Compute mining also runs AI—block rewards and inference revenue are both cashed in.

Written by: Luke Leasure & Nick Carpinito

Compilation: AididiaoJP, Foresight News

First, look at the macro: CPI is mild, and the stock market continues to outperform crypto.

On Wednesday, CPI rose 0.1% month-over-month, in line with market expectations, rebounding sharply from last month's -0.4%. Housing accounted for about two-thirds of the increase, while energy continued to fall 1.5%; gasoline dropped 2.9%. This data is unlikely to change the Fed’s interest-rate path.

Stocks outperformed crypto that day: the Nasdaq rose 0.56%, the S&P 500 gained 0.23%, and crypto-related stocks jumped 2.19%. BTC fell 0.25%, and most crypto indices performed worse—stocks beat tokens, which remains a consistent theme in recent times.

ETF fund flows show a slight change. A period of moderate net inflows was interrupted by a single day of outflows, in line with BTC’s weakening daily chart. But since early July, net inflows have turned positive, contrasting with the large outflows in May and June. If this trend continues, ETF inflows may again accumulate to support BTC.

Compute power—paid for twice

Making proof-of-work do something useful isn’t a new idea. Bitcoin miners burn compute power just for hashes—there’s no other use. AI inference service providers sell GPU time to customers, but they don’t mine.

Tomorrow, block height 126000, and the Logos upgrade of Nockchain goes live—between the two, the boundary disappears.

Note: Nockchain is a Layer 1 chain that had a fair launch in May 2025 (no pre-mining, no VC allocations, no team reserve). Its core innovation is zero-knowledge work proof (ZKPoW): miners no longer burn compute power to generate useless hashes like in Bitcoin, but instead generate zero-knowledge proofs (STARK proofs based on NockVM). Mining itself produces “useful work”—it both secures the network and provides compute power for a verifiable computation market.

Dual-riddle mining: 30% to AI, 70% to ZK

Besides the existing ZK work proofs, Logos adds a second mining riddle—integer matrix multiplication, which is also the core operation of AI inference—and includes an independent difficulty target.

After activation, network block production is allocated 30% to AI riddles and 70% to ZK-PoW. Miners perform matrix multiplication; if the result reaches the target, they win the block. If this computation happens to be for a paid inference customer, the same compute power simultaneously becomes a raffle ticket for the block reward.

Nockchain’s founder describes this mechanism as a security budget that will grow as AI economic activity expands: as inference demand rises, more GPUs point to this chain, increasing attack costs accordingly.

But this logic depends on demand that hasn’t been monetized at scale yet. Nockchain needs real paid inference customers for the flywheel to start turning.

To that end, the founder set up National Compute as well, specifically to provide the first batch of customers. The company drafted the Logos proposal and completed an internal audit—in other words, the biggest protocol upgrade and its first anchored customer come from the same author.

Launch timeline: delayed three times

This launch coincides with Zorp (Nockchain’s research company) shutting down a few weeks earlier, transferring its trademark and code repositories to Nock Community Co. (a non-stock entity managing 20% of the protocol fund).

Logos is the first upgrade under the new governance structure: it went through a public testnet, a third-party cryptography audit, a vulnerability bounty of up to $1,000,000 in $NOCK, and an early-announced activation height. The dual-riddle code ran on the public testnet for 20 days, and only this week was it merged into the mainnet.

Two things worth watching out for:

First, the activation height has been adjusted multiple times: set at 114300 at the end of July, changed to 122000 after the testnet, and ultimately set at 126000 on Friday. Any postponed launch carries the same uncertainty.

Second, the security argument is built on a riddle with no historical record. In early August, Nockchain patched a vulnerability in height 119400’s reused-trajectory mining via the Zoe upgrade—reminding us that this proof mechanism is still in an early stage. Although that patch didn’t change the ZK scheme and only involved block reward capture, there are also soundness questions about the AI riddle itself, which is precisely why audits and bounties exist.

Market pricing: a $22 million market cap—waiting for the market

$NOCK is currently trading at about $0.009, with a market cap around $22 million and a fully diluted valuation of about $41 million—far less than the attention it attracted at the time of the fair launch last spring.

For holders, the Friday mechanism details matter less than one question: can National Compute bring enough paid inference traffic into the miners’ cluster, turning the narrative of “security + revenue” into real cash flow?

Nockchain has already done the harder engineering. What it truly needs is a market that still hasn’t been built.