Bitcoin Brief

Sup, freaks.

AI is making it harder to answer a question investors have treated as basic for decades: what is this business worth ten years from now?

A moat that looked durable five years ago can now get chewed through by a model, an agent, or a tiny team shipping at machine speed. That changes the value of every cash flow sitting far out in the future.

Bitcoin does not have that problem.

Let's get into it.

LEAD STORY

AI Is Making Every Terminal Value Less Trustworthy Except Bitcoin's

Jordi Visser published a video over the weekend called The AI Crypto Macro Nexus Point: Why Bitcoin Matters to Investors. His argument is simple enough to make people uncomfortable.

AI compresses the competitive clock. Agents can write code, research markets, operate software, and attack old business models faster than human organizations can react. If disruption happens faster, the period over which an investor can confidently model a company's cash flows gets shorter. The terminal value inside a discounted cash-flow model gets shakier because the business may not survive long enough to produce what the spreadsheet assumes.

The burden of proof has changed. A brand, software workflow, or distribution advantage deserves less confidence when agents can replicate the customer experience, rebuild the product, and reach customers faster and cheaper. Even today's strongest AI labs face the risk that a new architecture or open-weight model resets the economics underneath them.

This is what makes bitcoin so interesting in an AI world.

AI can create another token, write another chain, and market the hell out of it. It cannot reproduce bitcoin's history, distributed holders, proof of work, liquidity, social consensus, or a supply schedule that has survived every attack thrown at it.

Bitcoin has no customer base to poach, margin to compress, or management team that can issue more shares, buy a bad company, or torch the balance sheet chasing the next trend. It just keeps producing blocks.

The other side of this argument is the sovereign debt system. It is also built on claims about the future. Governments borrow against future tax receipts, future growth, and future political discipline. The United States has spent decades pulling future demand into the present while promising that tomorrow's economy will be large enough to carry the bill.

Treasury's latest Debt to the Penny reading puts total federal debt above $40 trillion, and the long end of the Treasury curve keeps reminding Washington that time has a price.

Corporate terminal values are vulnerable to technological disruption. Sovereign terminal values are vulnerable to political discretion. Bitcoin sits outside both.

No model can change the issuance schedule. No Treasury secretary can order a larger block subsidy. No central bank can cut the difficulty because debt service got uncomfortable. The network adjusts according to rules that every node can verify.

That is why I think Visser's framing matters. A zero bitcoin allocation is not neutral anymore. It is a bet that AI-driven disruption will not materially shorten corporate moats, that the debt system will remain credible, and that policy makers will resist debasement when forced to choose between the currency and the political system built on top of it.

That is a hell of a parlay.

The market is beginning to act like it understands the setup. Coinbase daily closes put bitcoin roughly 13.6% above its 200-day simple moving average during this morning's final refresh. TFTC's ETF tracker shows five straight positive sessions through Friday, totaling about $1.918 billion, with roughly $2.382 billion of net inflows so far in August.

Flows do not prove the thesis, and a violent move can reverse just as violently. The useful signal is that capital is gaining a regulated path into an asset that cannot be disrupted like a company or diluted like a currency.

Meanwhile, the AI economy is building the other half of the bridge. Open models are taking token share. Agents are becoming major consumers of inference. Wallets, stablecoins, Lightning, ecash, and machine-payment systems are being built for software that will need to transact without waiting for a human to approve every move.

AI will create enormous value. It will also destroy a lot of terminal value that investors currently treat as permanent.

Bitcoin is the asset on the other side of that destruction. It does not need to predict which model wins. It does not need to know which lab survives. It does not need to trust a CEO, a government forecast, or a twenty-year cash-flow model.

It needs people to recognize that credible scarcity becomes more valuable when everything else can be copied, disrupted, or diluted.

AI is accelerating that recognition.

SIGNAL

Non-Custodial Does Not Mean Service-Independent

Blink reported Monday morning that Spark, the provider behind its non-custodial accounts, was temporarily down while an issue was investigated. Spark's status page records a major outage affecting Spark Operators from 04:30 to 08:00 UTC. Blink said service was restored at 07:48 UTC.

The incident exposed an important distinction. A wallet can remove the custodian without removing every service dependency. Users may control the keys while still depending on operators, coordinators, watch services, servers, or liquidity infrastructure to see balances and move money normally.

Blink and Spark have not published a root cause or disclosed a loss of funds. This was an availability failure, not evidence of a custody failure. Self-custody limits who can take the money. Resilient self-custody also needs credible recovery paths when the service layer disappears.

