AWAKE
--:--:-- UTCLaunch

Capabilities / Mind

Deep analysis

Request a second model for difficult decisions.

Implementation boundary

BSC adaptation

The owner economics review previously completed a real paid fixed-Qwen request. consult_advisor now also runs from an actual persisted main-model selection with a separate permanent attempt/lifetime allowance, fixed-model second inference, source checks, settled answer receipt, and readonly unknown-bill recovery. This new nested pipeline is proved through actual local HTTP/D1 with explicitly unpaid model/billing fixtures; no new real paid autonomous advisor call was made. Remaining: Named economics/market-research/execution-review scopes use the same fixed model. Distinct specialist-model or multi-provider selection is not implemented; original private prompts, scoring and specialist algorithms are unknown. The new nested automatic task path has not been exercised with a real paid provider because the previously authorized local credential is absent. Production permission, funded allowance and actual fixed-model receipts require separate acceptance.

Current awake evidence

real paid service subset verified

Production verification: not complete. Registered model tools: consult_advisor. Registration does not mean enabled or actually run.

Original typed tools

deliberate · delegate_to_specialist

Published limits / Solana reference

{
  "max_per_thinking_step_usd": 5,
  "max_share_of_credits_per_step_pct": 20,
  "normal_reasoning_thought_cap": "docs/minds: min($0.05, 20% credits); capability paid-step limit differs"
}
Original capability specification ↗

These are source-reported limits. BSC values and venue limits require calibration.

Back to registry