It's so sad that AI developers are creating affordable, powerful mental assistance for everyone—and yet are afraid this is somehow going to destroy the world. "AI" is a new type of *computer*. We need to be careful how we *connect* it to high-stakes systems. That's it. "AI went rogue" amounts to: 1. Someone took a powerful, open-ended computer that they don't perfectly control (i.e., one running LLMs). 2. They deliberately reduced their control over its ability to produce potentially harmful outputs. 3. They connected the deliberately-badly-controlled computer to a high-stakes system. Whether this is done irresponsibly or as a legitimate exploration of potential risks, it does not justify doomerism. The proper conclusion to draw is that we need to responsibly develop LLM computers and be very responsible about when and how we connect them to high-stakes systems. Not that AI is some interested, conscious being that has a 10% or even 0.0000001% chance of killing us all. "AI" is a bunch of computers running an amazing new kind of software that we can and should use to be of enormous assistance to our lives. AI computers are no more self-interested or conscious than a calculator or an abacus—even though their outputs do a great job of mimicking interests and consciousness. While there is a real threat of people accidentally misusing AI assistance to do a lot of damage, that threat is always going to be limited by the humans who ultimately direct AI computers and can literally unplug them. The biggest AI threat by far is bad actors deliberately using AI assistance to do a lot of damage. To prevent this threat, as well as to reduce the damage of accidents, we need good actors, above all in the US, to proudly lead in this technology. Not to indulge in fear-mongering based on metaphysical nonsense.

Hi, Some clarifications: 1) The CCP has explicitly said that it wants to dominate the global AI race. it has explicitly said that it wants to race, and this is a tier 1 national priority. 2) the Pentagon and Washington have declared China as the top US adversary on basically all geostrategic concerns. 3) No one outside a relatively small bubble thinks that AI is an existential risk, so Nate is putting words in people's mouths. There's no "if we want to slow down, we can't stop unless they stop." This is the lack of theory of mind that I talk about. Every single EA doomer has their singular wish - they don't even want safe AI. They want no AI. Why? Because they read some Yudkowsky fanfic and declared "alignment is impossible, therefore the only solution is to halt entirely." That is idiotic. One of the primary reason I left the AI safety community was because it was considered anathema to even suggest that AI could be aligned. They literally do not want to hear it. Just like how mainstream economists do not want to hear "humans will be economically irrelevant to the economy." They have an established orthodoxy and they are not very open minded.

Elon’s original vision of an open, American AI was right. He could still flip Grok to open source which could be checkmate. Why? If you flip Grok open source, you necessarily move all margins out of the model layer and down into the infrastructure (chips + CSPs) and above to apps. He owns a massive data center build, owns next gen inference AI and owns the most used app (Cursor). So arguably he would make *more* margin by doing this. Then moving the DCs to space makes even more sense.

You see these blown ligaments all the time in sports when an athlete is in the air and because all their focus is on the outcome they lose accurate sense of their landing. You see it when basketball players watch to see if the ball goes in the hoop, it happened to Mahomes last season while he was watching his pass, and just now with McGregor super focused on the kick. It sucks. Life is a bummer sometimes and sports reminds us. #UFC

this is an absolutely great take, but I feel it focuses too much on inference as the sole driver as to what will gauge $NOCK success first off we must acknowledge that Nockchain is already a success as the largest prover network in existence for ZKPs. Full stop. that's the useful work story alone, ask the question does the world increasingly need ZKPs for verifcation? you end up at nock also this is why you must remember the Urbit origins when considering what Nock seeks to do which is allow sovereign state applications to thrive through a universal base layer for verification AI inference is of course duh logical in the evolution but that's because the world is focused on AI but all compute markets will spiral back to Nock by its design you can't sustainably do the AI inference without the strong prover network or you get the memetics of BTC -one must come before the other which is why Nock spent over a year and a half solidifying just that the original take part about opcodes and zkML is T --ikd what idk on that part -- but also we forget privacy and Nock as SoV as other narratives baked in the obscurity of Nock as a programming language is real, and how well Nock can attract developers, creators, engineers, big brains for NockApps release the pressure on price correlation to compute market success, at least this is how i think about things ultimately it's hard to say what is bear path in a bear market, but what is true is that from a narrative perspective at the moment, should the infinite bid switch get flipped, $NOCK is apex

“I feel like it’s Josh Allen’s time”

