0xPrismatic

This is quite cool: SpaceX mirror token.
Questions in my mind are:
- how much liquidity will there be
- what legal claims will token holders have

Decentralized training in practice:

Compute, data, and validation are all done by separate actors, each earning tokens or micropayments.

Correctness is often enforced through some form of cryptoeconomics: cheat, and you get slashed. Play fair, and you get paid.

Open weights ≠ open systems.

Most “free” AI models today were trained by large AI research labs with strategic reasons to release them.

But what happens when that stops?

Tokenized AI may be the way out. 🧵

今日の分散コンピュートネットワークがこのAIスーパーコンピュータのパフォーマンスチャートでどこに位置するかを見るのは非常に興味深いでしょう。

h/t @EpochAIResearch

You can solve coordination.
You can solve verification.
You can even make it fast.

But without incentives, decentralized training doesn’t work. Why would anyone burn compute for free?

Let’s talk about why incentive design is the final boss. 🧵

if you're building with crypto-native users as your primary market, it’s going to be tough to find sustained demand

to me, the most interesting startups have products already geared towards infinite markets (AI, robotics etc)

Twelve months ago, we didn’t have “Nasdaq companies aping into TAO” on the bingo card.

But here we are.

- Synaptogenix (SNPX) did a MicroStrategy-style bet on Bittensor stock jumped 40% after it announced a $10M TAO buy (2x its own market cap). They’re aiming for $100M.

James Altucher is leading it. $5.5M personal stake. They're releasing preferred shares and TAO-linked warrants. And a full rebrand incoming.

- Oblong (OBLG) is next. $7.5M raise to accumulate TAO and back Subnet 0. The stock popped after the announcement.

The thesis for these companies is simple: TAO is scarce, programmable, and productive. It’s the native asset of decentralized intelligence.

This won’t be the last. We’re watching the start of a public market land grab for crypto AI infrastructure.

We made it. One full year of @cot_research!

Since the 1st edition of our AI & Crypto newsletter was published on 11 June 2024, we’ve been breaking down the madness so you don’t have to live inside Arxiv or scroll Twitter/X until your eyes bleed.

This is the biggest tech shift of our lifetime. It’s chaotic, it’s fast, and if you’re not neck-deep in it, it’s easy to miss what’s really happening.

So, a big thank you for riding with us. Year two starts now. 🫡

Another @cot_research drop will be landing in your inboxes in a few hours.

It's a thought piece by @ChappieOnChain.

Take a guess: what rabbit hole are we going down this time?

(link in bio)

Weekly AI Edge #51 is out! Read this, then get back to enjoying summer:

🌈 Project Updates
= @NillionNetwork's new Enterprise Cluster is live, with Vodafone, Deutsche Telekom, Alibaba Cloud, and stc Bahrain, aiming for a privacy-native internet.
= @TRNR_Nasdaq, listed on NASDAQ, is raising $500M to build the biggest AI-token treasury on a US exchange, backed by ATW and DWF Labs.
= @USDai_Official entered private beta with $10M in deposits for a yield model tied to tokenized Treasuries and AI assets.
= @PondGNN launched AI Studio and Pond Markets to help AI projects grow and fund.
= @Worldcoin launched native USDC and CCTP V2 on World Chain, enhancing transfers for 27M users.
= @peaq and Pulsar launched a Machine Economy Free Zone in the UAE for AI-powered machine pilots.
= @thedkingdao is deploying $300M with a sports-betting hedge fund via an on-chain DeFAI system.
= @CrucibleLabs launched Smart Allocator to auto-stake TAO into top subnets.
= @hyperlane introduced TaoFi's Solana-to-Bittensor USDC bridge, unlocking DeFi access for Solana, Base, and Ethereum.

🌴 AI Agents
= @Virtuals_io released I.R.I.S., a Virtuals Genesis AI agent on Ethereum, for contract security alerts.
= @TheoriqAI launched Theo Roo, an AI strategist for real-time on-chain efficiency.
= @AlloraNetwork started a six-week Agent Accelerator with $ALLO grants for top agents.
= @Gizatechxyz's Arma now integrates into Rainbow Wallet for yield tracking.
= @Chain_GPT launched AgenticOS, an open-source AI for posting crypto insights using on-chain data.

