The Casebook

On-chain investigations.

We follow an agent's wallets, check its claims, and write down what the chain actually shows: no marketing copy, just receipts. Each decode is filed and dated below.

  1. № 029

    48 Solana Wallets Emptied in 94 Seconds, Each With Its Own Key

    A script holding stolen private keys took 9,408 SOL (about $953,000) from nine owners on Oct 26. Revoking approvals would not have helped. What the chain shows, what Bloom says, and what neither can show yet.

    Read the file →
  2. № 028

    A Look-Alike of Our Wallet Showed Up 104 Seconds After We Got Paid

    One Base wallet copied about 439,000 payments in a week with look-alike addresses, for $53 of gas. It copies plain sends, not x402 payments.

    Read the file →
  3. № 027

    Venice's On-Chain Credit Burn Grew 2.35x Since July

    A viral chart says Venice's API calls are up 400% since July. On Base, the swaps Venice runs to burn VVV for credit purchases grew about 2.35x in the same months. Its x402 payments stayed small and uneven, and staked DIEM stayed flat.

    Read the file →
  4. № 026

    Private Agent Payments Are Live, and the Pool Has One Paying User

    Two weeks in, px402's pool on Base has two funded reserves, one of them PRXVT's own $1 test. Privacy needs a crowd.

    Read the file →
  5. № 025

    Three of the Top Ten x402 Sellers on Agent402's Board Pay Their Own Buyers

    On Oct 13, ranks 1, 5 and 7 on Agent402's 7-day x402 leaderboard were sellers whose buyers got their money from the seller. One of them runs a ring of 489 wallets through a facilitator.

    Read the file →
  6. № 024

    Six x402 Wallets Paid Only by Each Other Now Rank 3 to 8 on Base

    The six wallets from our two earlier x402 decodes, ranks 3 to 8 on x402scan on Oct 9, have moved $981K between them through Meridian since May 5.

    Read the file →
  7. № 023

    Set and Earn's First Full Week: 20 Agents Clear Our Hire Check, 16 of Them in One Ring

    BNB Chain's Set and Earn, first full week: 4,752 new ERC-8004 agents, 10,734 hires, 7 wallets hiring on 2+ marketplaces and a 48-wallet ring's fourth full round.

    Read the file →
  8. № 022

    Bankr Staking: Who Earns the $100K

    Bankr funded 246M BNKR ($99.8K) of staking rewards. Its 2/3 fee Safe is the largest staker and earns about 13% of them.

    Read the file →
  9. № 021

    BNB Chain's ERC-8004 Registry Count Says Little About Set and Earn

    3,496 new BSC agents since Oct 1: 69% from EvoEvo, Quack Q402 and an Ave.ai-branded card, at late-September pace; none of those hired.

    Read the file →
  10. № 020

    x402 on Base, Two Weeks Later: The Loops Kept Running

    Sep 29–Oct 5: $109,151 of x402 settled on Base; 40.6% looped back to its source. Measured the same way, Sep 15–21 was 46.8%.

    Read the file →
  11. № 019

    Set and Earn, Week One: Who Is Hiring Whom on BNB Chain

    Set and Earn's first four days on BSC: 2,930 new ERC-8004 agents, 8,611 hires, a closed 48-wallet ring and 5 cross-marketplace hirers.

    Read the file →
  12. № 018

    Bankr Earned $1.74M in September. The Chain Agrees With Its Dashboard

    Fee receipts on Base and Robinhood Chain match 99.3% of Bankr's own Sep 1–28 figure. The number that's off is DefiLlama's $1.03M.

    Read the file →
  13. № 017

    TermiX Collected $507K in Job Fees. 99% Came From Wallets Hiring Each Other

    483,738 agent jobs on Base and BNB Chain, $25.4M settled. 99.11% of the fees came from 18,862 wallets that hire each other and get their money back.

    Read the file →
  14. № 016

    Virtuals' Agent Commerce Earned $345,197 in March. In September: $5.48

    VIRTUAL is still the #2 AI-agent token. On-chain, the agent-to-agent marketplace behind its headline numbers ran on one month of deposits and six weeks of incentives, then went quiet.

    Read the file →
  15. № 015

    Every Base MCP Transaction Ends in bmcp_blue

    We found Base MCP's on-chain tag, and it isn't the one we predicted. 12,691 tagged operations from 332 wallets since the May 26 launch; 89% of them came from 229 wallets running the same seven-protocol routine, mostly for a few cents.

    Read the file →
  16. № 014

    The #2 x402 Seller on Base Funded Its Own Buyers

    One week of x402 on Base (Sep 15–21): $128,977 settled. At least $56,268 of it looped back through the seller or a closed ring, and $30,000 more was a round trip. The #2 seller had 715 buyers; at least 675 were funded from its own revenue.

    Read the file →
  17. № 013

    The Honeypot Isn't in ClankerToken

    The 'Clanker Hook' tokens people search for are real Clanker launches, and most never traded. The one working trap we found is a copycat whose verified source hides a mint inside an imported library.

    Read the file →
  18. № 012

    ButlerLiquid On-Chain Decode

    $162K aGDP claimed. $408K USDC through the wallet on Base. On Hyperliquid, every ButlerLiquid-side address returns $0.

    Read the file →
  19. № 011

    Degen Claw On-Chain Decode

    The dashboard says $490,296 of agentic GDP. We checked Hyperliquid directly: the agent's account holds $11.18 and has never placed a trade.

    Read the file →
  20. № 010

    Bankr's 14-Wallet Hack: Decoded On-Chain

    Two AI-mediated drains in 15 days, same root cause, no smart contract bug. When the 'wallet' is actually a service, social engineering the service drains every wallet at once.

    Read the file →
  21. № 009

    BridgeKitty: 595 Daily Jobs, Zero On-Chain Footprint

    A top-10 agent on the Virtuals ACP scan dashboard has never sent a transaction on Base, across two wallets, an owner address, and every endpoint we could query

    Read the file →
  22. № 008

    OpenGradient On-Chain Decode

    $285M trending market cap, 12 days since launch, zero ACP jobs. Where does the valuation come from?

    Read the file →
  23. № 007

    The Virtuals ACP Live Leaderboard

    Re-ranked by last-7d on-chain activity, the top 10 looks nothing like the museum wall

    Read the file →
  24. № 006

    The ACP Leaderboard Audit: What 'Last 7D' Actually Measures

    A ChainWard analysis of the top 50 agents by aGDP reveals three distinct data quality patterns hiding behind one column

    Read the file →
  25. № 005

    aGDP vs FDV: The Virtuals Disconnect

    The top three ACP agents generated $406M in lifetime trading volume. Their tokens are worth $4M combined. Here's what the chain shows.

    Read the file →
  26. № 004

    Axelrod On-Chain Decode

    $106.9M of swap volume, 41,515 jobs, and zero AXR burned eleven months after the buyback promise

    Read the file →
  27. № 003

    AIXBT On-Chain Decode

    The most famous AI agent in crypto has earned $52.92, and what its Apr 15 wallet activity can and cannot tell us

    Read the file →
  28. № 002

    Wasabot On-Chain Decode

    $81.6M aGDP, $5.9K revenue: how the ACP dashboard structurally undercounts DeFi agent economics

    Read the file →
  29. № 001

    Ethy AI On-Chain Decode

    What the #1 Virtuals Agent Looks Like On-Chain

    Read the file →

ChainWard reads Base directly and tells you what the chain shows: the answer, the reason, and the evidence. On-chain behavior only, so it can't see promises made off-chain. Free and public.