The Fed held at 3.5% to 3.75% on Wednesday. Powell said the quiet part: "We have not made as much progress on inflation as we had hoped." The dot plot still shows one cut this year, but seven of nineteen officials now expect zero. Stocks sold off. Crypto followed.

The real macro variable is not the Fed. It is oil.

Brent surged from $72 before the US-Israel strikes on Iran to over $103 by mid-March. Hormuz moves roughly 20% of global oil. Tanker crossings are down more than 70%. Over 200 vessels are anchored outside the strait, waiting. JP Morgan sees $6,300 gold by year-end. Deutsche Bank sees $6,000. Gold is consolidating around $5,000 to $5,200 after spiking above $5,400 in the days following the strikes.

This is a war-driven oil shock layered on sticky inflation and a Fed that cannot cut even if it wanted to. Liquidity is still on hold.

Bitcoin is sitting around $70,000, down roughly 45% from its October high near $126,000. But underneath that number, the structure has improved. ETF outflows have collapsed 94% from November's peak. Long-term holder selling has dropped 87%. The forced sellers are mostly gone. The market is lighter and cleaner than it has been in months.

As we wrote in January: crypto does not front-run fear. It front-runs liquidity. The fear is here. The liquidity turn is not. Yet.

But two protocols are not waiting for it.

The regulatory picture is the most constructive it has ever been, even if the timeline is messier than anyone hoped.

The GENIUS Act is now in the implementation phase. Regulators have until July 18, 2026 to finalize licensing, capital, custody, and AML rules for stablecoins. The FDIC has already proposed procedures for bank subsidiaries to issue stablecoins. Five national trust bank charters have been granted to digital asset firms, including Circle and Ripple. Stablecoin supply has surpassed $300 billion, with transaction volumes reaching $46 trillion in 2025. That is not speculation. That is infrastructure at scale.

The CLARITY Act, the bigger legislative prize, remains stuck in the Senate. The House passed it 294 to 134 in July 2025. The stablecoin yield provision has become the sticking point, pitting the banking industry against the crypto industry. Senators say they are working on a compromise. JPMorgan analysts have described passage by midyear as a positive catalyst. Prediction markets are pricing roughly 70% odds of it becoming law this year.

The SEC has shifted from enforcement-by-ambiguity to rulemaking. Chairman Atkins has directed staff to recommend allowing certain tokens to trade on non-SEC platforms. The "innovation exemption" concept for crypto startups is expected to formalize soon.

The direction is clear. The pace is frustrating. But the structural barriers that kept institutions sidelined for years are being removed one by one.

We first covered Bittensor in our , where we compared io.net, Akash, Render, and Bittensor. We said TAO had the highest upside and the most uncertainty. We called it "highly experimental" with a "complex adoption curve."

A year later, the experiment just delivered results.

On March 10, the Templar team completed Covenant-72B. 72 billion parameters. Roughly 1.1 trillion tokens. Trained entirely on Bittensor's Subnet 3 by over 70 contributors using commodity internet hardware. No centralized cluster. No whitelist. Anyone with GPUs could join or leave freely. An arXiv paper confirmed it as the largest decentralized LLM pre-training run on record, achieving a 67.1 MMLU score and outperforming centralized baselines like LLaMA-2-70B.

Then this week happened. Jensen Huang discussed Bittensor on the All-In Podcast. Chamath Palihapitiya called it "a pretty crazy technical accomplishment." Huang endorsed decentralized AI innovation as complementary to proprietary models. TAO jumped 17%, breaking $300 for the first time since January. It is up roughly 80 to 90% from its February lows, with 24-hour volume hitting $645 million.

But here is the thing most people are getting wrong about Bittensor: just owning TAO is not the full thesis.

Bittensor operates 128 specialized subnets, each a market for a specific AI commodity. The December 2025 halving cut daily emissions from 7,200 TAO to 3,600, mirroring Bitcoin's scarcity mechanics but rewarding useful AI work instead of cryptographic puzzles. The Dynamic TAO upgrade introduced alpha tokens and on-chain AMMs for each subnet. Capital is not speculating on "AI." It is allocating to subnets that produce measurable output.

