メインコンテンツへスキップ

The Market Sees a Downturn. Builders See the Beginning of a New Era.

Robert Mao
BlockchainAIDecentralizationIdentityVerifiability

Yesterday I posted a brief note on X. I was simply writing down how I felt as a builder watching the mood across the market and the industry lately, and it received far more likes and reposts than usual.

So many peers resonated with that short post because our industry is going through a rare wave of deep pessimism and panic.

Over the past twelve months, the bad news has barely paused: from the severe market disruption triggered by Binance last October that broke the industry's momentum, to Binance's massive security breach this October affecting roughly $6.3 billion in assets,[1] and then the CLARITY Act falling short on a 49-to-50 procedural vote in the U.S. Senate.[2] Now, with Democrats retaking both the House and the Senate in the U.S. midterm elections, equities and crypto markets have tumbled together as participants brace for a much colder political and regulatory climate.[3]

What cuts even deeper than falling prices is the double bleeding of talent and security. Over the past two or three years, the explosion of AI has acted like a giant magnet, pulling engineers, founders, and attention out of blockchain. Those who stayed look around and see peers leaving, while attackers have been the first to weaponize AI. Hackers are already using AI to scan massive codebases, chain together compromised developer tools, and exploit permission gaps with a speed and precision we rarely saw before. Talent is being drawn away by AI while attackers march in with AI weapons. That is the real chill hanging over crypto today.

My view is the exact opposite: when panic drives everyone toward the exits and turns AI into an excuse to abandon blockchain, it is the best possible time to stay and rebuild the foundation. Far from leaving blockchain for AI, we are using AI to double down on blockchain, decentralized identity (DID), and real decentralized applications.

We Didn't Leave Blockchain for AI. We Are Using AI to Double Down on Blockchain.

Many people know that ArcBlock has invested heavily in AI since 2023, building AIGNE, AFS (Agentic File System), and a series of production AI engineering tools. Friends occasionally ask me whether we have quietly pivoted away from blockchain to become an AI company, just as so many older crypto teams have done.

We chose the opposite path. While others left blockchain to chase AI, we are taking everything we have built in AI over the past three years and pouring it back into blockchain and decentralized identity with far greater force.

Why do we have more conviction now? For too long, crypto was hijacked by speculation and quick money. Its original promise, building real decentralized applications (DApps) where users own their identity and data, was pushed aside until much of the outside world began equating crypto with scams. As I wrote in Using the CLARITY Moment to Explain ABT Clearly, real applications need delivery protection mechanisms like SubGuard (Stake for X) and verifiable service accounting ledgers like Credit Token. And the deeper we go into AI, the clearer a fundamental truth becomes: when always-on AI agents begin calling services, managing assets, and collaborating across systems on behalf of people, the digital world cannot even establish basic accountability without decentralized identity (DID) and verifiable onchain boundaries.

After Binance's major security incident at the end of October, I published Crypto Can’t Go On Like This and Attackers Are Already Using AI. Our Defenses Have to Change., calling on builders across the industry to rebuild our security foundations together.

Now that attackers are using AI, defenders and builders cannot remain in the handicraft era. Starting this year, our entire company shifted to multi-agent collaborative software engineering. I have been writing code for more than thirty years, and honestly, the pace our team is moving at this year looks surreal even to me 🤣. Protocol overhauls, cross-chain rewrites, and native multi-platform rendering that used to take a dozen engineers a year or two can now be designed, implemented, red-teamed, and regression-tested in days by a handful of engineers orchestrating more than twenty specialized coding agents. AI never gave us a reason to leave blockchain. Instead, it handed a focused engineering team a sledgehammer heavy enough to tackle foundational problems we would have hesitated to touch in the past.

Hardcore Upgrades Powered by AI: Rebuilding a Chain for the Agentic Era

Saying we are doubling down is not a slogan. You can see what AI-powered engineering makes possible by looking at the foundational upgrades we have been shipping across our new blockchain stack.

The first major upgrade is a complete overhaul of our core transaction pipeline (VNext) and Composable TransactionBlock execution. We retired years of accumulated legacy baggage, archiving the old V1/V2/V3 write paths and the heavy RollupBlock state machine in favor of a domain-agnostic VNext protocol. At the same time, we introduced an in-memory snapshot engine (OverlayStateDB) paired with two-phase commit (2PC), allowing multi-step asset operations, ContractNFTs, and cross-protocol calls to execute atomically within strict resource and permission boundaries. If any step in the sequence fails, the entire block rolls back cleanly without leaving partial state behind.

