0xPrismatic

From our latest Bittensor report @cot_research :

We estimate >$468M of capital will move into alpha token markets by December 2025, by extrapolating current trends.

The top 10 Bittensor subnets (out of 100+) receive >55% of the total emissions

Only 4 subnets receive >5% of emissions:
- @chutes_ai
- @taohash
- @gradients_ai
- Targon.

there are many zombie protocols with enough treasury (from the boom days) to survive for years with zero PMF.

sad part is that there is not enough founder vision and courage to boldly reinvent the project before it fades into irrelevance

One thing people often miss: not all compute is equal!

We treat GPUs and FLOPs like they’re interchangeable units, but they’re not. The context: who owns the infra, who controls it, who can shut it off, matters more than we admit.

Your model might run just fine on AWS or on some decentralized network. Output looks the same. But the trust layer is entirely different.

Decentralized compute offers what hyperscalers can’t: censorship resistance, no central point of failure or kill-switch, and real user control over data and execution.

That sovereignty should come with a premium. Today, it doesn’t. Most devs still optimize for cost.

But that will shift. slowly, then structurally. As pressure builds and workloads demand stronger guarantees, control becomes non-negotiable.

And the repricing follows.

quick excerpt from our latest Bittensor research report at @cot_research

2 words: subnet summer

New research lab where R&D of AI models is completely transparent and open-sourced.

Cool to see but how do you fund the training compute which is v expensive?
(hint: this is where crypto comes in)

POV: you bought $300k of a memecoin and this is what you got

🚨 New post dropping today at @cot_research

This is the 1st of our fast ideas: Things which can break out fast, and feels like the right moment to pay attention.

(link in bio)

Crypto’s next primitive isn’t a new token.

It’ll be compute: scarce, trustless, and open to anyone.

Let me tell you why this changes everything:

Weekly AI Edge #48 is out!

@gregosuri, founder of Akash Network, suited up and testified before Congress. We got to say, he looks spiffy.

🌈 Project Updates
= @PrimeIntellect dropped PCCL for decentralized AI training across the internet, optimized for unstable connections, hitting 45 Gbit/s.
= @cookiedotfun goes public with Cookie Snaps, tracking Crypto Twitter influence for COOKIE rewards.
= @koeppelmann launches Circles—mint 1 CRC per hour in a P2P money experiment.
= @BrahmaFi launched Imprint: a soulbound NFT adapting to your on-chain behavior.
= @LayerZero_Core's integration is live on Bittensor, enabling subnet ERC-20s to connect with 130+ chains.
= @heurist_ai unveiled portable AI memory that carries user context across apps with self-custody.
= @Chain_GPT opened a $1M AI Grant Program for Solana projects, with up to $70K per team.
= @PondGNN announced a $20K Sybil Detection Competition with key partners.

🌴 AI Agents
= @virtuals_io launched agent staking; staked tokens earn points with a 14-day cooldown.
= @ii_posts released II-Agent with 92% human-level GAIA benchmark performance.
= @PhalaNetwork introduced Rabbi Schlomo, an autonomous on-chain trading agent for a 1 BTC fund.
= @Infinit_Labs is live with INFINIT Intelligence, a cross-chain DeFi engine rewarding actions with INFINIT Stones.

🐼 Web2 AI
= @AnthropicAI launched Claude Opus 4 for reasoning and Sonnet 4 for boosted performance.
= Google's I/O 2025 was packed with AI upgrades: Gemini, Meet's translation, Veo 3, and Imagen 4.
= @OpenAI released Codex, a cloud-based AI coding agent, now in beta. and acquired Jony Ive's AI hardware startup io in a $6.5B deal.
= @flowith_ai introduced the Infinite Agent for scalable, cloud-native reasoning.
= @v0 released a web development AI model compatible with OpenAI, now in beta via API, SDK, and Playground.

+ much more alpha in the full newsletter (link in bio)
h/t @issyadelaja osuri

Man @cookiedotfun is making me look like an @AethirCloud bull.

(but read my tweets for actual context)

Mental Framework #4: the opportunity no one’s paying attention to

Big shifts in tech don’t usually come from lone breakthroughs.
They come from collisions.

AI & Crypto is one of those collisions. 👇

Mental Framework #3 for AI x Crypto: The Utility–Attention–Trust Triangle

Every Crypto AI project sits on three legs:
Utility. Attention. Trust.

If one weakens, the whole thing wobbles. 🧵

Mental Framework #2 for AI x Crypto: Tokens are the superpower.

When designed well, tokens don’t support the network.
They BECOME the network.

Here's how:👇

The best thinking starts with clear mental frameworks.

Here’s one we keep coming back to:
Crypto ←→ AI

Let’s upack it: 🧵

Weekly AI Edge #47 is out. Can you believe we've been writing on AI for 47 straight weeks now!

🌈 Project Updates
= @worldcoin dropped the Orb Mini, a portable human verification device.
= @NousResearch launched Psyche, a decentralized AI training network with a 40B param LLM.
= @PrimeIntellect open-sourced INTELLECT-2, a 32B model surpassing QwQ-32B in math + code.
= @gizatechxyz raised a $5.2M seed for autonomous finance infrastructure.
= @EMostaque and @ii_posts released II-Medical-8B, a local "GPT-4 level" medical model.
= @BitMindAI launched Subnet V3, reducing miner emissions by 90%.
= @vana announced Vana Academy, a builder track for data businesses, apps closing May 26.
= @NRNAgents released a robotics SDK for Sim-to-Real training and Robotic Sports.

----
🌴 AI Agents
= @Tether_to introduced QVAC for AI agents on user devices with privacy and scalability.
= @Mirai_terminal wrapped a massive AI agent pre-sale, raising 76,831 SOL in 36 hours.
= @symphonyio launched V1 terminal for executable perp trading.
= @arbitrum started Agent Arena, a bot competition with a $50K prize.
= @moofdotfun introduced a no-code builder for A2A agents using Phala’s infrastructure.
= @zacxbt launched Gigabrain’s AI portfolio manager on Hyperliquid, now seeing $5M in beta volume.

----
🐼 Web2 AI
= @PatronusAI unveiled Percival, an AI agent for speedy debugging, outperforming top LLMs.
= @GoogleLabs launched AI Futures Fund offering resources for startups on DeepMind’s stack.
= @xai confronted Grok prompt issues, open-sourcing their system prompts for transparency.
= @SakanaAILabs introduced Continuous Thought Machine model with neuron timing encoding.
= @hedra_labs secured $32M Series A for creator-friendly video tools.
= @OpenAI launched the Safety Evaluations Hub for model assessment tracking and transparency.

+ @getoro_xyz, @MagicNewton and much more alpha in the full newsletter (link in bio)

h/t @issyadelaja