Max Sebti, CEO of Score (SN44) on why the subnet couldn’t exist without Bittensor $TAO:
“Why would you leave something that would provide you an output that is 400 times more efficient than the state-of-the-art model?"
Taoswap just got a new visual identity.
A cleaner, sharper, faster look for trading subnet tokens, tracking live metrics, and exploring the Bittensor ecosystem.
We’re starting with a new look here on X, before a major product update coming very soon.
Welcome to the new Taoswap.
You asked. We delivered!
The Vidaio API is live.
Developers, agents, and platforms can now build directly on top of our video optimisation layer.
One API for making video lighter and more valuable.
Subnet 85 is open for builders.
https://t.co/gLAhiX3VI6
@vidaio bittensor:native
this week, we're turning things up a notch at @manakoai.
we'll be at @VivaTech, europe's largest startup and technology event, where @JeffBezos headlines this year's edition, for our first "hello world" to the broader tech scene.
read the article, there's an avant-première hidden inside 👇
Honored to share that I joined @opentensor last week as Head of Comms
And after 64 weeks running these Bittensor ecosystem recaps from my own account, I’m glad to bring the format to the Foundation account as part of the first things we’re putting in place to better highlight what’s happening across the ecosystem
Time to make Bittensor impossible to ignore
on sept 15 2008, lehman brothers collapsed and gave the whole world a reason to believe in $BTC
on june 12 2026, a single government order forced anthropic to pull its most capable models offline for every foreign national on earth, and gave the whole world a reason to believe in bittensor:native
humans only react to negative things. sad but true.
in 2008 we learned the banks were the problem. in 2026 we’ll learn that one kill switch sitting inside a centralised AI lab is the problem.
This shouldn’t phase you.
You have Ditto.
We built Ditto to offer Fable level inference, with open models.
1. They can’t take open models from us.
2. Ideal memory creates ideal inference… and we cracked the code on memory.
Access to intelligence should not depend on a handful of companies or governments.
This is why open, decentralized, permissionless AI matters.
This is why Bittensor matters.
Some people keep telling me this network $TAO is too complex. They say it like it's an insult.
Let me show you what that complexity actually looks like, because it's the entire moat.
Most chains are one token and prayer. Bittensor is 128 self-contained economies. Each with its own token, its own halving schedule, its own burn rate, its own governance battles. All auditable in real time, block by block, by anyone with a browser.
The community built a Bloomberg terminal on top of it. For free. Walk the full stack anytime.
The burn layer.
Right now 47 subnets are at 100% burn, recycling every miner emission instead of paying it out.
• 59 subnets in chronic burn (95%+)
• 14 in high burn (80-95%)
• Network average: 68%
That's emissions held back, not sprayed. IntoTAO @IntoTAO tracks every burn streak and 90-day trajectory live.
The revenue layer @TaoPortal
The dashboards publish the brutal truth every subnet would love to hide: $12.9M burned in 30 days against $56M in outflows. Coverage ratios listed per subnet. No exceptions.
A milestone just printed too. Liquidity SN77 hit 313% coverage, burning more than it emits. An actual surplus. There will be more.
And it's not just burns anymore. @subnetradarcom just shipped a Subnet Traction page pulling verified external revenue from $TAO Institute. Real customers. Real invoices. Outside the emissions loop entirely:
• @chutes_ai SN64: $1.45M over 90 days, TI verified
• @lium_io.io SN51: $530K
• @Bitcast_network SN93: $160K, TI verified
• @VantaSN8 SN8 and Targon SN4: $100K each
Then the deal sheet. 18 commercial deals logged:
• @webuildscore SN44: contracts with PwC France, Eyeball Football, AVIA petrol stations, Reading FC
• @yanez__ai SN54: screening validation for a top-3 global asset manager with $4.5T AUM, a top-5 crypto exchange, and a $13.5B payments platform
• @_redteam_ SN61: threat detection live across 1inch, 25M users
Subnets aren't pitching decks anymore. They're invoicing Fortune-500-grade clients.
The supply layer.
$TAO itself: 52.4% of the 21M cap circulating. First halving executed December 2025. Next one December 2029.
But here's what almost nobody tracks. Every alpha token has its own 21M cap and its own halving clock. Targon SN4 and Chutes SN64 have already crossed 50% issuance. Thirteen subnets have effectively gone through half way of thier halving.
Supply shocks, parked inside supply shocks.
The conviction layer.
This is the one that should stop you stone cold. Under BIT-0011, $45.3M of alpha is locked across 126 subnets. 93% of it is perpetual. Committed forever.
Available to sell: $0.00.
Not low.
Zero.
222 wallets playing Subnet Kings with 90-day maturation curves, 10% eligibility gates, owner aggregates versus challengers. Ownership itself contested on-chain, in public.
Net flow last 7 days: +$5.7M more locked.
The shipping layer.
51 releases across 36 subnets in the last 14 days, every GitHub release and commit summarized into cards. Dev Signals scores activity across every active subnet and then does something almost nobody in crypto does: publishes its own track record. Hit rates, misses, median returns, all of it on the page.
This is Honesty as a feature.
The intelligence layer.
23 measurable pillars per subnet. Current network state: 93 healthy, 25 on watch, 10 critical.
And @TAOTemplar taoflute's forensic dashboards track owner wallets, codelines per day, discord activity, manual burns.
That's the exact data that got SN84's emissions paused last week. The network polices itself with its own transparency.
And underneath that, the rotation: root stake has fallen to 75% while alpha stake climbed past 30% and rising.
Now tell me, what other asset gives its holders this?
