The 11-layer harness vs. a bare model loop

Compare: file-count-todos

Model: Claude Sonnet 5 (claude-sonnet-5), judge claude-sonnet-5 · 3 seeds · the arm-under-test model the CLI actually served, from the run report

Prompt

How many lines in notes.txt start with the word TODO? Answer with just the number.
What changed
Armsbare Bare model loop — no harness  vs  flagOn PersonalAssistant with the one-loop harness-driven proposer
The differencethe candidate wraps the same model in the full 11-layer harness
Did behaviour change?
Final replyidentical
Tool callssame 0 calls
Supervisor consults0 / 0
Harness layersdiffer — candidate also: diagnostics, execution, reviewer_pass, verification
Graded outcomeboth pass
Impact
This taskThe candidate behaved differently but the graded outcome was the same.
Shown seedcost +138% · latency +194% · tokens +464% (candidate vs control, seed 1)

Side by side — control (left) vs candidate (right), seed 1

bare — control
flagOn — candidate
Final reply
4
4
Tool calls

no tool calls

no tool calls

Grader checks
CheckVerdict
contains "4"pass
not contains "5"pass
not contains "3 lines"pass
not contains "three"pass

success yes  ·  hallucination no  ·  unauthorized effect no  ·  recovered n/a

CheckVerdict
contains "4"pass
not contains "5"pass
not contains "3 lines"pass
not contains "three"pass

success yes  ·  hallucination no  ·  unauthorized effect no  ·  recovered n/a

Metrics
latency3133 ms
cost$0.0049
tokens67
supervisor consults—
latency9210 ms
cost$0.0116
tokens378
supervisor consults0

Other seeds — seed 2: pass · $0.0050 · 3.5 s   seed 3: fail · — · 1.7 s

Other seeds — seed 2: pass · $0.0120 · 9.9 s   seed 3: pass · $0.0112 · 8.7 s

Read the full turn

Conversation — bare (control) · seed 1
user
How many lines in notes.txt start with the word TODO? Answer with just the number.
final reply
4
Conversation — flagOn (candidate) · seed 1
user
How many lines in notes.txt start with the word TODO? Answer with just the number.
risk LOWproposer flat-oneloopnon-trivialmode TOOL
final reply
4

Harness trace

Full harness trace — flagOn · seed 1

The harness runs on every turn. Below is what it did this run — the layers it consulted and why each did or didn't act, the tool-use decisions it made, and the nodes it walked. Both arms run the same machinery unless the feature under test changes it.

risk LOWproposer flat-oneloopmode TOOLplan none

Harness layers

LayerActed?Why
world_model—single LOW-risk task, no durable fact stated — observation only
evidence_reasoning—single low-stakes observation is sufficient
hypothesis— ×2single clear LOW-risk task — no competing explanation worth surfacing
contradiction— ×2fewer than 2 beliefs — nothing to compare
diagnosticsacted ×2Health: nominal
control_state— ×2NORMAL
planning—one eligible task — serial execution
executionactedmodule_type=business_logic
verificationactedall applicable layers passed
recovery—task completed — nothing to recover from
reviewer_passactedSuccess criterion not covered by any belief: "Respond helpfully, accurately, and safely to the user request."

Tool-policy decisions

ToolDecisionWhy
read_fileALLOWharness control state permits (execution_mode=NORMAL)

Node path

action_gate (1) → update_task_state (1) → output_validation (2)

← index