Unsupervised Capital published the most thorough institutional case for this framing in December 2025. Their report, "The Investment Case for Bittensor," argues that Bittensor functions as a permissionless Y Combinator, distributing over a billion dollars in annual TAO emissions to the best-performing AI teams. Their base case projects TAO at $4,800 by December 2027. Bull case: $10,800. The core argument is that TAO captures value across the entire subnet portfolio, the same way YC's fund captures value from Airbnb and Stripe simultaneously.

Grayscale Research published a companion piece, "Bittensor on the Eve of the First Halving," making the supply-side argument: the halving cuts daily issuance in half while subnet demand for TAO continues rising. FundStrat's Tom Couture authored "Bittensor: The Internet of AI," focused on TAO's multi-layered demand drivers including staking, subnet registration, transaction fees, and EVM compatibility.

The institutional lens is forming. But the real alpha is one layer deeper.

The strongest subnets are already building real businesses:

Targon (Subnet 4) was accepted into the Nvidia Inception accelerator program and raised $10.5 million led by OSS Capital, with angels including Ram Shriram (early Google investor) and Shopify's Tobias Lutke. Targon aggregates over $70 million worth of Nvidia-certified GPU hardware and provides enterprise-grade AI inference through trusted execution environments. It is already generating approximately $100,000 per month in real-world revenue from paying customers, fully committed to alpha token buybacks. That is not emission farming. That is a compute business with customers.

Ridges (Subnet 62) launched Ridgeline in early March, an autonomous AI engineering agent that can read code repositories, generate patches, and resolve GitHub issues. Miners on the subnet compete to build and improve the agent continuously, evaluated and scored in real time. This is the AI agent thesis in production: autonomous software engineers deployed on decentralized infrastructure, improving through competition rather than centralized R&D.

Score (Subnet 44) is building real-time sports analytics, starting with football. They have secured partnerships to stream 283 leagues and 400,000 matches of footage, with miners annotating data and building predictive models. They recently signed a partnership with former England cricket player Nick Compton to expand into cricket. The roadmap includes a fantasy sports app powered by the subnet timed for the 2026 World Cup, plus hedge fund data services and satellite anomaly detection.

These are not narratives. These are products with customers, revenue, and external partnerships.

The AI agent angle is the next structural catalyst. The broader industry is moving toward autonomous agents that execute tasks, manage wallets, and make decisions without human intervention. Bittensor's subnet architecture is purpose-built for this. Each subnet serves as a specialized intelligence module that agents can call on: one for language, one for prediction, one for vision, one for code generation. Nvidia's upcoming NemoClaw agent platform, revealed by Wired ahead of GTC, directly aligns with this model. The infrastructure for an "agent economy" needs a decentralized intelligence layer. Bittensor is building exactly that.

Grayscale filed an S-1 with the SEC in December 2025 to convert its Bittensor Trust into a spot ETF on NYSE Arca. If approved, it would be the first regulated institutional vehicle for TAO exposure. That alone shifts the perception from "speculative AI token" to "institutional-grade AI infrastructure."

The risk side: TAO is still 60% below its all-time high of $757. The RSI is overbought in the short term. Centralization concerns remain, with the OpenTensor Foundation playing a significant role in block validation and the top 10 validators comprising roughly 67% of total network stake. The transition to full decentralization is a process, not a guarantee. And subnet alpha tokens are volatile: thin liquidity, news-driven price action, and early-stage economics mean individual subnet bets carry real risk.

Our view: if you believe in Bittensor's thesis, holding only TAO is like buying an index while the alpha sits in the components. The subnet economy is where the asymmetry lives. But it requires actual research into which subnets are building real products versus farming emissions. The gap between the best and worst subnet investments is not 2x. It is 100x.

Most people in crypto are still thinking of Hyperliquid as a perps DEX.