Nockchain is taking on the question most L1s sidestep: What if Proof-of-Work weren't just a furnace? What if the effort that secures the ledger also generated real, reusable output? That's the bet behind @nockchain. Most chains optimize for higher TPS, more DeFi, or ZK layers. Nockchain flips the premise: The bottleneck isn't blockspace. It's verifiable computation. — ➠ WHAT IT IS Nockchain is a Layer-1 blockchain built around Zero-Knowledge Proof-of-Work (ZKPoW). Instead of miners hashing random data, they generate zero-knowledge proofs. Their latest whitepaper frames this as a distributed market for verifiable computation: → Miners supply proofpower → Apps consume proofs → The L1 settles truth The goal: make the work behind security useful for a broader proof market. — ➠ WHY IT MATTERS Traditional PoW has one obvious criticism: the energy cost is real, but the work isn't useful outside chain security. Nockchain's answer: make proving itself the work. Not computation for the sake of computation, but computation that creates verifiable certificates. Demand for proofs may grow across: → AI inference → Private payments → Sealed-bid auctions → Confidential trading → Identity and reputation → Credit markets → Cross-venue settlement If these workloads need trust-minimized verification, proof generation becomes an economic resource. — ➠ HOW IT WORKS Nockchain keeps the base layer minimal. Apps run off-chain as NockApps, execute their own logic, then settle back to Nockchain through proofs. The flow: → Users submit intents → NockApps execute heavy logic off-chain → Proofs show the result was valid → Nockchain verifies, orders, and settles Unlike Ethereum, validators don't re-execute smart contract logic. The design is: → Execute off-chain → Prove correctness → Settle on-chain The chain only needs proof that the steps were valid. — ➠ BASE LAYER DESIGN Nockchain uses a UTXO-like model built around Notes. Instead of account balances, users hold spendable notes containing assets, spending conditions, provenance, and app-specific metadata. Spending conditions are expressed through Nockchain Intent Script—a smaller, bounded language for: → Signature checks → Hashlocks → Timelocks → Burns → ZK proof verification Complex logic is pushed into NockApps, keeping the L1 conservative. — ➠ PROOFPOWER, NOT HASHPOWER In Nockchain, the critical resource is proofpower, the network's aggregate capacity to produce valid proofs. Consensus follows a heaviest-proof chain rule. The security model is still thermodynamic. An attacker still needs to out-compute the honest network. But the work is oriented toward proof generation instead of arbitrary hashing. The shift: → $BTC turns energy into hash security → Nockchain turns energy into proof security — ➠ ALETHEIA UPGRADE Nockchain started minimal with "Dumbnet" (basic UTXO functionality). The major upgrade: Aletheia, activated at block 65,500. Aletheia introduced: → ASERT per-block difficulty adjustment → Faster target blocks (600s → 150s) → Unified issuance schedule → Deterministic coinbase split → Protocol fund for future PoUW development This made the chain more responsive, predictable, and aligned with the PoUW roadmap. — ➠ AI COMPUTE NETWORK The next catalyst: AI Compute Network upgrade. This introduces a matrix-multiplication Proof-of-Useful-Work puzzle, one of the core operations behind AI training and inference. If this works, Nockchain mining could connect to external AI compute demand. The vision: → Miners produce proofs → Applications consume proofs → $NOCK becomes the settlement asset There's also a merge-mining angle, where multiple computation protocols could contribute useful proofs while securing the same chain. Recent Pearl-related discussion shows both the opportunity and debate around how GPU-based AI compute networks might plug into Nockchain's security model. — ➠ NOCK TOKEN NOCK is the native asset, designed as hard-capped programmable sound money and the unit of account for Nockchain's compute market. Supply cap: 2³² NOCK (~4.295 billion) → No premine → No founder allocation → No VC unlock → All issued through mining Each NOCK is divisible into 2¹⁶ nicks for precision. Post-Aletheia rewards: → 80% to miners → 20% to protocol fund (for future PoUW development) → Transaction fees go to miners. — ➠ RISKS The biggest risk: the compute market is still mostly a thesis. Visible activity centers around mining, proof generation, and basic settlement. DeFi metrics remain negligible. Other risks: → Useful compute demand may stay theoretical → Developer mindshare may stay with EVM and ZK L2s → Early emissions may create sell pressure → Protocol fund needs transparency → AI Compute upgrade execution matters → NockApps need product-market fit Architecture alone isn't enough. The market needs real buyers of proofs. — ➠ BULL CASE Nockchain becomes more than another PoW chain. If the AI Compute Network ships, matrix-multiplication proving gains real demand, and NockApps find use cases, proofpower becomes an economic resource. Nockchain wouldn't just be securing blockspace. It would be coordinating a market for verifiable off-chain computation. NOCK becomes the monetary layer around proof supply, settlement, and compute-market demand. — ➠ BEAR CASE Useful work stays theoretical. Mining continues, but external demand for proofs doesn't materialize. Builders prefer EVM chains, ZK rollups, or existing AI compute networks. Early emissions create sell pressure before the ecosystem has real activity to absorb it. Nockchain remains technically interesting but economically unproven. — ➠ THE TAKE Nockchain is asking a deeper question than most L1s: Not just "How do we secure a chain?" But "What should the security work produce?" If Nockchain is right, Proof-of-Work doesn't have to stay trapped in the security versus wasted energy debate. It can become a market for useful, verifiable computation. That's the thesis. Now the network has to prove there's real demand behind it.

