NVK returns to break down the fiat tax farm, the ARCA data haven he's building at Coinkite, why UBI ends humanity, and how AI hands the unreasonable man the power of a thousand suns.
↓ Jump to the video and timestamps
NVK came back on the show and we picked up right where we always do, somewhere between hardware paranoia, monetary philosophy, and a shared conviction that the people running the fiat system are not confused about what they're doing. They know exactly what they're doing. You are the cow. The tax code is the milking machine. The only question is whether you're going to keep standing in the pen.
This conversation covered a lot of ground: why Bitcoin Twitter is boring and why that's actually the most bullish signal of the cycle, what NVK is building with ARCA, where AI tools are right now for people who aren't developers, why UBI would be the end of humanity, and why democracy without hard money constraints just degenerates into an open-air debt farm. NVK is one of the sharpest people I talk to. He builds hardware that assumes adversarial governments. He thinks in systems. And he has a way of saying uncomfortable things plainly that I find refreshing every time.
Here's what we got into.
NVK put it plainly: we're all tax cows. The fiat system's operators figured out centuries ago that the most efficient extraction mechanism is a monetary system that forces productive people to fund the whole operation while they play ponies in the stock market.
What makes it elegant, in a dark way, is how it compounds. NVK pointed out that the fiat market (global debt, derivatives, all of it) has grown to a scale where gold, which ran its 5,000-year run as the dominant monetary good, sits at roughly $33 trillion in market cap against a fiat edifice that dwarfs it by orders of magnitude. They obliterated gold in half a century. You have to give it to them. It was an impressive operation.
The income tax follows the same pattern. It came in with the lowest bracket sitting at 1%, not a top-bracket confiscation but a bottom-bracket entry point that bled upward over time into the permanent extraction machine it is now. Marty framed it directly: "When you consider what it was originally supposed to be, what it is now..." He didn't need to finish the sentence. The founding fathers would not have recognized what came after.
My line on it: the debt farmers need your tax dollars. Ukrainians aren't buying yachts for Monaco Formula One appearances on their own dime. The whole machine runs on your labor being converted into someone else's lifestyle, and the mechanism is tax policy dressed up as civic virtue.
Bitcoin is the first tool in a long time that the debt farmers genuinely cannot get their hands around. It has a fixed supply, no central issuer, and no override mechanism. That's why NVK's framing (that in a world where every central bank is tripping over itself to devalue its currency, Bitcoin is the victor among fiat alternatives) is correct and, honestly, conservative.
Nobody is having fun on Bitcoin Twitter right now. NVK's diagnosis: there are no memes. There were no memes on the pump, which is why there wasn't a proper pump. The laser-eye era, whatever its flaws, was at least a cultural signal. This cycle produced nothing equivalent, and the discourse has devolved into people knife-fighting over BIP110 and screaming at each other about things the market does not care about.
NVK's read on why: Bitcoin won. You can only be counterculture when you are counter to the culture. When pension funds start buying your asset, the revolution is over. You've been absorbed. That's not a failure. That's a victory. But victories feel strange to people who organized their identity around being outside the system.
I pushed back a little. I don't think the broader culture is actually on board with Bitcoin. I think we're in a weird interstitial period where everyone's bored and everyone's worried nobody else is going to care. NVK's answer was interesting. He thinks American hyper-productivity right now is temporarily counteracting the fiat disaster at the level of daily felt experience. If the world isn't visibly ending economically, people are less interested in the exit.
Which is fine. The flat bottom of the S-curve is where infrastructure gets built. And the real work right now is happening quietly. What Calle and the other developers are doing with Cashu, what the Ark team is building, that's moving. The noise on Twitter is not the signal.
The MSTR dynamic is part of the weird vibe too. NVK's view is that Saylor stacked as many sats as possible, which is internally consistent, but the dilution of the common stock pulled too many Bitcoiners into an overexposed position that then went underwater. A lot of people who bought MSTR post-split are in the red. That killed some of the cycle's energy.
Bear markets are always uncomfortable. But the 2015 sentiment comparison is the right one. This is what the flat bottom feels like. You put your head down, you build, you bring your head back up when the cycle turns.
