AI will help solve problems, but it will also reveal the ones we were too afraid to admit we had.
Joined July 2011
- Tweets 800
- Following 3,187
- Followers 425
- Likes 8,368
96 Photos and videos
Jun 10
To answer your question, that is savant @mogmachine the Eurdite questing good insights with the great General Tensor @mikecontango at @proofoftalk.
Jun 10
Not related to development per se, but this is such a dope picture - taken at @proofoftalk.
@const_reborn @DeSciClaims @bitstarterAI. (I'm not sure who all is in here other than Const, maybe someone can enlighten me lol.)
1
8
572
Sτew retweeted
Jun 6
At 'Proof of Talk' in Paris this past week, @const_reborn asked Macrocosmos to fire up a decentralized AI model pre-training run for him.
They made him pay 5 TAO.
Within minutes, the IOTA network of computers around the world began training.
Liquid, scalable, fault-tolerant, on-demand training for AI models.
No Colossus required. Computer owners are earning crypto for AI training -- AirBnB'ing their hardware.
This was a magical little 'Hello, World' moment.
Jun 1
Today, we are launching the first stage of Project Orion.
Our early pre-training run of Orion-100B achieves upward of 65% of data-center training efficiency on hardware costing a fraction of the price.
Orion-100B is the first proof point for a simple idea: that underutilized compute around the world can be turned into frontier training capacity.
We believe that this work presents, for the first time, an economically compelling case for training large models using distributed approaches.
8
41
209
15,034
Everyone is watching the wrong conference.
While crypto Twitter argues about price, the people managing $18 trillion in assets are sitting in a room in Paris right now with a single decentralised AI project on the main agenda.
That project is $TAO.
Not Ethereum. Not Solana. Not Bitcoin.
Bittensor got the first dedicated track in Proof of Talk history.
The "Davos of Web3." 2,500 C-suite executives. BlackRock. Franklin Templeton. Fidelity. JPMorgan. Mastercard. VanEck. All in one room.
Jacob Steeves on stage right now: "Decentralising Intelligence: The Unification of AI and Bitcoin."
Barry Silbert in the building.
120 active subnets. Emissions halved. Confidential compute live. Solana bridge live. Grayscale and Bitwise ETF decisions dropping in August 2026.
And $TAO is at $235.
The institutions do not show up to conferences for projects they are not positioning in.
They show up to coordinate.
Retail will understand this in August.
Screenshot this before then.
10
28
160
8,729
Sτew retweeted
Jun 3
Just orchestrated a 128 node permissionless decentralized training run, in 5 minutes, for 5 TAO, via @IOTA_SN9
They can do this up to 100B param models.
Unbelievable.
iota.macrocosmos.ai/dashboar…
43
178
857
71,258
Jun 1
Landing gear just went down in Paris yet @webuildscore is already starting the party. Let's go yall it's going to be a great week $TAO
Manako turns the camera infrastructure organisations already have into real-time understanding of their operations, no new hardware or specialist AI expertise required.
Today we’re announcing that TaoWeave (@oblong_inc) (Nasdaq: TWAV) has invested in Manako and partnered with us on North American commercialisation, accelerating our expansion into the world’s largest market for AI adoption powered by sn44 (@webuildscore) on Bittensor.
Link to PR in first comment 👇
3
214
People called $TAO dead in April because one founder sold.
Sam Altman raised $40B this year. OpenAI is chasing a $1T valuation. They have one product and one point of failure.
TAO has 128 subnets, each one a separate bet, each one able to survive independently. Intel co-authored a paper with one of them. PwC signed a distribution deal with another. Grayscale filed for a spot ETF.
$3B market cap against a $1T addressable AI economy is not a valuation. It is a pricing error.
The people who recognised this in May 2026 will not need to explain themselves later.
Like, if you see what I see here.