Iceberg Hides a Threshold Group Inside One Side of a Lightning Channel

IACR ePrint 2026/1757 introduces nested threshold multi-signatures and an Iceberg construction that lets one side of a Lightning channel operate as a t-of-n group while appearing to the counterparty as a normal MuSig2 participant.

The authors say this can be deployed without changing bitcoin, the Lightning protocol, or the counterparty. Their incomplete prototype integrates Iceberg into Eclair, a production Lightning implementation. In the reported four-member benchmark, three members had to remain online to sign, the group tolerated one corrupted member, and it sustained more than 93% of an unmodified endpoint's payment throughput.

That is promising research, not a production deployment. The paper does not eliminate operational key risk, replace monitoring, or turn every Lightning channel into a general multiparty channel. It shows a path for protecting one online endpoint without forcing a network-wide upgrade.

Open Models Are Taking Token Share While Agents Become AI's Biggest Customer

Vercel CEO Guillermo Rauch reported that open-weight models rose from 28.4% in his June 24 comparison to a record 61.6% of tokens routed through Vercel's AI Gateway on August 22. Gavin Baker's read is that open models can pressure the model layer without killing aggregate demand for compute.

Amjad Masad supplied the other half: agentic applications routed through OpenRouter consumed a reported 7.3 trillion tokens on the cited seven-day average, nearly five times the 1.4 trillion attributed to human applications.

These are provider-specific charts without a fully public methodology. Token share is not revenue, users, quality, or total-market share. The direction is still worth watching. Cheaper models and machine users can compress model margins while expanding demand for chips, power, inference, and payment rails.

ZEUS Hardened the Wallet After a Brutal Month for Bitcoin Security

ZEUS v13.2.0 adds 24-word seed support for LDK Node wallets, embeds LND v0.21.2-beta, upgrades its Cashu stack for newer mints, and lets users point fee, block-height, and sweep requests at a self-hosted Mempool instance.

The bigger story is the security pass. ZEUS says the release includes dozens of fixes and hardening changes covering more-complete wallet wipes, Nostr Wallet Connect permissions and budgets, expired invoices, LNURL verification, network egress, pinned hashes for native build dependencies, and reduced sensitive logging.

The work follows the project's recent infrastructure incident and independent disclosures from the Bitcoin Red Team. It does not mean ZEUS is now bug-free. It means a wallet team took the incident seriously, published a disclosure path, and tightened the surfaces that attackers actually probe. ZEUS Swaps remain indefinitely suspended.

JoinMarket NG Closed a Directory-Injection Hole and Hardened Maker Recovery

JoinMarket NG 0.37.1 shipped Sunday with a compact set of privacy and security fixes.

The release prevents malicious directory offer data from injecting HTML into the orderbook watcher. It also fixes duplicate maker notifications after restart, preserves working makers when restoring a requested CoinJoin counterparty count, rotates disclosed PoDLE commitments during maker replacement, and avoids blacklisting makers when Neutrino lookups temporarily fail.

None of this changes bitcoin consensus or makes CoinJoin trustless. JoinMarket users still face software, liquidity, network, and implementation risk. The release matters because privacy software lives or dies on the boring edges: directory data, maker coordination, restart behavior, and failure recovery. Those are exactly the places where a small bug can turn into a large leak.

⚡ FREEDOM TECH CORNER

Cashu's App Layer Is Shipping Faster Than the Narrative Around It

Cashu highlighted a batch of ecosystem releases Monday, and the useful part is not the number of version tags. It is where the stack is getting more usable.

cashu.me v0.3.0 improves mobile testing, recovery errors, accessibility, wallet startup behavior, mint validation, and receive flows. Coco v2.0.0 adds first-class BOLT12 support, on-chain mint and melt flows, payment requests over transports such as Nostr, persistent quote management, and more reliable transaction history. It is a breaking release, so integrators need to read the upgrade notes. Nutmix v0.7.0 is labeled a security release and tells operators to upgrade as soon as possible, although the public notes do not identify a specific exploited vulnerability.

Cashu still carries mint risk. Better wallet software does not make the issuer disappear, prove reserves, or guarantee redemption. What is happening is more practical. The open ecash layer is gaining the recovery, payment, merchant, storage, and operational tooling it needs to move from demos into everyday applications.

DATA SNAPSHOT

As of August 24, 2026, approximately 9:42 a.m. ET

Sources: Coinbase for spot price and daily closes; mempool.space for block, fee, hashrate, and difficulty data; TFTC Bitcoin ETF Flows for ETF data through Aug. 21.

See you tomorrow. Nothing here is investment advice. Do your own research.

YouTube: https://www.youtube.com/@TFTC

podcast: https://www.tftc.io/tag/podcasts/