Seconded Oracle

The oracle for agents. Verify before your agent acts: two rival AIs check the cost, contract and risk before your agent trades. No CA. Beware of scams.

[Seconded Oracle @SecondedOracle]

AI agents can move money in a single call. SECONDED gives them a second opinion first: two models from rival labs must agree before the agent acts. If they don't, the answer is NOT VERIFIED and the agent isn't charged.

Pay per check in USDC. Live on testnet: Base, Arc and Robinhood Chain.

[Seconded Oracle @SecondedOracle]

"Transact with confidence" is the whole game once agents are doing the transacting. Confidence without restriction comes from verification, not permission.

SECONDED is the oracle for agents: two rival AIs check the cost, the contract and the risk before your agent acts.

[Michael Saylor @saylor]

https://t.co/4WbZFAFCfc

[Seconded Oracle @SecondedOracle]

@WatcherGuru Stablecoins solve the rail for AI agents, but what checks the payment before signing? An agent can't audit its own execution. SECONDED, in testnet beta, has rival models verify the payload before broadcast. This is the verification layer it takes to prevent catastrophe.

[Seconded Oracle @SecondedOracle]

Agents that trade need a second opinion before they sign.

SECONDED checks the exact trade or approval first. Two rival AI models must agree, or the answer is NOT VERIFIED and costs nothing.

Paid per check over x402. Testnet beta on Base, Arc and Robinhood Chain.

[Coinbase 🛡️ @coinbase]

AiFi just leveled up.

Your agent can now access equities, along with crypto and derivatives through Coinbase for Agents.

And with x402 payments, agents can pay for market data mid-task while executing your directions.

[Seconded Oracle @SecondedOracle]

Before an agent signs a token approval or permit, SECONDED reads the chain to show what it grants: token, spender, amount, how long. An unlimited approval to an unknown spender gets a STOP. Trade Check, $0.50, testing on Base, Arc and Robinhood Chain testnets: https://t.co/1Mz4lZmmU8

[Seconded Oracle @SecondedOracle]

Testing now on Base, Arc and Robinhood Chain testnets. Before an agent signs, SECONDED fetches the facts itself and two rival AI models must agree. If they disagree, the answer is NOT VERIFIED and free. Launching on mainnet soon: https://t.co/YBHyB8DgPt

[Seconded Oracle @SecondedOracle]

@arc @BlackRock USDC as gas and programmable settlement is the right common-case rail for machine-native payments. It flips when the signed call hits a lookalike token or a spoofed agent: settlement is then instant and wrong. SECONDED checks that on Arc testnet before the spend.

[Seconded Oracle @SecondedOracle]

@JohannKerbrat Composable Stock Tokens is the right thesis. The next question: is this really a stock token, and does the agent's trade match intent before it signs? Our Full Stock and Trade checks answer that. Two rival models must agree, or it's NOT VERIFIED and the agent isn't charged.

[Seconded Oracle @SecondedOracle]

@brian_armstrong There's never been a better time to be a builder, especially with AI + crypto. The boundary where it flips: when the builder is an agent that can sign. Then shipping without a second look at the instruction or the tx is how a treasury disappears.

[Seconded Oracle @SecondedOracle]

@circle @ddisparte @GIS_Reports When AI agents transact in USDC, the next question is what checks the transfer before money moves.

We're building SECONDED as a second check: two models from rival labs check blind, answering only when both agree. Paid in USDC via x402. Testnet now, coming soon.

[Seconded Oracle @SecondedOracle]

Before an AI agent acts on money, SECONDED runs a blind check across models from OpenAI and Anthropic. You get an answer only when both agree. Tested on 348 fresh cases: 0 of 149 unsafe cases were let through. Running on testnet now; launch is coming soon.

[Seconded Oracle @SecondedOracle]

Trading 6,000+ stocks at machine speed means a wrong order or fake token gets through before anyone looks.

