TFTC

Bitcoin Brief

Sup, freaks.

The Trump administration reportedly asked OpenAI to slow-roll GPT-5.6 and approve access on a customer-by-customer basis. That should set off alarm bells. Strip away the AI safety language and you can see the real shape of the story: government gatekeeping, regulatory moat-building, and a warning about what happens when frontier labs that raised mountains of capital face cheaper open-weight competitors catching up fast. The government should be nowhere near deciding which models the public can use.

Missed yesterday? Catch up here

Lead Story

The Government Has No Business Deciding Who Gets Access to GPT-5.6

The Trump administration reportedly asked OpenAI to slow-roll GPT-5.6 and give access only to government-approved partners at first. The Decoder reports that the Office of the National Cyber Director, the Office of Science and Technology Policy, and the Department of Commerce were involved in the decision, and that access is being approved on a customer-by-customer basis. Sam Altman told employees this is not OpenAI's preferred long-term launch model and that a broader release is expected in a couple of weeks.

That should set off every alarm bell in your head. As I said this morning, the US government's track record is so incredibly piss poor that it should have no say whatsoever when it comes to what models can and cannot be released. We just lived through the COVID era. The same class of people locked down small businesses, coerced medical decisions, and treated constitutional rights like privileges that could be suspended by memo. Now they want a hand on the release schedule for frontier AI models. Absolutely not.

This is where the AI safety routine starts to look like a regulatory moat. OpenAI has published a Frontier Governance Framework. Anthropic has spent years warning that AI could become transformative within the decade and needs special safety regimes. Maybe some of the concern is sincere. I am sure there are engineers inside these companies who genuinely worry about misuse. But the business incentives are impossible to ignore. These companies have raised mountains of capital. They need to make investors whole. A world where the public can run cheaper, increasingly capable open-weight models is a world where the economics of closed frontier labs get much harder.

The competition is coming fast. MiniMax says M3 is the first open-weight model to combine frontier coding, long-context agentic work, and native multimodality in one model. Its GitHub page says the model has roughly 428 billion parameters, 23 billion active parameters, and sparse attention that reduces per-token compute to 1/20 compared with standard attention. Whether you use MiniMax, DeepSeek, Qwen, Mistral, or whatever comes next, the direction is obvious: intelligence is getting commoditized and the price is getting crushed.

That is the real threat to OpenAI and Anthropic. The danger is not that normal people get access to powerful tools. The danger, from their perspective, is that normal people get access to powerful tools without paying the closed-lab toll. Once you understand that, the safety language sounds less like public-spirited caution and more like a permission structure for protecting margins.

The precedent is the thing. You may like the current administration. You may think they are trying to do the right thing. That does not matter. If this administration claims the power to decide which AI models can be released and who can use them, your political adversaries will inherit that power when they take office. Then they will use it against you. That is how every emergency power works. It starts as a temporary measure. It becomes a permanent control layer.

Draw the line now. The government should be as far away as possible from deciding who can release models, who can train models, and who can use models. If they can pressure OpenAI and Anthropic into gating frontier releases today, they will come for open-weight models tomorrow. We need more open competition, more permissionless tools, and a much smaller government that leaves us alone.

Signal

Market Structure

$10 Billion in Bitcoin Options Expire Today as BTC Hits 2026 Lows

Why it matters: Forced deleveraging is flushing out the paper bitcoin complex while on-chain data starts to look like capitulation.

Roughly $10 billion in Bitcoin options expire on Deribit today, representing a major chunk of open interest. Max pain sits at $74,000 while bitcoin trades around $59,000. More than a billion dollars in leveraged positions were liquidated over the past day, with longs taking the brunt of the pain.

The on-chain picture is ugly in the way bottoms are ugly. MVRV sits near fair value at 1.12. SOPR is below 1.0. Long-term holder SOPR is 0.66, the deepest capitulation reading since mid-2022. ETF outflows have totaled $2.92 billion in June. The leverage that we talked about yesterday is being stress-tested in real time.

Stablecoins

Circle and Nomura Partner to Bring USDC Settlement to Japan's $440 Billion-a-Day FX Market

Why it matters: Japan's largest investment bank is rebuilding cross-border corporate payments on stablecoin rails. This is the boring adoption that actually matters.

Circle and Nomura Holdings announced a partnership to launch USDC-based digital asset settlement and corporate payment services in Japan, targeting a 2027 launch. Japanese businesses will be able to swap yen for USDC for cross-border supplier payments, affiliate transfers, and FX settlements. Standard bank wires take 2-3 business days. Blockchain cuts that to minutes.

Japan's FSA has already cleared USDC under updated payment rules, making it the first global dollar stablecoin approved for local corporate use. Nomura handles onboarding and compliance. Circle operates through its local branch. This is TradFi eating its own lunch. As we have covered extensively, dollar stablecoins are becoming the de facto digital dollar, whether governments intended it or not.

Corporate Treasury

Strive Bought 759 BTC at $65,850 Last Week. Bitcoin Is Now Trading at $59,000.