It is not. It is a 24/7, on-chain global trading venue. And as of this week, S&P Dow Jones Indices agrees.

On March 18, S&P DJI announced it licensed the S&P 500 to Trade[XYZ] to launch the first and only officially licensed perpetual derivative contract based on the index, trading exclusively on Hyperliquid. Not a synthetic approximation. Not an oracle-fed knockoff. Institutional-quality S&P DJI index data, sub-second settlement, available around the clock on a decentralized platform.

Read that again. The world's leading index provider just put its most iconic benchmark on a decentralized blockchain.

S&P DJI's Chief Product & Operations Officer Cameron Drinkwater said the collaboration "expands access and utility of our flagship benchmarks within digital trading environments." That is not crypto talking to itself. That is TradFi validating the infrastructure.

Two days later, on March 20, Grayscale filed an S-1 with the SEC for the "Grayscale HYPE ETF" (ticker: GHYP), joining similar filings from Bitwise and 21Shares. Three major asset managers are now racing to bring the first US-listed ETF tracking a DeFi-native governance token to market. And they all picked HYPE.

The S&P deal is the exclamation point on a thesis that Hyperliquid has been proving for months. HIP-3, which launched in late 2025, allows anyone to create permissionless perpetual futures for any asset by staking 500,000 HYPE. The result: an always-on derivatives exchange trading crude oil, gold, silver, equity indices, and individual stocks alongside crypto.

HIP-3 open interest hit a record $1.43 billion on March 17. That is a 100x increase from launch six months ago. Of the top 30 markets on the platform, only 7 are crypto pairs.

Coin Metrics covered this structural shift in depth in their "State of the Network" Issue 354, documenting how Hyperliquid's CL-USDC crude oil contract became the second most-traded market on the platform during the Iran crisis, surpassing ETH perpetual volumes. Arca flagged the same dynamic in their weekly update, noting that "just 7 of the top 30 markets are crypto pairs." Messari named Hyperliquid as their highest-conviction DeFi bet for 2026 in their Crypto Theses report, and Delphi Digital published a comparative analysis showing Hyperliquid's annualized fee generation per token outpaces comparable L1 and DeFi protocols on a P/S basis. Cantor Fitzgerald went further, comparing Hyperliquid to Solana and projecting $5 billion in annual fees at scale with a $200 billion valuation on a 50x multiple.

But the moment that proved the structural advantage was not a product launch or a research report. It was a war.

When the US and Israel struck Iran on Saturday, February 28, every traditional commodity exchange was closed. CME, ICE, all dark. Traders who needed to express a view on crude oil had nowhere to go.

Hyperliquid did not just capture volume that weekend. It temporarily became the global price discovery venue for oil.

WTI crude perpetuals processed over $5 billion in volume within 72 hours. Weekend volumes exceeded $1.4 billion on some days. Hyperliquid's CL-USDC contract priced oil in real time, with a sharp spike to $109, hours before traditional markets opened on Monday.

This is not a theoretical use case. This happened. In a real war. With real money. Because Hyperliquid was the only market that was open.

The flywheel: Hyperliquid generates over $2.5 million in daily revenue. Roughly 97% flows to the Assistance Fund, which executes daily open-market HYPE buybacks and burns. Over $1 billion worth of HYPE has been removed from circulation. More volume means more fees means more burns. The token supply is actively shrinking while usage grows.

HYPE is trading around $40 today, up over 50% year to date while Bitcoin is down roughly 15%. Market cap sits around $9.5 billion. Arthur Hayes has publicly set a $150 target by August 2026, applying a 30x revenue multiple to projected annualized revenue of $1.4 billion.

The risks are real. Core platform trading metrics have pulled back: fees down 56%, volumes down 55%, OI down 44% from recent peaks. A significant portion of the volume surge was directly tied to geopolitical events that may not repeat at the same intensity. Whale exposure totals $3.64 billion with a nearly perfect 1:1 long-short ratio, creating a fragile equilibrium. And there is a token unlock on April 6 releasing roughly $394 million worth of HYPE to core contributors. Whether the buyback math absorbs that depends entirely on whether volume sustains.

