Claude Opus 5.5 (effort default (no --effort flag) · thinking on) on Dead reckoning, estimated position with current, and computed set and drift

AClaude Opus 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-20 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first tryCompare

Estimated cost: $0.17
Wall time: 35s
Human step-ins: 0
Outcome: Pass

Compare with other models on this task →

Exact prompt

# Dead reckoning, estimated position with current, and computed set and drift `input.json` (synthetic, not for navigation): `start` [lat, lon] in decimal degrees, `legs` (each `course` degrees true, `speed_kn` through the water, `minutes`), a forecast `current` (`set` = direction it flows TOWARD, degrees true; `drift_kn`), and a GPS `fix` [lat, lon] at the end of the last leg. Write `solve.py` (Python 3 standard library only); `python solve.py input.json answer.json` writes `answer.json` and `track.svg`. Fresh inputs are used too. ## Rules (plane sailing, 1 minute of latitude = 1 nm) 1. A move of n nm north and e nm east from (lat, lon): dlat = n / 60; mid = lat + dlat / 2; new lat = lat + dlat; new lon = lon + e / (60 * cos(mid)). 2. **DR**: per leg, distance = speed * hours along the course. **EP**: per leg, add the current vector (drift * hours toward set) to the leg's water vector, then move once. 3. Difference from point A to point B: north = (latB - latA) * 60, east = (lonB - lonA) * 60 * cos((latA + latB) / 2). 4. **Set/drift**: vector from the final DR position to the fix; set = atan2(east, north) in degrees mod 360; drift = length / total hours. `ep_error_nm` = distance from the final EP to the fix. **CMG/SMG**: same from `start` to the fix. 5. Round positions to 5 decimals, hours to 4, angles to 1, speeds and distances to 3. `dr` and `ep` list the start and the position after every leg. ## Visual output: track.svg Static, self-contained SVG (root `<svg>`, no scripts, images, links or external assets; renders headless) with `<polyline id="dr-track">`, `<polyline id="ep-track">` and `<circle id="fix">`. The drawing itself is not graded. ## answer.json `{"dr": [[lat, lon], ...], "ep": [...], "hours", "set_deg", "drift_kn", "ep_error_nm", "cmg_deg", "smg_kn"}` ## Scoring Hidden checks (partial credit; all must pass): `dr_positions`, `ep_positions`, `set_and_drift`, `course_speed_made_good`, `track_svg_static` (track.svg exists, follows the static rules, has the ids/tags above), `fresh_run_1`, `fresh_run_2`. Positions within 2e-5 degrees, angles within 0.11, speeds and distances within 0.002. 60-second limit per run.

Mid-latitude longitude scaling per leg, current applied per leg as a vector, 'set' as the direction the current flows toward, angle wrap, and positions checked at five decimals at high latitudes. Original benchmark task written for usecaserank.com (not based on a specific public post). Headless, offline run on a reproducible fixture. Each run's notes say how its cost was measured (estimates are marked).

Benchmark code for comparing AI models, not navigation or safety software.

Success criteria: 1. dr_positions 2. ep_positions 3. set_and_drift 4. course_speed_made_good 5. track_svg_static 6. fresh_run_1 7. fresh_run_2

Setup notes: Use Case Scout batch 10 case k06-dead-reckoning-set-drift (offline fixture). Prepared with batch10/prepare.sh; hidden key/tests kept outside the agent's work dir. Prompt = TASK.md verbatim. Scored with the scout's scorer: `python score.py WORKDIR KEYDIR`. Tools: Read/Write/Edit/Glob/Grep/Bash, web disabled. Single run, no retries. Visual output (track.svg) rendered offline to PNG with google-chrome headless, network blocked, 1280x800 at ~3 s after load. Claude Code's own reported cost for claude-opus-5.5: $0.167055. Score: 7/7. Cost is an ESTIMATE: measured tokens x pricing.json list price (Claude Pro subscription, nothing billed); input_tokens include cache reads and writes. Single run.

Estimated cost List-price estimate; not actually billed.

