Mams (@MamsLBB) header
Mams (@MamsLBB) profile picture

| Sir Mams Musselton | τ, 𝼃 |

Clinging to a rock
Joined December 2021
Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)Photo by Mams (@MamsLBB)
The climb is over. Meet Ridgeline. Your own fleet of software engineers. Give it a task, let the agents work, and get a pull request you can review. GitHub-native PR reviews and fixes. x402 integration. No local IDE required. Put it to work: app.ridges.ai/
9
37
13
149
12,553
Not everyone who asks to see your wallet should see all of it. Decoy wallets are now live on the taostats app. Enter your decoy PIN and the app opens a wallet you’ve chosen, and nothing else. Pick your own decoy PIN or use your normal one backwards. Here’s how to set it up ⬇
7
13
3
57
5,004
Holy mother of GPU
We have nearly doubled our supply since 1 month ago. Now hosting ~$50m in hardware, 481 B300s, and over 25 different GPU types available now to rent. check us out for all your GPU needs.
18
Mams retweeted
i recreated THE ENTIRE WORLD in minecraft at 1:1 scale. every building on earth, including your house! to do this, i trained a SOTA world-to-minecraft generator on maps data, complete with furnished home interiors! play and check out your neighborhood: worldcraft.gg
1,718
6,450
2,635
123,753
20,709,151
We’re live on Bittensor Subnet 24. 🖥️ → HOST a website or app for free. → RUN your AI agent in its own sandbox. → CONNECT to our inference network (coming soon). Join the waitlist → ms.computer The internet’s computer, run by everyone. $TAO #Bittensor
7
22
6
104
11,154
Nearly $35M in annualized GPU rental spending from a single Bittensor subnet. @lium_io is becoming a serious player in the cloud compute market.
latest 30 day update: new paying customers +78% deposits +94% existing customer spend +46% annualized run rate +79% GPU spend + 87% rentals +67%
10
54
373
12,943
Mams retweeted
latest 30 day update: new paying customers +78% deposits +94% existing customer spend +46% annualized run rate +79% GPU spend + 87% rentals +67%
new month in review just dropped. +45% spend +16% active users +106% rentals +11% median rental time +20% charged +32% first time renters +15% providers -32% rental failures
13
46
16
232
22,120
What if intelligence had no owner? From one idea in 2019 to 128 subnets: the Bittensor story so far. Kusanagi. The halt. The restart. Finney. Upgrades without stopping. The first halving. All footage and the score made on @EveryFrame, Bittensor subnet 117. #bittensor $TAO
8
9
1
22
1,851
The wait is over. Hippius is now on iPhone and Android. It's been a long time coming, but now it’s here for everyone. Back up your photos and files automatically, encrypted on your phone before they upload. No one can read them but you. Your whole Drive, now in your pocket.
33
90
26
366
45,800
Mams retweeted
Gm. We’re cutting list prices.
3
9
1
59
1,525
Mams retweeted
1/ We’re entering a new phase of game generation on 404. The Game Jam showed that creators can already build engaging, playable games with the tools coming out of the subnet. Now, we’re building Found Games, a public platform where creators can generate games, build an audience, and monetise their work on Bittensor.
2
12
4
57
6,123
Mams retweeted
Using Claude, GPT or Gemini with client details in your prompts? Turn on guardrails in SayGM desktop. Detected emails and phone numbers are swapped for placeholders before the model sees them, and restored in your reply.
3
8
56
1,223
four things from this week in Kaspa dev. intern has kept the technical words to a minimum. 1 / kaspa-redteam @KasperoLabs released kaspa-redteam, an open-source testing tool for SilverScript contracts. helps devs find bugs in their own contracts and reproduce them while building. xitter.gq/KasperoLabs/status/211… 2 / Argent Argent merged a fix so devs can use multiple contracts together without running into errors caused by clashing variable names. github.com/argent-lang/argen… 3 / covenant indexing work continued on letting apps ask a Kaspa node which coins belong to a particular covenant. useful for wallets and explorers tracking tokens. this week’s changes keep those searches from asking the node to do too much work at once. still under review. github.com/kaspanet/rusty-ka… 4 / Mid-Chain talks @hashdag spoke about “Opening the Strategic Pandora Box”, work with @OriNewman. @michaelsuttonil covered DAGKnight. both at Mid-Chain this week. recording linked in Michael’s post. xitter.gq/michaelsuttonil/status…
12
76
2
348
17,024
🍁Fall is the best time of the year if you're a mycophile. 🧺🫳🍄 What if I'd share one my spot with you guys ? 🤫 🍄 lium.io 🔸26 available pods for B300 4$/GPU/hr 🔸7 available pods for B200 6$ /idem 🔸9 available pods for H200 3$ /idem 🔸‍& many more.