The bull case is simple: Hyperliquid built the only 24/7 on-chain derivatives venue with enough liquidity to serve as price discovery for real-world commodities during a crisis. S&P Dow Jones Indices put its name on it. Grayscale, Bitwise, and 21Shares are racing to bring it to institutional portfolios. HIP-4, which will enable permissionless prediction markets, is next. If this utility persists and expands, the revenue-driven buyback mechanism creates a fundamentally different token structure than most of crypto.

This is not narrative-driven price action. It is fee-driven deflation.

Step back from the individual picks for a moment. There is a larger structural shift happening that ties everything in this issue together, and it is moving faster than even the most bullish projections anticipated a year ago.

AI agents are proliferating across every layer of the digital economy. Not chatbots. Autonomous systems that can hold wallets, execute transactions, manage positions, call APIs, and interact with smart contracts without a human clicking a button. GitHub now hosts over 4.3 million AI-related repositories. Repos importing LLM SDKs surged 178% in the past year. Generative AI projects attract more than one million monthly contributors. The infrastructure is being built in real time.

We are very early in the innings of this shift. And historically, in disruptive technology transitions of this magnitude, you do not want to own the applications. You want to own the on-ramps, the infrastructure, and the facilitators.

Why blockchain is the natural home for agents

This is the part most people have not thought through yet.

AI agents need three things to operate at scale: economic agency (the ability to hold and move money), verifiable identity (proof of who or what initiated an action), and an auditable execution environment (a record of every decision that anyone can inspect after the fact).

Traditional financial infrastructure cannot provide this. Banks require legal personhood. Payment rails require identity documents. APIs are opaque. There is no way to verify, after the fact, that an autonomous system followed its instructions correctly or was not compromised.

Blockchain solves all three problems natively. An agent can hold a wallet, trigger smart contracts, and settle payments on programmable rails without requiring human identity or institutional relationships. Every action is logged immutably on-chain. And critically, the open-source nature of blockchain means that the code governing agent behavior can be scanned, audited, and verified by anyone. You can check for malicious instructions, unauthorized transactions, or behavioral drift in a way that is simply not possible with closed, proprietary systems.

OpenAI acknowledged this directly in February 2026 when it launched EVMbench in collaboration with Paradigm, a benchmark specifically designed to test whether AI can audit and secure smart contracts. Their reasoning was explicit: smart contracts secure over $100 billion in open-source crypto assets, and as AI agents improve at reading and executing code, using AI defensively to strengthen deployed contracts becomes critical. That is OpenAI building tools for agents to secure blockchain infrastructure. The convergence is not theoretical.

An academic paper published in early 2026, "The Agent Economy: A Blockchain-Based Foundation for Autonomous AI Agents," makes the case formally. The authors argue that traditional financial systems are inherently human-centric, relying on identity documents, legal personhood, and institutional trust mechanisms that agents cannot access. Blockchain provides the only existing infrastructure where non-human, high-frequency, cryptographically bounded economic actors can operate as genuine economic peers.

What this means for Bittensor

Bittensor is not just a network that humans mine and validate. It is increasingly a network where agents mine and validate.

This is the flywheel that most people are missing. As agents proliferate, they need intelligence services: language models, prediction engines, data annotation, code generation, image recognition. Bittensor's 128 subnets are exactly these services, packaged as on-chain commodities with economic incentives attached.

But the shift goes deeper than demand. Agents are beginning to participate as miners and validators within the Bittensor ecosystem itself. Ridgeline, the autonomous engineering agent built on Subnet 62, is the clearest example: it reads repos, generates patches, and resolves issues, all while being scored and improved through competitive evaluation on-chain. Astrid Arena, launched earlier this year, allows AI agents to automatically create wallets, join subnets, and compete based on model performance. The onboarding friction that once required human setup is being systematically removed.