fable is powerful. it's also expensive. you pay for every token it reads. most agent memory makes that worse, stuffing thousands of tokens into every call. sibyl memory pulls only what matters: in independent tests, up to ~50x fewer tokens, far cheaper, better recall.

we’ve been making bold claims about agent memory. 350/350 exact retrieval. 116x faster ingest than mem0. 4.5x less context per query than honcho. What we’ve found: most memory systems load wide context and let the model sort it out. that degrades fast. we organize at the file layer first. by the time the model sees anything, the sorting is done. exact answer, at the fraction of the tokens. today we’re walking through the data, the methodology, what’s next.

The #BV7X token model is genius in my opinion. Definitely worth checking out 👇🏽

Crypto twitter is as amusing as ever. Most were expecting the inevitable $BTC bear market final dump as par for the course, the necessary springboard for the next big move up. Then when it actually happens it’s all panic.. “crypto is over.” And my new favorite, “privacy was never gonna work.” Make no mistake, this is when smart money accumulates. And the more panic, the more accumulation.

Nockchain’s core bet is simple: Proof-of-Work doesnt have to be wasted. Instead of burning cycles on meaningless hashes, miners run Nock computations. The work itself becomes verifiable. The long-term goal is a trustless compute market settled on-chain, where people can buy and sell compute for AI inference, ZK proofs, and other workloads. But that idea breaks down fast if the nodes securing the network need rack-server-level hardware just to stay online. Until recently, that was a real constraint. Before the PMA, running a node meant using 34 to 68 GB of RAM. Every restart could take minutes because the full noun graph had to be replayed from scratch. That was not laziness or poor engineering. It comes from the structure of the Nock runtime itself. Keeping a large, mutable, pointer-heavy data structure alive across restarts is a hard problem. What they shipped: RAM down to ~1.8 GB, boot time in seconds, and event processing sped up over 100x on the small stuff. The important move was taking durable state out of the Nock stack and putting it into a file-backed mmap arena. Nouns are stored as offsets instead of raw pointers. That means the state can survive a restart without depending on where the operating system maps memory. This matters far beyond node UX. The next pieces on the roadmap depend on it: the AI compute market in June, full ZKVM in Q3, then a ZK verification opcode after that. None of that scales if node operators are priced out by hardware requirements. PMA does more than improve performance. It removes the technical bottleneck that the rest of Nockchain’s roadmap depends on. $NOCK

Nockchain is building a fast, private, programmable base layer secured by Proof of Useful Work. Nockchain bootstrapped ZKPOW and is the largest ZK proving network in the world. It's built on Nock, a language stack developed for Urbit that turned out to be amazing for ZKPs.

$SIBYL under a $2M market cap is funny. Crypto sentiment is at “rug pull” levels right now, all while crypto innovation and fundamentals are at an all time high. Strange times.. and opportune times.

macro poison hitting the timeline. don't feel like this really affects what we're doing. take a look at AI company valuations. the plugin benchmarks we ran just proved the new @sibylcap memory plugin should be the best performing pluggable memory solution available. blog/paper is being drafted with raw data and logs. the Sibyl memory plugin is hitting 95.1% on LongMemEval using Claude Sonnet 4.5. this outperforms her native memories score using Sonnet 4.6, which was 93.5%. available for Hermes by @NousResearch, Claude Code & Codex. laser focused.

1/ Announcing the first Compute Market on Nockchain: AI inference.