This is the thing I'm most excited about from the conversation.
The origin of ARCA is simple: Cold Card solved the seed-storage problem. But there are all the other secrets. Passwords. SSH keys. Inheritance letters. Agent API keys. A backup of your phone wallet seed. The kind of data that should not live in iCloud, not because Apple is sloppy but because iCloud is not safe from governments and it's not safe from backdoors.
That category of problem had no consumer-grade answer. NVK is building one.
The product, as he described it: a steel box running FreeBSD. They ported FreeBSD to the RockChip themselves, which NVK says was a lot of work but was non-negotiable because FreeBSD is the right foundation for a hardened network device. On top of that they added TrustZone chips for the heavy-lifting security work, tamper-detection circuitry, a soldered-down battery for the secure clock, and a motion sensor. If someone moves the box, it kills the keys. That's Cold Card logic applied to the network layer.
The user experience: you plug it in near your router, connect it to your network, and a QR code initiates enrollment. From there you get a vault for passwords (they're still deciding between Bitwarden or Vaultwarden as the server), a place for seeds from your phone wallets, SSH key storage, and agent key management. It runs Tor so you can access it from anywhere without punching holes in your home network in ways you'd regret. They're also looking at Tailscale support.
The redundancy model is what separates this from anything else on the market. You can sync multiple ARCA boxes, one in your house, one at your mother's, one in a different jurisdiction entirely. They authenticate each other with a key exchange and then sync over Tor in encrypted tunnels. Neither box sends plaintext to the other. If one goes down, you fall back to the others.
Flag theory, applied to your data.
The ultimate recovery mechanism is a BIP39 seed phrase. Per-user, per-tenant. Each person on the device has their own user space with their own BIP39 backup. That's the break-glass.
The inheritance piece deserves its own mention because NVK clearly thought hard about it. You can configure a deadman switch: set a time window and conditions, and if you don't check in, a key you left somewhere (safe deposit box, wherever) becomes valid and your family gets access to the secrets you left for them, including instructions, account information, everything. That's a problem that currently has no good consumer answer. People either have nothing or they have a janky manual system held together with hope.
On AI agent use: NVK has extended his envchain-xtra repo to inject keys from your key storage into the prompt, scoped by agent profile. ARCA takes that further. You can enroll new agent SSH keys directly on the device, so your agent has only the access you explicitly granted and can use ARCA as an SSH hop to reach other machines. Agents shouldn't have access to everything. The way most people run agents right now is, in NVK's word, scary. ARCA is the tool that puts a fence around that.
Ship timeline: NVK said if the stars align, this year, but he was explicit that the margin of error is three to four months in either direction. Prototype boxes exist, production test boards are ordered, tooling is in motion. First batch will likely be limited quantities. If you want to be in it, follow ARCA Safes on Twitter or arcasafes.com and put your email in.
NVK's framing here hit hard: every morning he stares at his laptop with anxiety he didn't have before. Not dread, more like the overwhelm of standing in front of infinite possibility. He called it the power of a thousand suns at your fingertips. What do I build today?
His current stack lives at learntoprompt.org, which he keeps current publicly, and that's generous. He sandboxes LLMs on his systems, uses a separate set of machines with slightly more permissions, and relies on AI heavily for code review, web GUI work, contract review, and research. He doesn't use self-agentic tools like full autonomous loops because, as a technical person, he finds they produce too much junk that costs more time to fix than they save. But he's explicit that for non-technical people, those extra layers of agents talking to other agents are exactly what makes the whole thing accessible.
My own setup is the proof point here. I run TFTC on voice-to-text only. I don't type prompts. I don't use CLI. The whole operation (graphics, push, transcripts, research) runs on agents I can direct with my voice.
A media company. If I can do that, the barrier really is just getting started.
NVK's advice to the 35-55 TFTC audience asking how to get in: just start building. Free accounts are still capable. Your imagination is the limit. He used the Bitcoin comparison directly: it's like Bitcoin at $60,000. It feels late. It isn't. You are at the flat bottom of the S-curve for this technology too.
