MO§ES™ · Sovereign Signal · Five Measured Kernels

Five measured kernels.
One operator.

Operator-augmented Claude Code + Opus 4.7 measured against the Artificial Analysis Coding Agent Index field average. Seven-day window. Raw JSONL extraction across 98 session files and all subagents. MO§ES leads every measured economic category: cache discipline, compression cascade, token efficiency, throughput, and cost per line of code shipped.

Live HTML · all 13 published field models · full fidelity · open standalone ↗

Note · Numbers above reflect dashboard window (21 parent sessions). Revised kernels below reflect raw JSONL extraction across 98 session files including all subagents.

Honest Comparison · MO§ES vs. Raw Field Average

Raw JSONL · 98 session files · all subagents · 2 bars per chart · open standalone ↗

Honest Comparison · Portrait · With Cross-Cost Analysis

Token breakdown · field cross-cost · per-kernel significance · open standalone ↗

The Five Kernels
MO§ES™ ranks #1 in all five measures
Cache hit
94·66%
#1 · field avg 90·68%
Out : In
17·9×
#1 · 110× field avg
Tokens / task
810K
#1 · field avg 4·67M
Time / task
1·84 min
#1 · 6·5× faster
$ / LOC
·0007
#1 · < 1¢ per line
From Five Token Categories
Input
123,246
fresh tokens in
Output
3·90 M
3,902,803
Cache Create
34·83 M
34,826,779
Cache Read
1·084 B
1,084,399,183
Total Tokens
1·123 B
1,123,252,011
Methodology · Not the Same Thing Being Measured
AA field: per-task isolated runs on SWE-Bench-Pro-Hard-AA, Terminal-Bench v2, and SWE-Atlas-QnA benchmarks. Each task = one bug/issue, agent runs in a clean session. MO§ES: sustained operator work on a real product (Application Hub) over a 7-day measurement window (2026-05-08 → 2026-05-14), 98 sessions, 7,327 turns, 35,242 LOC shipped in 5 build days. Task-equivalents derived from 7,327 ÷ 5 turns/task = 1,465 tasks. $/LOC: AA = cost_per_task ÷ 20 LOC (industry SWE-Bench convention). MO§ES = $23.33/wk subscription ÷ 35,242 actual LOC = $0.000662. The convergence of leadership across both regimes, isolated benchmark runs and sustained product builds, is the structural result.
VI · The Full Instrument
Five kernels and five token categories — updated raw JSONL numbers
VI · Five kernels → five token categories · updated to raw JSONL · 98 sessions · all subagents
Downloads
Full poster · PDF Portrait · PNG Landscape · PNG Honest comparison · PDF Combined slide · PNG Field source · Artificial Analysis

Field data extracted from artificialanalysis.ai/agents/coding-agents on 2026-05-14. MO§ES values measured from operator-reported 7-day window. Operator: DJM · burnmydays · Ello Cello LLC.