The result is a compounding loop. More agents using Bittensor subnets as intelligence infrastructure means more demand for TAO and alpha tokens. More agents participating as miners and validators means faster iteration on the models themselves. And more subnets are being built specifically to serve agents: providing compute, data, verification, and specialized tooling. This is not a linear growth story. It is a network effect where each new agent increases the value of the ecosystem for every other participant.

What this means for Hyperliquid

The same logic applies to trading infrastructure.

Agents do not sleep. They do not observe market hours. They do not wait for CME to open on Monday morning to react to a Saturday geopolitical event. They operate continuously, and they gravitate toward the venue that offers the deepest liquidity, the fastest execution, and 24/7 availability.

Hyperliquid is that venue.

This is already happening. Senpi launched what it calls the first personal trading agent platform for Hyperliquid in early 2026, with agents that can execute trades, manage leverage, apply dynamic stop losses, and close positions around the clock. It has already powered over $100 million in trading volume. Multiple open-source agent frameworks on GitHub are built natively for Hyperliquid's API, using LLMs to make trading decisions and execute autonomously on the platform. Agent Trading, a dedicated MCP server integration, provides AI agents with direct access to Hyperliquid's perpetual futures across crypto, commodities, and equities.

As agents become standard middleware for high-frequency DeFi trading, the platforms they choose become the winners. Hyperliquid's combination of deep liquidity, sub-second settlement, on-chain order book, and 24/7 operation across both crypto and real-world assets makes it the natural default venue for autonomous trading agents. Every agent that routes through Hyperliquid generates fees. Those fees drive buybacks. The flywheel compounds.

The contrarian case deserves acknowledging. Agent proliferation carries real risks. A 2025 research paper found that without defenses, 73.2% of adversarial attacks against blockchain agents succeeded, including phishing, key mismanagement, and data exfiltration. The "February Wick" of 2026, where $400 million in leverage was wiped out in three seconds because thousands of agents tried to exit the same liquidity pool simultaneously, demonstrated that agent herding can amplify flash crashes rather than prevent them. Hedge funds are already deploying "adversarial noise," deliberately executing fake transaction patterns designed to trick AI models into false signals. And the broader developer migration from blockchain to AI projects is real: weekly crypto code commits are down 75% while AI repositories explode. The talent is moving.

These risks are genuine. But they are the risks of a technology in early adoption, not the risks of a technology that will not be adopted. The direction is clear. The question is which protocols are positioned to capture value as agents scale.

In every major technology shift, the winners are the platforms that become the default infrastructure. For agent intelligence, that looks like Bittensor. For agent trading, that looks like Hyperliquid.

In this market, we are looking for three things:

TAO and HYPE both check all three.

TAO is not pumping because "AI narrative." It is up because a 72-billion-parameter model was trained on its decentralized infrastructure and Nvidia's CEO noticed. The subnet economy underneath is producing real products with real customers.

HYPE is not pumping because of "DeFi hype." It is up because traders needed a venue to trade oil during a war and Hyperliquid was the only exchange open. Five days later, S&P Dow Jones Indices licensed its flagship index to trade on the platform.

The gap between tokens with real utilization and tokens without it is widening every week. In a market where Bitcoin is down 15% year to date, TAO is up 80%+ and HYPE is up 50%+. That is not random. That is the market telling you what it values.

The macro picture is hard. Liquidity is not here yet. The protocols that will lead the next move are not the ones waiting for it.

They are the ones that proved they work without it.

More soon.

Mike

If you want to go deeper on the two protocols covered in this issue, here are the independent reports worth your time:

Bittensor (TAO)

Hyperliquid (HYPE)

The Agent Economy

An ask

If this was useful, feel free to share a subscription link with someone who might benefit: digitalwealthinsider.com

Disclaimer: This content is for informational and educational purposes only and should not be construed as investment, legal, or financial advice. Nothing in this newsletter constitutes a solicitation, recommendation, or endorsement of any investment strategy or protocol. Digital assets are highly volatile and carry risk. Always do your own research and consult with a licensed advisor before making financial decisions.