The AI War Is Being Fought at Every Layer

The AI war is not coming. It is here. It is being fought over power lines, county zoning boards, model access, cyber safeguards, developer tooling, and the invisible plumbing between your machine and the model provider.

That is the thread running through a new report from the Bitcoin Policy Institute that tracks the ground game against American AI infrastructure. BPI argues that genuine local concerns about electricity prices, water use, land use, and grid strain are being amplified by a Singham-linked activist network with documented ties to organizations that routinely line up with Chinese state interests. The report tracks 21 campaigns across 14 states and argues that PSL-linked organizing has played a role in data center moratoria, bans, rejected projects, and delayed or scaled-back investment.

The legitimate concern here is electricity prices. Americans are right to notice that power bills are moving higher, and data center demand is absolutely eating into excess capacity in places where the grid was already stretched. That does not mean data centers are evil. It means America has a generation problem and an interconnection problem. The answer is not to kneecap the AI buildout. The answer is to build more power, faster.

The water panic is mostly FUD. Yes, data centers use water in certain cooling systems. No, it is not remotely the same category as thirsty agricultural use cases like almonds or avocados, and much of the cooling infrastructure is closed-loop or recirculating. The real bottleneck is generation. We need more coal plants back online where it makes sense. We need modular nuclear reactors. We need more natural gas. We need more wind, solar, batteries, transmission, substations, and interconnection capacity. The productivity gains from AI are too important to let a shortage of electrons become the excuse for letting America fall behind.

The same fight showed up at the keyboard this week. International Cyber Digest reported allegations that Claude Code silently encoded route, timezone, and possible China-proxy signals into the system prompt it sent upstream. ICD notes a clarification at the bottom from Anthropic technical staff saying the code is being rolled back. Fine. Roll it back. But the fact pattern matters either way. Anthropic may have good reasons to detect unauthorized resale, sanctions exposure, or model distillation attempts from Chinese proxy networks. That does not erase the trust problem. Developers are letting coding agents sit inside their repos, read files, inspect project structure, and in many cases run commands. Trust is the product.

Then there is Anthropic’s own Fable 5 update. The company says the June 12 export controls on Fable 5 and Mythos 5 were lifted on June 30, and Fable 5 returns globally today. Anthropic also says it worked with the government, Amazon, and other partners after a report about bypassing Fable 5 safeguards. The important detail: Anthropic trained an improved cyber classifier, and if a request to Fable 5 is blocked, users will be notified and the request will be sent to Opus 4.8 instead. That is a big guardrail on the most sensitive edge of the model, and it came out of direct coordination with the government.

We wrote last week that government gatekeeping of model access is a red line. This is the next chapter. The chokepoints are multiplying. The county board can slow the data center. The federal government can shape who gets the model. The lab can route sensitive requests through a weaker model. The client can inspect your routing path. The proxy can leak your location. The open-source model in China can keep getting better anyway. Different layer. Same war.

Freaks should pay attention because this is the map of the next decade. Money, energy, compute, jurisdiction, and trust are converging. If you do not own your money, your infrastructure, your tools, and as much of your intelligence stack as possible, you are renting your future from people who are already under pressure from governments, competitors, and foreign adversaries. Bitcoin taught us the lesson first: do not outsource sovereignty to institutions that can be coerced. AI is teaching it again, faster.

The 58k Gang Rides Again

Why it matters: Bitcoin is testing a meme level from the last cycle, but the market structure is completely different this time.

James Check’s latest Checkonchain note is the bitcoin market signal of the day. He calls it the return of the 58k Gang, a nod to the cohort that top-blasted the 2021 highs around this same level and then had to survive the 2022 bear. The difference is that $58,000 in 2021 was euphoria. $58,000 today looks more like time-pain capitulation. Live market data had bitcoin near $58,560 while I was drafting this, which makes the psychological level impossible to ignore.

Check points to waning downside momentum across several indicators, including short-term holder MVRV, short-term holder SOPR, realized losses, and RSI divergences. He also notes that much of the loss realization is coming from higher-cost buyers, while older coins are mostly sitting tight. That is how bear markets usually do their work. They bore people, punish leverage, flush recent buyers, and transfer coins to people who know what they own. None of this is financial advice. It is simply the market reminding everyone that conviction is easy near the highs and earned near the lows.

PBoC Liquidity Is Helping, But the Dollar Still Bites

Why it matters: liquidity is improving at the margin, but the global system is still fighting dollar strength and QT.

Michael Howell’s latest Capital Wars update says global liquidity has edged higher to roughly $194 trillion, helped by renewed PBoC liquidity injections and low bond-market volatility. His 3-month annualized growth estimate sits at 13.6%, which points to better near-term momentum than the annual figure alone would suggest.