The second major upgrade is a ground-up rewrite of our ARC-native stateless bridge and a "Published Indirect Multi-Sig" cross-chain control plane. Historically, running a cross-chain bridge meant maintaining clusters of always-on servers and periodically pushing Merkle roots to Ethereum contracts just to amortize gas, an architecture that was clunky and prone to single points of failure. We rewrote our bridge from scratch on top of our ARC runtime architecture: bridge nodes can now run directly on Cloudflare edge infrastructure (Workers and Durable Objects), using standard EIP-712 typed structured signatures for zero-prewrite verification on destination contracts. Going a step further, we turned our own OCAP blockchain into a security control plane capable of governing external treasuries on networks like Ethereum and Base. Signers complete public governance, voting, and cooldown verification on the OCAP chain to produce a cryptographic certificate with human-readable parameter previews (Clear-Signing, eliminating blind-signing traps), which then drives external contract execution with two-stage release gates for large transfers.

The third major upgrade is native 3-Tier Scoped Delegation for AI agents and verifiable state proofs. In an agentic world, no sensible user will hand a master private key directly to an AI agent. At the base layer of our chain, we implemented a fine-grained three-tier scoped delegation protocol, a unified Token-as-an-Asset model, and our ArcSpace architecture (verifiable checkpoints onchain paired with payload storage in the user's private space). An AI agent can act on its owner's behalf only within explicitly declared operation categories, spending caps, and expiration windows, and every ledger entry carries an inclusion proof that the client can verify independently.

What excites me most is that we have welded Multi-Agent Red-Team Adversarial Review directly into our blockchain engineering pipeline. Whenever an engineer and a coding agent finish a cross-chain multi-sig or state-machine protocol, our workflow immediately spins up three or more independent-context Red-Team agents. They attack the implementation concurrently from angles including blind-signing semantic deception, cross-chain replay, signer-rotation lockouts, and proposal collusion, turning every edge case they uncover into an automated regression test. If attackers are using AI to hunt for vulnerabilities, we use multi-agent offense and defense to hammer our defenses ten times over before code ever ships.

Next-Generation ArcWallet and ArcSphere: Returning Identity and Data to Users

Every upgrade to the underlying chain ultimately has to serve the wallet and applications people open every day.

As I shared a few days ago in From a Call to Action to Engineering: Passkey-Native Accounts and Wallet Design, the next-generation ArcWallet we are building alongside our new chain brings together native Passkey accounts, a unified multi-account DID view, clear transaction-intent previews, and client-side ledger verification. Users should not have to spend their lives acting as custodians of paper seed phrases and stamped steel plates, and they should not have to bring the whole bank home just to verify their own accounts.

At the application layer, we are pushing forward at full speed with ArcSphere and DApps. By bringing AI agents, personal knowledge and bookmark spaces like Aside, decentralized identity, and everyday applications into an environment controlled by the user, your identity remains your DID and your data stays in your own DID Space. AI works around your data to serve you, instead of locking your digital life inside a tech giant's private database.

A Whitepaper in the Depth of Winter, and the Start of a New Era

Look back to October 2008. When Satoshi Nakamoto posted the Bitcoin whitepaper to the cryptography mailing list, the global financial crisis had just erupted, Lehman Brothers had collapsed weeks earlier, and panic gripped both Wall Street and Silicon Valley.[4] Almost nobody at the time recognized those nine pages of PDF as the doorway to a new era.

Standing here in November 2026, political headwinds, market crashes, talent drain, and AI-armed attackers have pushed our industry into another deep trough. If you only watch price charts, the landscape looks bleak. Walk onto the engineering floor where we are using AI and multi-agent workflows day and night to forge a new generation of chains, wallets, and decentralized applications, and you will see a completely different picture. Whatever the political weather looks like outside, returning identity, data ownership, and asset security to users through better technology is an unstoppable trend. Where the market sees winter and a downturn, builders looking back will recognize the beginning of a new era.


  1. Binance: preliminary findings and response update, November 3, 2026. Binance's preliminary disclosure estimated roughly $6.3 billion in affected assets, with about $3 billion frozen; affected value is not final net loss. ↩
  2. U.S. Senate Roll Call Votes, 119th Congress, 2nd Session. For context on the procedural cloture vote on the CLARITY Act, see The CLARITY Act Is Dead. DAOs Died Long Ago. Is AI the New Hope?. ↩
  3. Following the November 2026 U.S. midterm elections, equities and crypto markets fell sharply amid shifting congressional control and heightened regulatory uncertainty. ↩
  4. Satoshi Nakamoto, Bitcoin: A Peer-to-Peer Electronic Cash System, October 31, 2008. ↩