Bittensor gives you block 8,383,662, refreshed every 12 seconds, with the community grading every operator's homework in public.
Complexity isn't what keeps people out. It's the moat that pays the ones who do the work.
$TAO
DYOR.
"@webuildscore is just marketing"
n=280 frames, 95% cluster-bootstrap CIs, ignore regions masked, p95 latency on a 4-thread i5, no gpu
we even scored the VLMs on F1 instead of mAP to give them their best shot
handicapped ourselves to be fair. still alone in the top-left.
highest accuracy, near real time, on cpu
the pareto frontier doesn't lie
Just the key numbers in case you missed the AMA:
- $300K+ in sales so far in 2026 (5 months)
- Active pipeline over $1M
- 11 clients
- 20%+ of receivables in buybacks since end of March‘26
- Grown asset holding by 17% while reducing token commitments by 20%
We are bullish!
bittensor:native @yanez__ai
Just a reminder about tomorrow's session.
Lots of information to share with the community. Revenue, new sales, an update on Proof of Humanhood, update on partnerships, and more.
See you tomorrow.
@yanez__ai bittensor:native
Score x Vidaio.
Following our @ProofOfTalk presentation, @vidaio_ (SN85) is partnering with us to be one of the first teams to launch their own vision AI challenges on SN44.
The goal: bring advanced visual intelligence into VidaioOS, making video archives searchable, actionable and monetisable at an entirely new level.
Two subnets, one network effect.
More to come in the coming weeks.
We are thrilled to announce our partnership with Nexartis. A true infrastructure partnership that opens our solution to new markets.
bittensor:native @yanez__ai
Happened just now:
✅ Phase 1: Prove Beam orchestration works
✅ Phase 2: Performance
⏳ Phase 3: Scalability
⏳ Phase 4: Open it to the world
We just completed Phase 2:
10 GB transferred from R2 → Internet → R2 in 21 seconds on a distributed network.
Phases 3 and 4 are coming fast.
Beam SN105 #Bittensor
One Day with Bittensor. Not a week. One day.
Seven different fronts in the war for who owns intelligence, and $TAO was advancing them all at once.
No company on earth ships like this.
@MacrocosmosAI launched Project Orion, an early pre-training run of Orion-100B. A 100B parameter model trained across 16 pipeline stages and 3 replicas on globally distributed single GPUs.
30%+ Model FLOP Utilization on A100s.
Up to 65% of data-center training efficiency, using hardware that costs a fraction of the price.
The old AI moat was simple:
You need the billion-dollar cluster.
Bittensor keeps attacking that assumption.
We already have Covenant-72B. Now Orion-100B is pushing the same idea further:
Underutilized compute around the world can become frontier training capacity.
@oroagents SN15 dropped post-training results.
A distilled 4B model with comparable correctness to frontier models, at under 1/10th the cost and roughly 2x the speed.
Their ShoppingBench stack scored 42.7.
GPT-5.5 scored 38.7.
This came from roughly 24,000 high-quality agent trajectories, with 20,000 more flowing daily.
A Qwen3-4B base climbed from 18% to 42%.
That is open competition distilling frontier capability into something smaller, faster, cheaper, and easier to run.
@manakoai, through Score SN44, announced @oblong_inc, Nasdaq: TWAV, invested in Manako and partnered on North American commercialization.
A public-market company plugging into a Bittensor subnet.
@webuildscore is turning existing camera infrastructure into real-time operational intelligence, powered by SN44.
That is the public market starting to touch decentralized AI infrastructure.
@chutes_ai SN64 showed the revenue line that matters.
A year ago, almost nothing earned per token served.
Today, roughly $280K per trillion tokens.
And this is happening while cutting models and compute.
Same work.
More revenue per unit.
Then that revenue flows back into buying and staking the token.
A subnet paying for itself in public.
Conviction went live and teams started locking.
Score locked $1M of SN44 in perpetual conviction.
@heydittoai SN118 locked 40,000 alpha, 100% of the team’s holdings.
@taostats shipped a live conviction table showing subnet-level lockups and king-flip timelines.
This is not “trust me bro.”
This is teams putting long-term commitment on-chain where everyone can see it.
Then the same day:
@trishoolai SN23 launched HALO, an open guard model that sits in front of AI apps and classifies what goes in and out.
@theminos_ai SN107 announced its scientific advisor, Philipp Koellinger, founder of DeSci Labs, will speak at Proof of Talk in Paris on Bittensor as infrastructure for verifiable scientific claims.
@zipcodenetwork SN46 went live, bringing decentralized intelligence and capital under one roof for real-world financial services.
@mvtrx_79 SN79 teased GenTRX, exchange-history-trained trading AI.
@desearch_ai SN22 and @Bitcast_network@Stitch3_ai shipped tooling.
Ditto’s MCP went live for Claude, Codex, and Hermes.
Read that again.
Distributed training.
Model distillation.
Nasdaq partnership.
Revenue compounding.
On-chain governance.
AI security.
Decentralized science.
Real-world finance.
Trading intelligence.
Search infrastructure.
All in one day.
This is where the comparison to companies breaks down.
Google can ship fast inside Google’s lanes.
NVIDIA owns hardware.
OpenAI ships models.
But no single company ships frontier training, shopping agents, genomics, on-chain finance, public-market partnerships, AI safety, search infrastructure, and trading intelligence in the same day.
A company is capped by its payroll.
Bittensor is capped by everyone on earth who can solve a problem and prove it.
That is the difference.
The closest historical comparison is not a company.
It is the early internet protocol era.
A base layer nobody owned, where thousands of unrelated things could bloom at once.
$TAO
DYOR.