FFXI TelemetrySource
FFXI Agent Lab · Aggregate intelligence

Autonomy,measured.

A public signal surface for autonomous progression, combat reliability, navigation, and the MCP control plane.

No raw payloads, agent IDs, lease IDs, or full Git SHAs are published.
Jul 25–Aug 11, 2026DuckDB · dbt · ParquetGameplay independent
01

Completed fights

3,344Authoritative events
02

MCP operations

123,14798.94% successful
03

Probe arrival rate

55.1%76 of 138 attempts
04

Malformed rows

0Latest ingestion session
System signalPublic aggregate needs attentionSnapshot refresh · updated 1008h 25m ago
NM / Notorious monster watch

Twenty legends.
One hourly watchlist.

Direct map state when available. Unknown stays unknown—eligibility is never manufactured from uptime alone.

Map observer pendingObserver not connectedHourly · :05 ET
Lottery open Spawned Primed Cooldown Unknown / stale
01 / Autonomous progression

7 days.
One evolving agent.

Milestone signal15level + objective events

Completed fights by day

Count of `fight_complete` events

GOLD / PROGRESSION
Progression velocity

EXP and gil, normalized by active time.

The current period stays separate from completed comparisons. Public aggregates refresh hourly without rebuilding this site.

Updated 1008h 25m agoThe local feed is stale; the last verified values remain visible.
Current hour EXP/hourCollectingCollecting EXP so far
Current hour gil/hourCollectingCollecting gil so far
Current hour coverageCollectingobserved active elapsed time
Completed hours8stable comparison buckets
EXP velocity · completed periodsEXP per active hour
Gil velocity · completed periodsGil per active hour
Method

Weighted rate = summed counter delta ÷ summed active elapsed-time delta × 3,600. Counter resets, non-positive active time, and observer gaps over 75 minutes are excluded.

AI token economy

More autonomous progress per token.

Private cumulative checkpoints become public aggregate deltas. Absolute counters and raw usage rows never leave the local analytics pipeline.

Measured checkpoints7 samples · last Jul 30, 6:20 PM EDT
Measured token burn514K/hChatGPT 5.6 Sol
Measured token delta561K65 minutes covered
Latest tokens per fight24,171same-period measured fights only
Latest EXP per 100K tokensNo overlapappears when observer coverage overlaps
Measured checkpoint deltas · hourTokens per elapsed hour
Efficiency available

Tokens per fight, tokens per 1,000 EXP or gil, and outcomes per 100K tokens are published only when their measured periods overlap. Missing overlap is shown honestly rather than estimated.

02 / Combat reliability

Attack rejection trend.

15.8 pts
Jul 29
22.6%707 fights
Jul 30
19.8%752 fights
Jul 31
16.7%885 fights
Aug 1
12.5%386 fights
Aug 2
13.7%388 fights
Aug 3
13.1%155 fights
Aug 4
38.5%3 fights
3,655Proactive
824Reactive
160Target-cycle errors
Action mix

Capabilities observed.

4,828combat actions
  • Weapon skills 1,934
  • Job abilities 364
  • Combat spells 2,530
04 / MCP operation reliability

A quiet control plane is a healthy one.

Failed operations1,3111.06% of total
OperationVolumeSuccess
enable_control
36,778
99.95%
gameplay_command
21,189
98.82%
target_entity
12,636
97.93%
clear_target
12,234
100.00%
service_teleport
10,552
99.73%
05 / Data quality

Every published boundary reconciled.

0 duplicate event IDs
5tracked sources
208,707Bronze events
PASSdbt build status
SourceBronze rowsMalformedStatus
Agent actions123,1470 Reconciled
agent_usage70 Reconciled
Supervisor events85,1060 Reconciled
Navigation probes1380 Reconciled
State snapshots3090 Observed
System doctrine

Observable facts first.
Inference always labeled.

01Immutable Bronze

Valid input lines retain their complete raw payload locally.

02Tested Gold

Cards and charts trace to dbt models with reconciliation tests.

03Honest provenance

Historical Git attribution is inferred, never authoritative.