2
1
2
126
Also, if you are not interested in 🍄 but have friends who rent or provide, you can set up an account and a referral link in 2 clicks at lium.io/refer-earn
21
How come you don’t use @say_gm_ ?
1
1
1,178
Mams retweeted
Jason Calacanis told you to sell Bitcoin for AI. Bittensor's own cofounder just called that idea offensive. That is the most surprising thing to come out of this week's Novelty Search, the weekly Bittensor call hosted by cofounder Const and Mark Jeffrey. But it is not the only thing. In under an hour, they covered Bitcoin, quantum computers, a new wallet upgrade, privacy, which subnets break through first, and why the price looks ugly while the network looks stronger than ever. Here is all of it, step by step, in plain English. Step 1: The tweet that started it Earlier this week, Jason Calacanis posted that Bitcoin will never break its all time high again. He called it a dead network and told people to rotate into AI instead. Then he followed up with something along the lines of buy $TAO. You would expect Bittensor's leadership to celebrate that. They did the opposite. Const said he wanted to reply, but his wife stopped him. Her words: "Don't touch that dynamite." Mark, who is friends with Jason, joked that people came at him like his dog had peed in their yard. His answer: "That's not my dog. It's a wild wolf." Step 2: Why Bittensor defends Bitcoin Const was clear. He thinks it matters more for humanity that Bitcoin survives than almost anything else in crypto. His reasoning is simple. Bitcoin was not built inside Silicon Valley. It was not venture funded. Nobody owns it. That neutrality is the whole point. If Bitcoin had come out of a San Francisco startup, countries like China or Russia would never accept it as a neutral reserve asset. Because it belongs to no one, it can still become digital gold for everyone. And that potential is what funds the rest of the industry. If Bitcoin loses, Const argued, everything built on top of crypto loses with it. Ethereum, Solana, and yes, Bittensor too. Mark put it in one line: "Bitcoin is the primary asset, and everything else is secondary, including us." Step 3: Bittensor is not trying to replace Bitcoin People often call Bittensor "the Bitcoin of AI." Const flipped it. He sees Bittensor as "the AI of Bitcoin." It borrows Bitcoin's ideas, like a fixed supply and open mining, and points them at useful AI work. It is not trying to become the world's reserve currency. He contrasted that with projects trying to out mine Bitcoin and replace it. His view: you can never recreate Bitcoin's neutral, ownerless start. Trying is pointless. So when someone asks, "What do you need Bitcoin for when you have TAO?", his answer was blunt. He finds the question offensive. Step 4: Scarcity and abundance, the yin and yang This was one of the best ideas on the call. AI and Bitcoin are opposites. • AI makes everything abundant and changeable • Bitcoin makes one thing perfectly scarce and unchangeable AI is sold as an exciting future with a scary downside. Bitcoin is sold as a scary present with a hopeful future. You need both. Const went further. AI will create enormous wealth, but it could also widen the gap between the people who control it and everyone else. In that world, a money with only 21 million units that nobody can print becomes more important, not less. Nobody can raise a trillion dollars and dilute your share of it. His point: the Bitcoin thesis is quietly getting stronger every day the AI boom accelerates. Step 5: The quantum problem, explained simply Now to the biggest product news. Every crypto wallet is protected by math. A powerful enough future quantum computer could, in theory, break that math and forge your signature. Here is the part most people do not know. Bitcoin is better protected than most chains. When you receive Bitcoin, the network only sees a scrambled version of your key, called a hash. Your real key only appears when you spend. An attacker would have to break two things at once. Most other chains, including Bittensor until now, show the raw key on chain at all times. If that math ever breaks, those wallets are exposed. Step 6: What Bittensor just built The Bittensor team built two layers of protection. Layer one: hashed wallets. You can swap any key on Bittensor into a hashed key. Your address stays the same. Explorers and the command line tool work the same. The difference happens under the hood. Your real signing key stays off chain, and it gets replaced with a fresh one every time you sign a transaction. That gives Bittensor wallets the same kind of protection Bitcoin has. Layer two: quantum resistant signatures. Bittensor also added two newer signature types, ML DSA and Falcon, which are designed to survive quantum attacks. You can combine them in a multisig wallet that also sits behind a hash. To break it, an attacker would need to break several different kinds of math at the same time. Const went out on a limb and said this could make Bittensor wallets the most secure in crypto. But he added a big asterisk himself. • The chain's own node keys have not been migrated yet • Bittensor is not as decentralised as Bitcoin • You only get the protection if you actually move to the new wallet type Step 7: What you should actually do For almost everyone, the answer is calm and