The Chopping Block: Live at TOKEN2049 feat. Balaji Srinivasan & Arthur Hayes on Kazakhstan, Crypto Bunker Mode, and the L2 Shutdowns

Balaji said The Network State has moved its headquarters to Kazakhstan, while retaining an office in Singapore and considering another location. He expects Kazakhstan to become “the next El Salvador, the next UAE, the next Singapore,” and said more than 10 countries had bid to host the project. He cited Kazakhstan’s crypto law, low energy and construction costs, and room for data centers and startup cities. He put its population at about 21 million, its land area at roughly 3,000 times Singapore’s, and said it has about 40% of the world’s uranium. He also pointed to its performance in physics Olympiads as evidence of its capabilities. Arthur Hayes said he had already set himself up there.

Balaji argued that smaller countries can offer territory and legitimacy, while crypto and social networks can bring capital and connections. He said Kazakhstan has enough land for a 10,000-acre city for Hayes, and named Paraguay and Uruguay as other possibilities for people looking beyond the United States. His advice to investors was to visit countries rather than rely on media portrayals, and to treat backing a country as a capital investment rather than buying a politician.

A reported OpenAI release contained roughly 700 AI-written mathematical results, some formalized in Lean, but reportedly no cryptography results. Speakers speculated that cryptographic findings might have been withheld for responsible disclosure or because they could be damaging, while emphasizing that neither explanation was confirmed. Justin Drake urged crypto to enter “bunker mode,” freeze activity, and move significant assets to addresses that had never signed a transaction. Others questioned whether that would help if signatures were already practically breakable, and stressed that a theoretical advance is not automatically an actionable attack if it still requires too many resources. One speaker said Drake’s estimated chance of a catastrophic quantum outcome had reportedly risen from 0.1% to 5%, arguing that even a 5% risk of wiping out crypto warrants insurance.

The proposed insurance included a possible on-chain social contract to rewind a blockchain by about a week and reverse transactions if specified cryptography were compromised. Speakers said this might be more feasible on some proof-of-stake or simpler chains, but bridges, cross-chain dependencies, concentrated ownership, and damage to confidence could make a rewind difficult. They said the ecosystem was not ready for a new architecture. Tom argued that actions would reveal more than public safety statements: concern would rise if exchanges, protocols, or large holders actually moved assets on-chain. Other suggested ways to limit trust risk included relying on a government, reducing online exposure and trusting the physical world, or diversifying across chains and assets such as physical gold and Singapore dollars.

Blast and Abstract announced shutdowns in the preceding week, putting pressure on the assumption that block space is effectively unlimited and permanent. Abstract, associated with Pudgy Penguins, reportedly said running its L2 had cost tens of millions of dollars, including indexers and integrations, not just the sequencer. Speakers said indexing, transaction inclusion, listings, and oracles can all be expensive, and that launching and maintaining L2s likely destroyed capital. One described Blast and Abstract as copy-paste L2s with little technological or economic innovation; another said Robin’s approach made distribution the innovation because it brought clients and economic activity.

The speakers disagreed over whether L2s had improved Ethereum: one defended their positive externalities, while another said not all L2s produced them. One argued that Vitalik and the Ethereum Foundation loosely promoted L2s as a solution, encouraging teams to build in expectation of demand and low operating costs. A broader defense of experimentation was that failed investments can be worthwhile if successful outcomes outweigh losses; the speaker said L2 experimentation helped produce improvements and projects including Solana, Near, and stablecoins, and that a scalable L1 such as Solana was a necessary test of the L2 thesis. The discussion also noted that users often leave projects when prices fall, captured by the line “when number goes down, people leave.”

Other ideas included physical states that generate secret keys on demand and blockchains that bundle AI models with nodes, using tokens and compute for continuous self-testing and defense. Speakers cautioned that AI defense is not foolproof, depends on better prompting, and may not cover new mathematics that formal verification misses. The Flop Network’s theory is that an AI-payment currency should be directly convertible into compute, giving people a reason to hold it.