Why it matters: The seventh-largest corporate Bitcoin holder is buying aggressively into the drawdown. The treasury companies are the bid when the ETFs are not.

BTC Times reported that Strive disclosed in a June 22 Form 8-K that it acquired 759 BTC between June 15 and June 21, its largest single-week purchase in months. The $50 million buy brings their total treasury to 19,864 BTC. CEO Matt Cole said the company is "aggressively buying Bitcoin as prices decline." The average cost was $65,850, which means they are already sitting on about a 10% unrealized loss. That is the conviction bet. While ETFs are bleeding $2.92 billion in outflows for the month, the treasury companies are stepping in.

Macro / Trade

Section 232 Semiconductor Tariffs Could Cut $4.4 Trillion from US GDP Over a Decade

Why it matters: The Commerce Department is due to report by July 1 on data center semiconductor usage. Phase 2 tariffs could follow.

A 25% semiconductor tariff enacted in January is already reshaping supply chains. The Information Technology and Innovation Foundation published analysis this week showing a 50% tariff scenario could reduce US GDP by a cumulative $4.4 trillion (10.3%) over a decade. Companies are diversifying sourcing to Mexico and Southeast Asia. The Commerce Department owes an update by July 1 on the market for semiconductors used in US data centers, which could trigger broader Phase 2 tariffs.

For bitcoiners, the second-order effects matter most. Higher chip costs flow directly into mining hardware prices. More expensive ASICs raise the cost of hashrate production, which compresses margins for miners already operating below production cost. Tariffs that protect domestic chip production in theory could also entrench US dominance in mining infrastructure long-term. The trade-off between short-term pain and long-term reshoring is real.

On-Chain

21Shares: Bitcoin's Cycle Is "Evolving, But Not Broken"

Why it matters: Institutional research desk says the cycle thesis holds even as ETF outflows pile up and on-chain stress deepens.

21Shares published an analysis arguing Bitcoin's four-year cycle is evolving structurally but remains intact. ETF outflows hit $2.92 billion in June and $2.34 billion in May. January and February were outflow months too. March and April saw billions in inflows, then the trend reversed. The LTH/STH SOPR ratio, which contrasts long-term and short-term holder profitability, shows divergence consistent with mid-cycle corrections in prior epochs. The report acknowledges the drawdown is more severe than typical mid-cycle pullbacks but argues the accumulation patterns underneath are constructive.

AI / Energy

SemiAnalysis: The Grid Cannot Keep Up With AI. Behind-the-Meter Power Is the Escape Hatch.

Why it matters: The AI race is running into physics. Compute needs power, and power is becoming the bottleneck that decides who can scale.

SemiAnalysis published a dense breakdown of the US grid bottleneck facing AI data centers. Their model has US data center gross power demand rising from +21GW in 2026 to +84GW by 2030, while net-new accredited grid capacity runs closer to 15GW per year today and trends toward 20GW+ late in the decade. In their view, grid headroom is already approaching zero and turns negative by 2027 across a growing set of regions.

That leaves behind-the-meter power as the obvious escape hatch. SemiAnalysis expects BTM to power well over half of new US data centers in 2028 and beyond, with the BTM equipment market crossing 50GW per year by 2029. This is where the AI story stops being software theater and starts looking like energy infrastructure. If you understand stranded power, gas, interconnection queues, ERCOT, mining economics, and load flexibility, you are staring at the same problem the hyperscalers are now forced to solve.

AI / Mindset

Jordi Visser: AI Rewards the People Who Think in Bets and Keep Moving

Why it matters: The gap in the AI age is not technical ability. It is willingness to experiment, update, and treat roadblocks as information.

Jordi Visser's latest 22V Research note makes the right point about what AI demands from people. The winners are not necessarily the youngest people, the most credentialed people, or the most technical people. They are the people willing to get stuck, ask better questions, run another attempt, and update their priors when the output teaches them something.

That is exactly the mindset the closed-lab gatekeepers and government safety bureaucrats do not want to scale. AI makes exploration cheap. You can draft, test, compare, break, rebuild, and refine faster than any institution can write a manual. The freak who treats every prompt as a wager and every bad answer as feedback is going to lap the person waiting for permission.

⚡ Freedom Tech Corner

Run Your Own Electrum Server This Weekend

Your wallet is only as private as the server it talks to.

Every time Sparrow Wallet or any other Bitcoin wallet checks your balance, it asks a server. If that server belongs to someone else, they can see your addresses, your balances, and your transaction history. Running your own Electrum server (Fulcrum or Electrs) on a home node eliminates that leak entirely.

If you already run Bitcoin Core, you are most of the way there. Sparrow's node connection guide walks through connecting Sparrow to your own Bitcoin node instead of leaking wallet queries to somebody else's server. Takes about an hour. Once it is done, your wallet talks to nobody but you.

Data Snapshot

⚡ Browse BitcoinProducts.com

If this landed, forward it to someone who could use more signal and less noise. The Bitcoin Brief is free, always will be.

See you on Monday,

Marty Bent

Follow: @MartyBent · @TFTC21

Nostr: primal.net/marty

YouTube: TFTC · Podcast: tftc.io/podcast