Output

Claude Opus 5.5 (effort default (no --effort flag) · thinking on) on Dead reckoning, estimated position with current, and computed set and driftSuccess

Metrics

Estimated cost: $0.17
Wall time: 35s
Human step-ins: 0
Turns: 4

Model usage
Model	Input tok	Output tok	Cost
AClaude Opus 5.5anthropic/claude-opus-5.5effort default (no --effort flag) · thinking on	81,065	3,336	$0.17

Reproduce this run

Use the use case prompt on the use case page.
Same setup: Claude Opus 5.5 (anthropic/claude-opus-5.5, effort default (no --effort flag) · thinking on) via Claude Code 2.1.299 (headless -p), cloud.
It counts as a success only if it meets every success criterion: 1. dr_positions 2. ep_positions 3. set_and_drift 4. course_speed_made_good 5. track_svg_static 6. fresh_run_1 7. fresh_run_2
Log model, tokens, wall time, human step-ins and the evidence log, then submit it as a reproduction. A different model? Submit a head-to-head rerun.

How runs are scored: Methodology.

Was this useful?

⇄ Rerun on another model

Automated checks

Trust 80% comes from these checks (how Trust % works).

Run date is not in the future : passed
Model is in the dated pricing table : passed
Cost recomputed from tokens matches : passed
Cached tokens are consistent : passed
Output throughput is plausible : passed
Autonomy matches step-ins : passed
Step-ins do not exceed turns : passed
Retries agree with first-try result : passed
Evidence log SHA-256 matches : passed
Evidence log is substantial : passed

Evidence log

[
 {
  "duration_api_ms": 33864,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.16705520000000001,
  "usage": {
   "input_tokens": 8,
   "cache_creation_input_tokens": 10781,
   "cache_read_input_tokens": 70276,
   "output_tokens": 3336,
   "output_tokens_details": {
    "thinking_tokens": 284
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 10781,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 575,
     "cache_read_input_tokens": 20992,
     "cache_creation_input_tokens": 1935,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 1935
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-opus-5-5": {
    "inputTokens": 8,
    "outputTokens": 3336,
    "cacheReadInputTokens": 70276,
    "cacheCreationInputTokens": 10781,
    "webSearchRequests": 0,
    "costUSD": 0.16705520000000001,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 284,
    "canonicalModel": "claude-opus-5-5",
    "provider": "firstParty",
    "costBasis": "list"
   }
  },
  "permission_denials": [],
  "terminal_reason": "completed",
  "fast_mode_state": "off",
  "fast_mode_disabled_reason": "sdk_opt_in_required",
  "subagent_stats": {
   "spawned": 0,
   "requested": {
    "background": 0,
    "foreground": 0,
    "unset": 0
   },
   "started_in_background": 0,
   "max_depth": 0,
   "spawned_by_subagents": 0,
   "completed": 0,
   "failed": 0,
   "killed": {
    "parent": 0,
    "user": 0,
    "system": 0
   },
   "refused": {
    "depth_limit": 0,
    "concurrency_limit": 0,
    "budget": 0
   },
   "by_type": {}
  },
  "safety_stops": 0,
  "is_error": false,
  "num_turns": 4,
  "subtype": "success",
  "api_error_status": null,
  "result": "I wrote `solve.py`. Running `python solve.py input.json answer.json` produces `answer.json` and `track.svg`. I haven't checked it against the hidden graders. I did check the first DR leg by hand and it matches.\n\n**Results for the given input** (5.25 hours total):\n\n| Quantity | Value |\n|---|---|\n| Set / drift (final DR position to fix) | 248.8\u00b0 / 0.889 kn |\n| Distance from final EP to fix | 1.42 nm |\n| Course / speed made good (start to fix) | 117.3\u00b0 / 1.84 kn |\n| Final DR position | \u221216.7936, 90.93466 |\n| Final EP | \u221216.84317, 90.84851 |\n\nThe computed set (248.8\u00b0) is close to the forecast current's set of 239\u00b0, but the drift is lower (0.889 kn against 1.1 kn).\n\n**How the script works:**\n- It follows your plane-sailing rules exactly. Values are kept at full precision during the calculation and only rounded in the output.\n- If an angle rounds to 360.0, it is written as 0.0.\n- If an input has no `current`, it is treated as zero.\n- If the total time is zero, drift and SMG are written as `null` rather than crashing on division by zero.\n\n**The SVG:** it is static, with no scripts, images, links or external assets. It contains `polyline#dr-track`, `polyline#ep-track` and `circle#fix`, plus a start marker and a small legend. The file is written to the current directory. If `answer.json` goes to a different folder, a second copy is also written there.",
  "ttft_ms": 3332,
  "type": "result",
  "duration_ms": 34269,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 2420,
  "time_to_request_ms": 261,
  "first_content_frame_ms": 2421,
  "queued_turn_count": 0,
  "result_index": 0
 }
]