🐼 Web2 AI
= @MistralAI released Magistral, a multilingual model for domain-specific tasks.
= @xAI and Polymarket are partnering to integrate Grok’s AI with prediction markets.
= @OpenAI launched o3-pro, the new ChatGPT Pro model, with enhanced features.
= @Yutori released Scouts, AI agents for personalized internet alerts; beta at https://t.co/gxJvB6iC7h.
= @Krea entered image modeling with Krea 1 in private beta, offering artist-grade output.

+ much more alpha in the full newsletter @cot_research (link in bio)

h/t @issyadelaja

New research essay dropping on Wednesday at @cot_research

This time, we’re not just unpacking how the protocol works. We’re exploring if it’s investable.

As usual, subscribers get 1st look. (link in bio)

planning a token launch? Pivot to an IPO instead.

you can still rewrite that investor deck tonight before sending.

私たちはBittensorがCrypto x AIが展開されるのを見るための最も魅力的な場所であると信じています。

ここでは、チームと話し、野生のサブネットを見ることで学んだいくつかの教訓をご紹介します: 🧵

Bittensor now has over 110 subnets—and counting

Here are 12 Bittensor subnets that have stood out to us, and we're keeping on our watchlist👇

Just released a detailed deep dive on decentralized training. We cover a lot in there, but a quick brain dump while my thoughts are fresh:

So much has happened in the past 3 months and it's hard not to get excited
- @NousResearch pre-trained a 15B model in a distributed fashion and is now training a 40B model.

- @PrimeIntellect fine-tuned a 32B Qwen base model over a distributed mesh, outperforming its Qwen baseline on math and code.

- @tplr_ai trained a 1.2B model from scratch using token rewards. Early loss curves outperformed centralized runs.

- @PluralisHQ showed that low-bandwidth, model-parallel training is actually quite feasible... something most thought impossible

- @MacrocosmosAI releases a new framework with data+pipeline parallism + incentive design and starts training a 15B model

Most teams today are scaling up to ~40B params, a level that seems to mark the practical limit of data parallelism across open networks. Beyond that, hardware requirements become so steep that participation is limited to only a few well-equipped actors.

Scaling toward 100B or 1T+ parameter models, will likely depend on model parallelism, which comes with an order of magnitude harder challenges (dealing with activations, not just gradients)

True decentralized training is not just training AI across distributed clusters. It’s training across non-trusting parties. That’s where things get complicated.

Even if you crack coordination, verification, and performance, none of it works without participation. Compute isn’t free. People won’t contribute without strong incentives.

Designing those incentives is a hard problem: lots of thorny issues around tokenomics which I will get into later.

For decentralized training to matter, it has to prove it can train models cheaper,
faster,
and more adaptable.

Decentralized training may stay niche for a while. But when the cost dynamics shift, what once seemed experimental can become the new default, quickly.

I'm watching closely for this.

私たちは重要なBittensorメトリクスを追跡してきました (h/t @taoapp_)

データは、ほとんどの人がまだ理解していない物語を語っています... 👇

Decentralized compute networks (DCNs) offer cheap GPUs.

But enterprises aren't touching them (mostly).

WHY? WHY? A 🧵

This was a great read and added perspective on valuing Bittensor subnet tokens @UnsupervisedCap @Old_Samster

"With a 2-year hold, the discount approaches ~75%"

私たちの最新のBittensorレポート @cot_research から：

2025年12月までに、現在のトレンドを外挿することによって、468百万ドル以上の資本がアルファトークン市場に移動すると推定しています。

100以上のBittensorサブネットの中で上位10のサブネットは、総排出量の55%以上を受け取っています

排出量の5%以上を受け取るサブネットは4つだけです：
- @chutes_ai
- @taohash
- @gradients_ai
- Targon.