The constraint is that the dollar is still doing damage. Howell flags weak Fed liquidity growth, continued ECB and BOJ QT, and dollar strength against major currencies as brakes on the expansion. This is the macro backdrop behind the bitcoin chop. China can inject. Collateral values can firm. Treasury can try to dampen volatility. But when the dollar tightens the noose, risk assets feel it. Bitcoin is not broken. It is trading inside a liquidity regime that keeps changing underneath it.

Stablecoins Are Becoming the New Banking Cartel Rail

Why it matters: the biggest names in payments and finance are moving to make stablecoins useful, compliant, and safely inside their perimeter.

TFTC flagged the Open USD announcement because the partner list tells you exactly where the dollar system is going: Visa, Mastercard, BlackRock, Stripe, Coinbase, Google, Ripple, Shopify, and more than 140 other companies are lining up behind a shared stablecoin called Open USD. Stablecoin Insider reports that OUSD charges zero mint and redeem fees with no volume caps and shares most reserve earnings with partners after a management fee. Yahoo Finance framed it as a possible threat to USDT and USDC.

Do not fall for the branding. They can call it open. They can call it consortium-governed. The obvious point is that the largest payment processors, asset managers, exchanges, and tech platforms are admitting stablecoins are useful money pipes while trying to make sure the pipe terminates inside their perimeter. Stablecoins are useful, especially for people trapped in broken currencies. But they are still dollars. They are still permissioned at the edges. They are still subject to compliance pressure. Bitcoin is bearer money. OUSD is a corporate dollar wrapper with a better distribution strategy. Useful? Sure. Sovereign? Absolutely not.

Enterprises Move From Tokenmaxxing to Token Budgeting

Why it matters: the AI spend story is shifting from novelty consumption to operating discipline.

SemiAnalysis has a useful frame on enterprise AI usage: the first half of the year was defined by tokenmaxxing, but the next phase is token budgeting. Companies pushed employees to use AI aggressively, then realized that unbounded agentic workflows can burn through budgets fast. Now they are setting limits, downgrading default models, turning off premium tiers, and forcing teams to justify spend.

The lazy take is that this means the AI boom is over. I do not buy it. It sounds more like a market growing up. The most serious usage is concentrated among power users, coding teams, data teams, and AI-native companies. Median enterprise adoption is still early. That means the next leg will not be driven by everyone asking chatbots for email drafts. It will be driven by companies that figure out where AI creates real operating leverage and where it is just expensive theater. That is exactly the distinction the Roundtable is going to keep hammering.

Jordi Visser: Signal, Alpha, and Agency in the AI Cycle

Why it matters: nobody knows exactly how this AI cycle ends, which makes humility and adaptation more valuable than certainty.

Jordi Visser’s latest 22V note is a good antidote to the confident takes flying around AI. Bulls are convinced AI will change everything. Bears are convinced the market has already gone too far. Both can sound smart. Both can build a story. But the cycle is moving faster than the human brain can comfortably model.

Jordi’s framework is Signal, Alpha, and Agency. Signal means separating what is happening from what people are saying is happening. Alpha means acting when the facts, earnings, and price confirm the thesis. Agency means keeping enough flexibility to change your mind when the evidence changes. That is a good mental model beyond investing. It applies to operators, founders, and freaks trying to build in this environment. Narratives are cheap. Reality is expensive. Respect the data.

Citizenship Is Becoming the Next Legitimacy Fight

Why it matters: the birthright citizenship ruling exposes a deeper fight over jurisdiction, allegiance, and political membership.

The Supreme Court struck down President Trump’s executive order seeking to end birthright citizenship for children born in the United States to parents who are here unlawfully or temporarily. Chief Justice Roberts wrote for the majority that those children satisfy both elements of the Citizenship Clause and are citizens at birth. AP reports that Roberts, Sotomayor, Kagan, Barrett, and Jackson formed the majority, while Thomas and Alito dissented.

The dissent is the tell. Justice Clarence Thomas wrote a 91-page dissent, joined by Justice Gorsuch. Thomas does not normally write for that long unless he is furious, and you can see why. His argument is that the Fourteenth Amendment was meant to overturn Dred Scott and guarantee citizenship to people born and domiciled here, not to create a global birth-tourism loophole for anyone who can get inside the border before giving birth.

Matt Walsh captured the cultural fault line in blunt terms. It is objectively insane that a woman can fly here on vacation while nine months pregnant, have a child on American soil, leave immediately, raise that child in a country that may not share American values, and still hand that child a future citizenship claim. That is an abuse of citizenship. The legal fight is about the Fourteenth Amendment. The political fight is about whether citizenship is merely territorial or tied to allegiance, domicile, and membership in a political community.

Run BTCPay Server for Your Business

Why it matters: if you sell anything online, you should be able to accept bitcoin without asking a payment processor for permission.

Download BTCPay Server or spin it up through a hosted deployment path and connect it to your store. The official BTCPay deployment docs walk through the options, from self-hosting to third-party hosts. The point is simple: invoices, checkout, and bitcoin payments should not depend on a company that can freeze you because a bank or politician got nervous.