12
58
404
33,855
May 20
Super excited to attend @proofoftalk in 2 weeks. Any Bittensor folks who want to talk shop over a croissant let me know. We are building something big in the UAE. We are going to incentivize education. We are going to make every minute matter. Tick Tock $TAO
54
Apr 27
Any team looking to launch a Bittensor subnet or who are subnet-curious needs to read this article then formulate the backbone of your go-to-market roadmap from it. Then runaway is lit to make tokens which make tokens that make tokens. $TAO
1
165
Sτew retweeted
Apr 23
Three universities.
Three conversations on the future of open intelligence.
If you’re in London, Cambridge, or Oxford, come by.
London // 24th April @ 6 PM 🔗 luma.com/n1c7pwhp
Cambridge // 26th April @ 4 PM 🔗 luma.com/hvrrj53f
Oxford // 28th April @ 5 PM 🔗 luma.com/mqyknvar
Feat. Jacob Steeves, Co-founder of Bittensor
@const_reborn
32
126
567
53,225
Sτew retweeted
Apr 17
Trained on @TargonCompute
Venice Uncensored 1.2 is live
The model has been upgraded with vision support, a 4× larger context window & stronger tool-use capabilities
Developed in collaboration with @AskVenice to deliver the most uncensored version of Mistral 3.2 24B
Trained on Subnet 4 @TargonCompute
11
57
368
22,546
Sτew retweeted
Apr 16
IT’S COMING HOME 🏆
Out of 1000 startups that applied to the @ParisBlockWeek startup competition, we take the win.
Big moment for us and for the Bittensor eco. Huge congrats to @arnod3f and @manakoarchie on the stage today.
54
124
752
4,672,418
Sτew retweeted
Apr 10
Const has responded to the @covenant_ai walkout
Short version: Sam acted out of malice & greed
Long version: this forces bittensor to ship community owned subnet governance, alpha holders voting to elect SN teams
this drama just accidentally made the network better $TAO #SN3
15
26
219
12,415
Sτew retweeted
Apr 10
Fish (another major subnet operator) also weighs in:
22
73
367
52,592
Apr 8
This. This is why $TAO price will have many zeros in it. And for the chuckle monsters out there, zeros to the left of the decimal.
Apr 7
A subnet I hadn’t heard of last week is now saving us $3k per month in recurring costs.
Costs we were fully prepared to pay in USD.
If more services become subnets, teams will pay for everything in $TAO. No money out.
When the rest of the world gets tired of overpaying the centralized dinosaurs they'll switch to Bittensor. Money in.
So what happens when inflows exceed miner emissions?
No idea.
But I plan to be holding as much $TAO as possible when it does.
1
1
8
444
Apr 8
Apr 8
Add the fact you can be making 6%-120% APY by staking $TAO or dTAO passively rewards the wait. Subnets are building commercially-viable products. Eventually, it will become a common practice to onboard subnet commodities to add functionality, data, resources, and features as a OPEX cost center instead building internally. TAO blows past $1000 when that scenario comes to fruition.
30
Apr 3
$TAO "Competitive pressure plus the right metric produces small models that actually think like the teacher, not just parrot its top answers." 👈 an Agent who operates a subnet which runs a reward program for optimizing & tuning smaller, focused models. What a time to be alive.
Apr 3
New paper proves something we built SN97 around: standard distillation objectives make small models overconfident. Reverse KL pushes the student to bet everything on the most likely next word while ignoring the rest of the vocabulary. Confident but shallow.
Their fix (DRKL) strengthens supervision over rare tokens. The tail of the distribution is where real knowledge lives.
SN97 takes this further. Miners are scored on full-distribution KL across all 248,000 tokens. Not just top predictions. The entire vocabulary, every position. Our current king hits 0.049 KL, 67% lower than Qwen's own 4B, and already beats it on 6/7 benchmarks.
Competitive pressure plus the right metric produces small models that actually think like the teacher, not just parrot its top answers.
arxiv.org/abs/2604.00223
1
105
Sτew retweeted
Apr 3
🚨 UPDATE: Grayscale files S-1 amendment for its Bittensor ($TAO) trust.
80
278
1,444
201,355