix, AI2’s FlexoLMoE, and DeepSeek’s ESFT paper, which showed expert-specific training can outperform LoRA.
The new part is combining those ideas with a decentralized incentive mechanism.
Isabella made a point that stuck with me. The current miner-as-GPU model is limiting. If you want a subnet’s output model to match OpenAI, you would need the owner’s research team to be as big as OpenAI’s.
That is not happening.
But if miners have flexibility in what they train on, and you treat Bittensor’s talent pool as the research team, then you have a real shot.
@ConnitoAI is designed around that thesis.
The team is Isabella and George, her university alumni who did post-grad research in distributed training. They are backed by Crucible Labs for packaging and go-to-market.
The roadmap is research-first.
Next deliverable is a research paper analyzing design choices from the whitepaper. Purpose is to build credibility and attract collaboration with research firms. Then comes a working pilot. Then the goal is to convert collaborations into paying customers.
They are open to B2B and B2C, anyone who needs a model trained on specialized data.
Long term, the vision is training-as-a-service with the modular advantages MoE provides.
Isabella was direct about marketing. Product before marketing. Engineers before marketing hires. The whitepaper dropped before this AMA.
The honest gaps:
Data privacy for defense and healthcare is still unsolved, and she said so.
Man hours are a real constraint. Two-person research team with more to build than they can currently cover.
No paying customer yet. Go-to-market is a plan, not a pipeline.
This is early stage, research-heavy, and technically ambitious.
But Isabella has been thinking about this problem longer than most people have known Bittensor exists. The architecture is differentiated from every other distributed training attempt in the ecosystem. I am willing to bet that she will succeed
We’re excited to share the Connito whitepaper V1: a framework for decentralized, composable MoE adaptation.
We trains sparse expert subsets, validates updates through Proof-of-Loss, and turns open-model improvement into a distributed expert-level market.
Read the whitepaper: https://t.co/KKkGVnMIRF
Honestly, after taking some serious hits in the Templar+ saga, I’ve become a lot more cautious about how I approach subnets now. Painful lessons, but they forced me to actually learn the system properly instead of just chasing emissions.
Over the last while I’ve spent a lot of time watching your TaoFlows updates, plus TaoStats, TauFlute, TaoMarketCap, TaoRadar and other analytics platforms, trying to understand what’s really happening underneath the surface.
Whale concentration, miner activity, burn rates, emissions stability, validator behaviour, capital flows ffs, all of it matters now.
What I’ve realised is that most subnets should probably be treated more like immature startups than mature businesses. Some teams are brilliant technically, but many still don’t seem to fully understand traditional business fundamentals, communication, investor confidence or long-term trust. That combination matters when my capital is involved.
Because of that I’ve moved roughly 70% back into root, earning hardly anything, and only about 30% is spread across subnets. Even then, it’s active daily management and hard to keep on top of.
It’s not really “stake and forget” anymore unless you’re willing to accept serious losses.
I run my own business outside crypto, and honestly, subnets have become almost a full-time job if you want to protect your capital properly. You can’t disappear for a week and expect it to be fine.
Emissions rotate mad overnight, liquidity shifts, miners vanish, unlocks happen, sentiment changes really fast.
Affine was a good example for me recently. I liked the project, saw emissions improving around the 1%+ mark, moved in, woke up the following morning and emissions had collapsed back to zero.
That’s the reality right now.
At the same time, there are subnets I still believe in long term. I like Score, Targon, Resi, Lium Gradients plus few more , but Jesus, you have to watch them all the time. Targon went from almost 5% emissions to 0% overnight.
What really opened my eyes was seeing discussions around some subnets burning 100% of alpha and others potentially heading toward deregistration, despite looking strong from the outside.That tells you how unstable this entire ecosystem still is.
I still have belief in Bittensor long term but it may be buy and hold for awhile. But I think retail needs to understand this isn’t simply “buy TAO and pick a subnet you like.”
There’s a huge amount going on under the bonnet, and if people don’t understand the mechanics, they can get cleaned out very quickly. That’s kind of where my head is at this week anyway.
Well, you did ask @TAOFlows 😏 keep up the good work #Bittensor #TAO #DeAI #Subnets #dTAO #TAOFlows
1/ QuickPerps Falkor has been unleashed! 🐲 🟣
Offering a powerful perp experience on the @0xPolygon PoS network
Enjoy up to 50x leverage on selected tokens and zero gas fees
Trade Now: https://t.co/7z2hO5lWFO
For details, check here👇🧵
150 LXPs again? uh yes please.
Linea OG Rush is here with @AcrossProtocol@MendiFinance and @LynexFi!
Explore the new additions to the Linea Ecosystem & complete onchain tasks to WIN 150 LXP! 🤩
Ends June 4th: https://t.co/L6MZDogRQJ
We've been cooking up something BIG 🌋
Now bridging to @LineaBuild via Across can earn you 50 LXP!! The Linea OG Rush campaign is live from May 22 - June 4 on @IntractCampaign 💫
Earn your LXP here: https://t.co/cjIB5cvlX7
The Call of Taiko with @Timelesswallet is live on @galxe! 🥁
Complete testnet warm up tasks to earn points & prepare yourself to claim your Taiko Factions badge 🙌
🔍 Ready to become part of the #CallofTaiko movement?
Start your adventure right here: https://t.co/OUaoA0y1v5