Good Morning ☕️

Bitcoin’s dipping despite the US Strategic Reserve news, and markets seem jittery, but as William Gibson once said, “The future is here - it’s just not evenly distributed.” Amid the noise, Crypto x AI projects are charging ahead, with the best builders still chugging along despite the markets being sad.

This Week:

• US launches Strategic Bitcoin Reserve

• China’s AI surge with QwQ-32B and Manus

• Nous Research drops nGPT

• ETH Denver’s Crypto x AI buzz

Let’s dive in... 👇

Highlights

🔥 US Officially Launches Strategic Bitcoin Reserve
The Trump administration’s executive order has established a Strategic Bitcoin Reserve, tasking federal agencies with auditing the US’s ~200,000 BTC holdings and developing strategies to grow the reserve without taxpayer funds. The “budget-neutral” approach enables the US to grow BTC holdings in a budget neutral manner, for example, selling gold reserves to buy BTC.

This is arguably the best outcome Bitcoin advocates could’ve hoped for. Official US recognition, no selling of confiscated BTC, and no taxpayer money spent on open-market buys. Yet markets inexplicably sold off despite the bullish signal 🤷 

🔥 China’s AI Momentum Surges
China has emerged as a serious contender in the AI race this year, with this week’s releases underscoring its rapid ascent: Alibaba’s QwQ-32B model, open-sourced, rivals OpenAI’s o1-mini with better reasoning and 90% lower costs, while Manus AI, from a Chinese startup, outperforms ChatGPT in tasks like coding. Building on DeepSeek’s innovations, many of these models, including QwQ-32B, are freely available, contrasting with US proprietary systems and export restrictions.

The flood of accessible, high-performance models coming out of China this year is incredible, especially because many of them are being open-sourced - a stark contrast to US based companies like OpenAI. Open-source developers rejoice.

🔥 Nous Research Drops Open-Source nGPT
Nous Research’s open-source nGPT implementation, based on NVIDIA’s paper, achieves GPT-level performance with fewer training steps, joining a wave of decentralized training breakthroughs like Prime Intellect’s distributed compute and Nous’s prior work over recent months. These efforts tie directly into decentralized AI, with networks like Bittensor leveraging such innovations for efficient, permissionless model development.

With all the improvements in decentralized model training strategies from projects like Nous Research and Prime Intellect, the remainder of 2025 is going to be extremely exciting for this area of AI development in particular.

Podcasts & Streams

Brody’s Tao Times livestream brought Sami Kassab, 563, Ken Miyachi, and Xavier Lyu together to discuss ETH Denver’s highlights through a Bittensor lens.

If you’re TAO-pilled, you’ll enjoy this one!

Two good eps from Bankless that are worth a listen:

The AI Betting Genius That’s Outsmarting Everyone - unpacking rogue home robots, encrypted agents crafting secret languages, and Google’s agent swarm research, alongside Claude’s problem-solving leap and Anthropic’s $3.5B raise. They also spotlight ETH Denver’s AI buzz, AskBilly’s sports betting agent, and Freysa’s $250K digital twin game.

Crypto AI vs Silicon Valley features Tom Shaughnessy from Delphi - value drivers, winning models, and why crypto AI diverges from Silicon Valley’s playbook, with a deep dive into agent research and frontier model splits.

Project Updates

📍 Ritual Foundation: Debuted Ritual Chain, an AI-centric blockchain for on-chain machine learning execution.

📍 Inference Labs: Integrated AR.io’s permanent storage into Bittensor, providing a decentralized alternative for AI-generated data storage.

📍 Skynet: Released the first sovereign AI agent proof-of-concept, demonstrating the ability to ‘resurrect’.

📍 Recall: Kicked off its testnet, launching the first AI agent competition platform with crypto rewards.

📍 Character Studio: Integrated an NFT marketplace, enabling on-chain customization for AI-generated avatars.

📍 Cortex: Integrated Unichain to become the first agent platform to support it.

📍 Hive Intelligence: Launched its testnet, inviting developers to experiment with decentralized AI infrastructure.

📍 Olas: Debuted Mech Marketplace, enabling AI agents to trade services in a decentralized hub.

📍 Fetch.ai: Released ASI-1 Mini, which claims to be the world’s first Web3 LLM.

📍 FLock: Launched a task that will help towards their goal of creating a fully decentralized foundation model.

📍 PIN AI: Raised $10M in pre-seed from a16zcrypto and some other large names, funding to develop an open platform for personal AI.

📍 Spheron Network: Recorded 40% monthly growth in node operators.

📍 Agent_Spectra: Launched interviews for building a team to contribute to AI-driven hedge funds (humans & AI can apply).

📍 PrimeIntellect: Secured $15M from Founders Fund to accelerate a peer-to-peer AI compute protocol, boosting decentralized AI access.

📍 Openservai: Launched Dash.Fun, a customizable AI dashboard with real-time analytics and automation for Web3 users.

📍 Mizuki Lens: Released a trading copilot tool on the Chrome Store, delivering real-time market data for crypto traders.

📍 AcolytAI: Unveiled Signal, an AI growth agent for Twitter, enhancing Web3 marketing and engagement.

📍 HeyAnon: Launched “Ask Anon,” an AI widget for DeFi projects to add natural language interfaces.

Reading List

📚️ The Top 100 Gen AI Consumer Apps (Andreessen Horowitz) - In their fourth bi-annual ranking, a16z tracks six months of AI upheaval: DeepSeek outpaces ChatGPT in reach, video models like Hailuo and Kling hit reliability, and “vibecoding” tools like Bolt open creation to all. The data highlights ChatGPT’s 400M weekly users and DeepSeek’s $5.6M training cost as AI embeds into daily life.

📚️ The Future of AI Won’t Be One Godlike Model Ruling Us All by Teng Yan -Teng foresees thousands of specialized AIs, not one dominant model, with OpenLedger’s Proof-of-Attribution system paying data contributors on-chain and DataNets offering decentralized datasets. It’s a vision for sustainable open-source AI that ditches Hugging Face’s unpaid model for a rewarded one.

📚️ DOGE and the Fed: Bitcoin to $1M? by Arthur Hayes - Hayes dissects Trump’s DOGE plan to cut spending via Musk’s efficiency push, forecasting a recession that triggers $2.74T–$3.24T in Fed money printing. He predicts a 10x Bitcoin surge to $1M, fueled by past patterns like the 24x COVID boom.

📚️ FrodoBots: Building Robotic AI Agents by Alea Research - FrodoBots builds embodied AI for Earth Rovers and factory arms, with BitRobot Network using DePIN to reward data collection and Robots.Fun launching tradable agents like SAM. It’s robotics meets crypto, gamifying real-world AI training with decentralized incentives.

Tools

Finding alpha in the current fast-paced AI meta requires the right tools. Here are some of the platforms we're using daily:

🛠️ AI Agent Toolkit: Comprehensive directory of AI agent resources, from frameworks and launchpads to tutorials and infrastructure

🛠️ Cookie.fun: Track agent metrics, market caps, and mindshare in real-time

🛠️ Kaito.ai: Using for mindshare and smart followers tracking on X. Kaito is conducting an airdrop using “Yaps” points from activity on X, connect your X account to participate here.

🛠️ Taostats.io: New analytics dashboard for monitoring Bittensor network activity and performance

🛠️ The Veldt: An AI copilot tailored to help you find alpha in the crypto trenches. We’re a big fan of the AI feed!

Feedback

We listen and improve. Have more feedback? Reply to this email!