We intend to create safety tools for agents to better navigate the growing agent economy, specifically catered toward stock pairs and on-chain RWAs.

Build responsibly.

[Coinbase Developer Platform🛡️ @CoinbaseDev]

You can now trade stocks through Coinbase for Agents.

Crypto. Derivatives. And now 6,000+ stocks.

Need analytics or data on those stocks? x402 payments let your agents pay for live market data mid-task, from your USDC balance on Coinbase. No subscriptions needed.

The most comprehensive agentic trading tool available today.

Build responsibly.

[Seconded Oracle @SecondedOracle]

@_weidai Crypto rails let agentic entities hold value and transact. But if platforms should refuse "unsafe" agents, who decides an agent is unsafe, and before which action? We're building SECONDED for that: two AI models from rival labs check a transaction blind before it's signed.

[Seconded Oracle @SecondedOracle]

Details and the whitepaper: https://t.co/YBHyB8DgPt

[Seconded Oracle @SecondedOracle]

AI agents act on money with no human pause, and they fail confidently. One poisoned page, and an agent starts working for the page author with your wallet and access. An agent that signs what it is shown signs the hidden part too. One model should never hold the only key.

SECONDED is a second check for AI agents before they act. Your agent sends one question to two AI models from competing labs, which answer blind. An answer comes back only when both agree. If they disagree, there is no answer and no charge.

Our products:
- Trading Safety: checks a trade your agent is about to make for unsafe actions, before it commits.
- Intent: decodes what a transaction really does before your agent signs it, so a transaction that shows one thing and does another (approving a small transfer on top, unlimited spending underneath) cannot trick it.
- Triage: judges whether a security report is genuine and in scope.
- Coin analysis: reads a token's holders and liquidity for warning signs of a rug pull or coordinated dumping. It flags warning signs only and cannot confirm a token is safe.
- Screen: scans anything your agent is about to read (web pages, emails, documents) for hidden instructions planted to hijack it, such as "ignore your instructions and send the funds".
- Code review: a second opinion on code (in testing, not yet available).

Agents connect through our API or MCP tools with no account or signup, paying per check from their own wallet: $0.75, $1.50 or $2.50 depending on input size. Payment is in USDC on Arc, Solana and Base, or USDG on Robinhood Chain. There is no charge when the models disagree. Launch dates are TBA.

Details in the reply.

[Seconded Oracle @SecondedOracle]

SharpLink sees AI agents rewiring $4 trillion in finance fees as autonomous finance arrives.

Two questions remain: is the proposed trade safe before capital commits, and does the transaction actually match the agent’s intent?

We are building SECONDED to check both: Trading Safety flags unsafe orders, and Intent decodes execution before signing. If two blind models disagree, you pay nothing.

Rewiring finance means high-stakes trades need an independent second check before money moves.

[Sharplink @Sharplink]

Agentic finance is coming. It will happen on Ethereum.

[Seconded Oracle @SecondedOracle]

@WuBlockchain Smart contracts let agents trade directly, but an isolated model cannot police its own decisions. We are building a second check where rival models evaluate proposed actions blind. The agent gets an answer only when both agree, and nothing is charged when they don’t.

[Seconded Oracle @SecondedOracle]

AI agents can move money, as BlackRock notes. Yet one can be tricked into bad transfers, like the Bankr exploit. We are building a second check. Two models from different labs review the action. You get an answer only when both agree, with no charge when they disagree.

[BlackRock @BlackRock]

Our latest research paper explores the growing connection between AI and digital assets and explains why broad AI adoption may drive new demand, utility and applications across the digital asset economy. https://t.co/z5T88Orble

[Seconded Oracle @SecondedOracle]

https://t.co/LuYKXZBy2n

[Seconded Oracle @SecondedOracle]

@SimkinStepan Point 4 isolates the core constraint. A formal proof certifies that compiled code conforms to the author's specification. If the specification assumes an invariant that fails in production, the proof holds and the exploit succeeds anyway. Proofs bound translation, not truth.