simple. When the upgrade is live in your wallet, consider moving your larger holdings to a hashed wallet. Small mining wallets matter much less. As Const joked, if someone is powerful enough to break the internet, they are probably not going after your miner on a small subnet. Nothing is forced. Ignore anyone telling you to rush or send funds anywhere. Wait for official wallet and tool updates. He also reminded everyone why this is not just a crypto problem. If that math breaks, email, online banking, and shopping all break with it. It would feel like 1993 overnight. Step 8: Will Bittensor add privacy? Mark asked if Bittensor would add Zcash style private transactions. Const said it is technically possible. They have already looked at how to do it. But he leans against it. His view is that Bittensor works best as a transparent system where everyone can see who earns what. That openness makes it harder to cheat, which is the whole point of a fair AI network. Mark agreed with him. If you want a digital Swiss vault, use Zcash. Bittensor is doing AI. And if someone really wants privacy, they can bridge their TAO elsewhere anyway. Step 9: Which subnets break through first Mark's view is that Bittensor wins the way Ethereum did. One to three standout subnets reach a billion dollars in market cap and pull the whole ecosystem forward. And it now looks like inference will be first, meaning subnets that serve AI models to users. Two trends are behind that. Trend one: frontier models only stay ahead for about five weeks before an open model catches up. After that, value goes to whoever serves it cheapest without losing quality. Trend two: big centralized compute companies like CoreWeave carry heavy debt with fixed contracts and variable financing. If rates move against them, they get squeezed. Bittensor miners do not carry that risk. They plug in machines and supply compute on demand, the same way Bitcoin miners do. The world, as Mark put it, reorganized itself in Bittensor's favor. Step 10: The hard part right now Const was honest about the pain. Demand for top GPUs is exploding. Because Bittensor miners rent compute at today's prices, miners are paying roughly twice as much as before. Unlike CoreWeave, the network cannot raise a big pile of capital and lock in cheap capacity ahead of time. But he sees the other side of the cycle coming. When the buildout overshoots and there is too much compute, Bittensor's open markets get to soak up all that excess supply. Survive the squeeze, and the network comes out stronger. Step 11: Why research subnets look quiet Commodity subnets like inference and compute rental were always going to earn first. The problem is obvious, and the work is about efficiency. Research subnets are different. They look like nothing is happening until suddenly everything works. Const gave an example. Subnet 97, Ingot, switched to model quantisation, which shrinks AI models so they run on smaller machines. Within about a week, he said, its miners produced the best model you can run on a MacBook, beating a well funded team of Stanford and DeepMind alumni. He also pointed to subnets 10 and 14, which now turn money into inference optimisation. Their next challenge is turning that into steady paying customers. Step 12: The ugly price and the strong network This was the emotional close of the call. Const admitted the price is ugly. But he called this the most aligned and best performing moment in Bittensor's history. His claim: almost no one in crypto produces real revenue except Bittensor and Hyperliquid. Mark compared it to Jeff Bezos watching Amazon's stock fall 95% while every internal metric kept climbing. Bezos knew it would come back. He just did not know when. That is exactly how Mark says he feels watching the subnets today. My read on all of this The headline is not the quantum upgrade, even though it is a real and meaningful step. The headline is the stance. When a well known investor told the world to sell Bitcoin and buy TAO, the people building TAO refused the compliment. They chose to defend the asset their whole industry stands on. That tells you how they see Bittensor. Not as a Bitcoin killer, but as a productive network built on Bitcoin's ideas, chasing real revenue instead of memes. Agree or disagree with their bullishness, that is a more serious pitch than most of this market offers. Do you see $TAO as a rival to Bitcoin, or as the AI layer built on its ideas?
15
22
4
119
13,657
Every company shipping AI agents asks one question: how can I stop my agent from going rogue? To answer it, HaloGuard Output Guard now sets a new state of the art (SOTA): 83.63 average F1, ahead of the previous best at 82.5. All measured on public benchmarks. And we did it with a 0.8B model, a quarter of the size of the previous best. Small enough to run efficiently. Cheap enough to make checking every response practical. This was the focus of Phase 3 on Trishool: • 7,500+ jailbreaks submitted • 125+ unique miners • 15 weeks of challenges It comes at the right time. We’re taking our product to enterprises now, and this makes our security stack stronger. We’ll share the models on Hugging Face, the arXiv paper and a full explainer soon. There’s much more to come, and we’re looking forward to an exciting week ahead. SOTA AI security, on Bittensor. Ψ
7
19
12
88
10,738