The gray-beard observation is the one I keep coming back to. Developers who were already 100x engineers are becoming 1000x engineers. Even if all they do is use the AI to confirm obscure things they already know, the speed compounds.
The people who will be left behind aren't the ones who started late. They're the ones who never start.
One company worth watching in the open-source model space: NVK mentioned that high-quality developers are starting to pay attention to the open models in a way they weren't before. When people at that level start optimizing the open-source stack, things move faster. DeepSeek is already impressive for local inference, even if it's slower than frontier. The escape hatch from depending on commercial APIs is getting more viable every month.
On Dario Amodei and the regulation-seeking AI companies: I agree with NVK completely. The fear-mongering, the begging for regulation, the "we're going to take all your jobs" messaging, it's dumbfounding. It's the same play Coinbase ran when they went to Congress asking to be regulated, because they wanted to do things that would otherwise land them in legal trouble.
The answer in both cases is the same: just don't do the stupid thing. You don't need regulatory capture to avoid liability if you're not trying to do something you shouldn't. NVK's formulation: Dario somehow makes Sam Altman look good. That takes real effort.
NVK tweeted recently that there is no retirement in nature. Animals don't stop competing for resources. If they stop, they starve or their brain rots. That's not a bug in the design of living systems. It's the feature.
The competition for the banana is what keeps the capacity sharp.
His UBI argument follows directly from that: if you remove the requirement to do something for resources, you remove the drive that gives life meaning. The argument is biological before it's economic. Humans need the mild existential pressure of having to produce something the world values. Without that, you get the Walmart scooter people. Not a cruel joke, but an honest preview of what happens when the pressure is fully removed: dependency, disconnection, atrophy.
I co-sign this completely. We already have the preview. The welfare state, as it currently exists in Western countries, is a partial UBI. And we can already see the effects in the populations it has touched longest and hardest.
It's not that those people are bad. It's that the incentive structure removed the thing that generates the behavior we're calling their absence of drive. You can argue about the politics of how we got there, but the outcome is what it is.
NVK added a point I hadn't thought about in exactly these terms: there probably won't be civil wars in Western countries anymore because people are too fat, too tired, and too comfortable in their basements to organize one. The population that might once have constituted a critical mass for a cultural pushback has been de-energized. Soft countries don't revolt. They just slowly accept whatever the government decides, including ID mandates, financial surveillance, and the criminalization of political opposition.
Canada's trucker crackdown was NVK's example. He characterized a large portion of the Canadian public as supporting it. Whether or not you agree with the exact numbers, the directional observation is hard to argue. The truckers were widely framed as the villains by the people whose opinion is broadcast most loudly. A country where that framing wins has already lost the cultural capacity for serious resistance.
The US is different, and NVK is direct about why: the Second Amendment serves primarily as an anti-tyranny provision, not just a self-defense right. That's basically unimaginable to the rest of the world. Most countries don't have it written down anywhere that citizens have the right to resist a government that becomes tyrannical.
The US does. Whether that right is actually exercisable given how the government defines its opposition is another question. But the fact that it exists on paper at all is, in NVK's read, the last meaningful institutional constraint against the slide.
NVK's read on democracy is that it feels like it's in its final innings, and his fear is that whatever comes next is worse, not better.
The structural problem: democracy without hard money constraints is just efficient flip-flopping. A sufficiently well-run parliamentary system can change everything with a single majority. Canada's Charter of Rights has a literal notwithstanding clause (Section 33 of the Charter). It effectively means constitutional rights apply unless the government decides they don't. That's not a constitution. That's a suggestion.
The Quebec 1995 referendum result was 50.58% No, 49.42% Yes. NVK called it totally stolen, which is his characterization, not a legal finding. But the deeper point stands: a country that nearly split on a one-point margin, held together by a constitution that can be overridden, is not operating from a stable institutional foundation.
NVK's positive formulation: democracy without central banking is actually very palatable. If you have a genuine democratic republic with hard money constraints (a fixed supply that can't be printed to fund whatever the current majority wants), you can't do the asymmetric damage. The problem isn't voting. The problem is that fiat money lets the winners of each election borrow against the future indefinitely.
