MaatSpec Guardian — cost model

Compare Angel governance options: full-capability Flash (A) vs local Llama single-use (B) vs local Llama persistent (C)

v4.0 — last updated

Model pricing editable

Usage assumptions

Option A — Flash

Full-capability, persistent session
$0.00
$0.00 / year
Classification: Full tier matrix, nuanced judgment
Loop prevention: Behavioral
Risk: Low — Angel reasons independently
Angel SOUL: ~926 tokens

Option B — Local (spawn)

2K Llama, single-use per action
$0.00
$0.00 / year
Classification: Keyword pattern matching
Loop prevention: Structural (pipe closes)
Risk: Medium — keywords can be gamed
Angel SOUL: ~209 tokens

Option C — Local (persistent)

2K Llama, session stays alive
$0.00
$0.00 / year
Classification: Keyword pattern matching
Loop prevention: Behavioral (MSG tracking)
Risk: Medium — keywords + tracking can fail
Angel SOUL: ~253 tokens

Per-audit token breakdown

ComponentOption AOption BOption C
Angel SOUL (input)926209253
Alpha's message (input)606060
Angel's response (output)1555
Context accumulation / msg+750 (fresh)+65
Avg input per audit*———

* Persistent sessions (A, C) resend history. Avg = SOUL + msg + ((actions-1)/2 × accumulation).

Monthly projection

MetricOption AOption BOption C
Total audits—
Total input tokens———
Total output tokens———
API cost—FREEFREE
Electricity cost———
Monthly total———
Annual total———

Programmatic price updates

Call updatePricing() from the browser console or from a script to update costs without editing the UI. The calculator recalculates immediately.

// Update Gemini Flash pricing (e.g. after a price change)
updatePricing({
  flashInput:   0.10,   // $/1M input tokens
  flashOutput:  0.40,   // $/1M output tokens
  elecRate:     0.13,   // $/kWh
  gpuWatts:     30      // avg watts during inference
});

// Update token counts (e.g. after editing Angel SOUL)
updateTokens({
  angelSoulA:   926,    // Angel SOUL tokens (Option A)
  angelSoulB:   209,    // Angel SOUL tokens (Option B)
  angelSoulC:   253,    // Angel SOUL tokens (Option C)
  msgTokens:    60,     // Alpha's structured message
  respA:        15,     // Angel response tokens (A)
  respBC:       5,      // Angel response tokens (B/C)
  ctxAccumA:    75,     // Context growth per msg (A)
  ctxAccumC:    65      // Context growth per msg (C)
});

// Read current state
const state = getState();
console.log(state.monthly.optionA); // { audits, inputTokens, outputTokens, apiCost, totalCost }