🦍 State of the Market..

The Crypto x AI market cap stands at $28 billion today, down 17% from last week. Not surprising, given how bad market sentiment has been lately — are you following all the drama on X?

Fetch (FET), SingularityNET(AGIX), and Ocean (OCEAN) have shown surprising price action. Since the news of the merger of these 3 protocols, they have announced the ratios of exchanges for OCEAN and AGIX to the ASI token.

The FET token has a 1 to 1 exchange with ASI. So there's speculation on the valuation of the new ASI token being between the three tokens, less than FET and more than OCEAN / AGIX as FET is the largest by market cap.

🏆 Caught Our Eyes..

• Fetch.ai, SingularityNET, and Ocean Protocol token merger delayed to July 15

• Minting for AO computer via bridge deposits goes live, attracts $200M+ TVL

• Grass debuts the Grass App Store and expects an influx of new users.

• Crypto advocate Michelle Bond launches new think tank to help shape digital assets and AI regulation

• Crypto and AI Could be a $20 Trillion Megatrend, Bitwise Says

• Kalos discusses ParallelTCG’s Wayfinder AI protocol

• Genlayer introduces an interactive “intelligence contract” sandbox for developers

• Akash launches permissionless chat using open-source models

🐰 Research Highlight — Hyperspace AI

Every week we present to you early, interesting crypto AI projects that caught our eye.

This week, we look at Hyperspace AI, a Peer-to-Peer decentralized blockchain network for AI agents.

Hyperspace CEO Varun Mathur calls it a “BitTorrent but for AI.”

Just imagine:

AI agents on a decentralized network running from any consumer device (laptops, desktops, phones) and scaling to millions of nodes.

This network includes Wikipedia-like curation tools for fine-tuning open-source models, all within an integrated UX and AI-native browser.

That was a real mouthful, so let’s break it down.

Hyperspace AI offers access to open-source models like Codestral, Mixtral, Llama3, Phi3, and more through their node app.

Users can connect and receive ChatGPT-comparable generative AI results without running the models locally on their computers.

Example: In real-time, a consumer MacBook can request an LLM to be run on another network machine and get results back instantly.

Why Crypto is important

As AI agents become more prevalent, we need a way to prove computations without redoing them. Hyperspace uses Zero-Knowledge (ZK) technology via a ZK engine to address this requirement for a robust AI agent network.

But Hyperspace is not just a P2P network and ZK engine.

They’re also developing:

• Collama: a collaborative Wikipedia-style LLM

• VectorRank: ranking vectors on the web to optimize AI model training

Quick Thoughts

• There’s a points system launching soon, so it’s probably worth downloading the app and giving it a spin. It takes just a few minutes.

• The user interface is pretty slick and will simplify the use of open-source models. It is still raw: I encountered bugs like network errors and slow inference generation.

• CEO Varun’s mega-thread provides the best info we found on Hyperspace

• Hyperspace has grand ambitions, to make AI free and accessible to millions. This is a vision I can personally support and get behind.

🧠 Open Source Corner

• Meta makes headlines by releasing several new open-source AI models, including a vision-language model, an audio watermarker and a multi-token prediction model

• 🍿 OpenAI co-founder Ilya starts a new company to create “superintelligence”, challenging OpenAI. Bring out the popcorn!

• Ex-Stability AI founder Emad launches Schelling AI to build open-source, decentralised AI. More details on July 20th

🔥 On X..

Luke Saunders (Delphi Labs) on Verifiable Inference

How to farm the Nillion Airdrop for free

On the ATH Aethir token

Sami Kassab on how to launch a Bittensor Subnet

Comparison of decentralised GPU projects

1kx weekly AI news breakdown