That's why Bitcoin matters here beyond the monetary argument. Bitcoin is a constraint on state power that doesn't require winning an election or appointing the right judges. It exists outside the system that can be rewritten. Open source software is the same logic applied to code. You can build things that no government can tax or shut down because there's no central entity to compel.
NVK's closing call: stop talking about politics. Go build. Productivity can't be stopped. Open source can't be taxed. The unreasonable man now has all the tools.
ARCA is a hardware data haven NVK is building at Coinkite. It's a steel box running FreeBSD on a RockChip processor, secured with TrustZone chips, tamper-detection circuitry, and an accelerometer that kills keys if the device is moved without authorization. You plug it into your home network, enroll via a QR code on its touch screen, and get a vault for passwords, seeds, SSH keys, and agent credentials. It connects back to you over Tor and can sync with other ARCA boxes you control across different physical locations. The ultimate recovery mechanism is a BIP39 seed phrase per user. NVK says if the stars align, it ships this year, with a three-to-four month margin of error on that timeline.
A NAS is a Linux box running on a commodity CPU with no meaningful hardware security. It backs up your photos. ARCA is built to Cold Card standards of paranoid security: TrustZone secure enclaves, tamper-detection circuits, a soldered-down battery for the secure clock, and an accelerometer that triggers key destruction if the box is physically moved. It runs FreeBSD, not Linux. It handles agent SSH keys, BIP39 seed recovery, multi-jurisdiction sync over encrypted Tor tunnels, and a deadman-switch inheritance mechanism. The closest commercial equivalent would be an enterprise HSM costing tens of thousands of dollars that requires specialized knowledge to operate. ARCA is designed for Bitcoiners who already understand Cold Card.
NVK's framing is that citizens in fiat monetary systems are effectively livestock (tax cows). Their labor is extracted through income taxes, currency debasement, and debt issuance, and the proceeds fund whoever the current government decides deserves them. Bitcoin is the first monetary tool in decades that sits outside that extraction mechanism. It has a fixed supply, no central issuer, and no override switch. The debt farmers cannot build a fence around it. That's the core of the bull case.
His argument is biological before it's economic. Animals don't retire. They compete for resources or their capacity atrophies. The mild existential pressure of needing to produce something the world values is what keeps human drive sharp. Remove that pressure entirely with a universal basic income and you remove the engine. He points to the existing welfare state as a partial preview: the populations most fully insulated from economic consequence over the longest periods show the effects. Dependency, passivity, and disconnection from the skills and agency that make a life meaningful are the result. UBI taken to its logical conclusion doesn't liberate people. It removes the thing that makes them function.
Not yet. NVK said prototype boxes exist and production tooling is in motion, but the first batch will be limited quantities. His timeline is "this year, if the stars align" with an explicit caveat of three to four months of margin in either direction. To get notified when it's available, follow ARCA Safes on Twitter or sign up at arcasafes.com. Given the expected limited first run, being on the list early matters.
NVK's advice: start with free accounts and just build something. The tools are capable enough at the free tier. His own public stack is at learntoprompt.org if you want a starting point. My personal workflow is voice-to-text only. I don't type prompts, I don't use CLI, I direct agents with my voice to run TFTC. If a media company can operate on that, the accessibility barrier is lower than most people assume. The paralysis is the only real obstacle.
You configure a time window and a set of conditions on your ARCA. If you don't check in within that window, a key you left for your family in a safe deposit box or another secure location becomes valid. That key gives them access to the secrets stored on the ARCA, including passwords, seeds, inheritance instructions, and account information. NVK described this as one of the core personal motivations for building the device. It's a well-secured, hardware-backed version of the kind of inheritance letter most Bitcoiners should have but most don't.
NVK (Rodolfo Novak) is the founder and CEO of Coinkite, the company behind the Coldcard hardware wallet and Opendime. He builds Bitcoin security hardware designed to assume adversarial governments, and he is now developing ARCA, a hardware data haven for seeds, passwords, keys, and inheritance. He publishes his AI-tooling stack publicly and keeps it current.
