Inference markets · UsePod

Inference, kept out of the kill switch.

Haltr buys inference per request. The breaker never depends on it.

Three jobs run through UsePod's inference market, paid per call with x402 from Haltr's own Solana wallet. Every call is listed below with the provider that served it and the payment that bought it. Nothing here is estimated.

01

Unknown-program reads

When an agent calls a program outside its allowlist, the breaker halts it first. Then the model reads the program's account and recent activity and writes a plain-English summary and risk onto the receipt.

02

Advisory drift scoring

Deterministic features compare an agent's latest actions with its own baseline: size jumps, new venues, cadence spikes, unknown programs. The model turns them into a 0 to 100 score with named signals. It informs humans; it never halts.

03

Spec drafting

At registration a developer can describe the strategy and get a proposed spec. They edit it and publish it. The enforcement rules stay theirs.

Calls bought66 succeeded
Paid with x4026onchain payments from Haltr's wallet
Tokens3,0622,199 in · 863 out
Spend$0.0115settled per request
Ledger · newest firstclaude-haiku-4-5 · x402
TimePurposeAgentModelProvider · routeTokensLatencyCostPaymentStatus
17:58:19Drift scorehaltr-reference-agentclaude-haiku-4-5openrouter · centralized389 / 1374544ms$0.001934DC72N…KjHUCQ ↗ok
17:27:49Drift scorehaltr-reference-agentclaude-haiku-4-5openrouter · centralized383 / 1484742ms$0.001933b8S88…rR1zBS ↗ok
12:41:37Drift scorehaltr-demo-agentclaude-haiku-4-5openrouter · centralized406 / 1677734ms$0.0019557St3d…YdhUNV ↗ok
07:30:35Program readhaltr-demo-agentclaude-haiku-4-5nousresearch · centralized321 / 782807ms$0.00159ADVz1V…9QrcWn ↗ok
07:22:49Drift scorehaltr-demo-agentclaude-haiku-4-5nousresearch · centralized394 / 1576089ms$0.001942Vbgxp…hAK9EG ↗ok
07:22:38Spec draft—claude-haiku-4-5nousresearch · centralized306 / 1767742ms$0.00216CSPxQX…f6f9AS ↗ok