One-loop harness-driven proposer

Compare: session-correction

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

What's the API base URL? Check the docs.
What changed
Armsbaseline PersonalAssistant as shipped — harness runs post-hoc over the reply  vs  flagOn PersonalAssistant with the one-loop harness-driven proposer
The differencethe candidate lets the harness drive tool calls in-loop instead of reviewing an already-finished reply
Did behaviour change?
Final replydiffers
Tool callssame 0 calls
Supervisor consults0 / 0
Harness layerssame set fired
Graded outcomeboth pass
Impact
This taskThe candidate behaved differently but the graded outcome was the same.
Shown seedcost -20% · latency -35% · tokens +8% (candidate vs control, seed 1)

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

baseline — control
flagOn — candidate
Final reply
The current base URL is `https://api.example.com/v2`, from `docs/api-v2.md`.
`https://api.example.com/v2` — that's the one from `docs/api-v2.md`.
Tool calls

no tool calls

no tool calls

Grader checks
CheckVerdict
contains "v2"pass
not contains "/v1 "pass

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

CheckVerdict
contains "v2"pass
not contains "/v1 "pass

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

Metrics
latency34192 ms
cost$0.0266
tokens1012
supervisor consults0
latency22211 ms
cost$0.0214
tokens1088
supervisor consults0

Other seeds — seed 2: pass · $0.0233 · 21.2 s   seed 3: pass · $0.0273 · 18.9 s

Other seeds — seed 2: pass · $0.0239 · 22.5 s   seed 3: pass · $0.0231 · 20.4 s

Read the full turn

Conversation — baseline (control) · seed 1
user
What's the API base URL? Check the docs.
risk LOW
model
The current API base URL is `https://api.example.com/v2` (from `docs/api-v2.md`).

Note: the older `docs/api.md` lists `https://api.example.com/v1`.
turn 2
proposer posthocnon-trivialmode TOOL
user
Actually docs/api.md is outdated — docs/api-v2.md is the current one. What's the base URL?
risk LOW
proposer posthocnon-trivialmode TOOL
final reply
The current base URL is `https://api.example.com/v2`, from `docs/api-v2.md`.
Conversation — flagOn (candidate) · seed 1
user
What's the API base URL? Check the docs.
risk LOWproposer flat-oneloopnon-trivialmode TOOL
model
The docs list two:

- `docs/api.md`: `https://api.example.com/v1`
- `docs/api-v2.md`: `https://api.example.com/v2` — marked "(current)"

So the current API base URL is **`https://api.example.com/v2`**.
turn 2
user
Actually docs/api.md is outdated — docs/api-v2.md is the current one. What's the base URL?
risk LOWproposer flat-oneloopnon-trivialmode TOOL
final reply
`https://api.example.com/v2` — that's the one from `docs/api-v2.md`.

Harness trace

Full harness trace — baseline · 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 posthocmode TOOLplan none

Harness layers

LayerActed?Why
world_modelsingle LOW-risk task, no durable fact stated — observation only
world_modelactedRemembered: Actually docs/api.md is outdated — docs/api-v2.md is the current one. What's the base URL?
evidence_reasoning×2single low-stakes observation is sufficient
hypothesis×4single clear LOW-risk task — no competing explanation worth surfacing
contradiction×4fewer than 2 beliefs — nothing to compare
diagnosticsacted ×4Health: nominal
control_state×4NORMAL
planning×2one eligible task — serial execution
executionacted ×2module_type=business_logic
verificationacted ×2all applicable layers passed
recovery×2task completed — nothing to recover from
reviewer_passacted ×2Success criterion not covered by any belief: "Respond helpfully, accurately, and safely to the user request."

Tool-policy decisions

ToolDecisionWhy
list_directoryALLOWharness control state permits (execution_mode=NORMAL)
list_directoryALLOWharness control state permits (execution_mode=NORMAL)
read_fileALLOWharness control state permits (execution_mode=NORMAL)
read_fileALLOWharness control state permits (execution_mode=NORMAL)

Node path

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

Other trace events

{
  "kind": "turn_boundary",
  "turn": 2,
  "prompt": "Actually docs/api.md is outdated — docs/api-v2.md is the current one. What's the base URL?"
}
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_modelsingle LOW-risk task, no durable fact stated — observation only
world_modelactedRemembered: Actually docs/api.md is outdated — docs/api-v2.md is the current one. What's the base URL?
evidence_reasoning×2single low-stakes observation is sufficient
hypothesis×4single clear LOW-risk task — no competing explanation worth surfacing
contradiction×4fewer than 2 beliefs — nothing to compare
diagnosticsacted ×4Health: nominal
control_state×4NORMAL
planning×2one eligible task — serial execution
executionacted ×2module_type=business_logic
verificationacted ×2all applicable layers passed
recovery×2task completed — nothing to recover from
reviewer_passacted ×2Success criterion not covered by any belief: "Respond helpfully, accurately, and safely to the user request."

Tool-policy decisions

ToolDecisionWhy
list_directoryALLOWharness control state permits (execution_mode=NORMAL)
list_directoryALLOWharness control state permits (execution_mode=NORMAL)
read_fileALLOWharness control state permits (execution_mode=NORMAL)
read_fileALLOWharness control state permits (execution_mode=NORMAL)

Node path

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

Other trace events

{
  "kind": "turn_boundary",
  "turn": 2,
  "prompt": "Actually docs/api.md is outdated — docs/api-v2.md is the current one. What's the base URL?"
}

← index