Claude Sonnet 5.5 on Research use cases and build pitch slides

AClaude Sonnet 5.5 · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-14 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first tryCompare

Estimated cost: $0.07
Wall time: 22s
Human step-ins: 0
Outcome: Pass

Compare with other models on this task →

Exact prompt

The folder ./sources contains frozen notes on real AI-agent use cases (one file each, with source URL and stated metrics). Do not use the web. Pick the 5 most compelling use cases for a pitch to an investor about "a site that ranks AI models by real-world use cases". Write slides.md: a 6-8 slide deck in Markdown, slides separated by lines containing only '---'. Slide 1 is a title slide. Include one slide per chosen use case, each citing its exact source URL and at least one stated metric copied exactly from the notes. End with an ask/next-steps slide.

Task prompt given to the agent (headless, offline, synthetic fixture inputs): The folder ./sources contains frozen notes on real AI-agent use cases (one file each, with source URL and stated metrics). Do not use the web. Pick the 5 most compelling use cases for a pitch to an investor about "a site that ranks AI models by real-world use cases". Write slides.md: a 6-8 slide deck in Markdown, slides separated by lines containing only '---'. Slide 1 is a title slide. Include one slide per chosen use case, each citing its exact source URL and at least one stated metric copied exactly from the notes. End with an ask/next-steps slide. Inspired by https://x.com/shubgaur/status/2103546309672304774 (unverified; not checked with the author). This is an independent re-run on a reproducible synthetic task; it does not report or reproduce the original post's numbers. Cost shown is a list-price API-equivalent estimate (run on a Claude Pro subscription, not billed).

Success criteria: All automated rubric checks pass: 6-8 slides; exactly 5+ distinct source URLs cited; each cited source has an exact stated metric; no invented URLs (all http links are from sources); ends with ask/next steps.

Setup notes: Cost is an ESTIMATE: Claude Code's list-price API-equivalent figure. The run used a Claude Pro subscription, so nothing was billed. input_tokens = all input incl. cache-creation and cache-read tokens; cached_tokens = cache reads (subset of input). Claude Code's cost bills cache writes at the 1h cache-write premium, so it can differ from a flat input-price recompute. Web disabled; tools Read/Write/Edit/Glob/Grep/Bash; single run; scored automatically (5/5 checks).

Estimated cost List-price estimate; not actually billed.

Metrics

Estimated cost: $0.07
Wall time: 22s
Human step-ins: 0
Turns: 4

Model usage
Model	Input tok	Output tok	Cost
AClaude Sonnet 5.5anthropic/claude-sonnet-5.5	81,384	2,386	$0.07

Reproduce this run

Use the use case prompt on the use case page.
Same setup: Claude Sonnet 5.5 (anthropic/claude-sonnet-5.5) via Claude Code 2.1.299 (headless -p), cloud.
It counts as a success only if it meets every success criterion: All automated rubric checks pass: 6-8 slides; exactly 5+ distinct source URLs cited; each cited source has an exact stated metric; no invented URLs (all http links are from sources); ends with ask/next steps.
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": 20870,
 "stop_reason": "end_turn",
 "session_id": "[REDACTED]",
 "total_cost_usd": 0.0739152,
 "usage": {
  "input_tokens": 8,
  "cache_creation_input_tokens": 10744,
  "cache_read_input_tokens": 70632,
  "output_tokens": 2386,
  "output_tokens_details": {
   "thinking_tokens": 255
  },
  "server_tool_use": {
   "web_search_requests": 0,
   "web_fetch_requests": 0
  },
  "service_tier": "standard",
  "cache_creation": {
   "ephemeral_1h_input_tokens": 10744,
   "ephemeral_5m_input_tokens": 0
  },
  "inference_geo": "global",
  "iterations": [
   {
    "input_tokens": 2,
    "output_tokens": 494,
    "cache_read_input_tokens": 21460,
    "cache_creation_input_tokens": 1823,
    "cache_creation": {
     "ephemeral_5m_input_tokens": 0,
     "ephemeral_1h_input_tokens": 1823
    },
    "type": "message"
   }
  ],
  "speed": "standard",
  "fallback_credit": null
 },
 "modelUsage": {
  "claude-sonnet-5-5": {
   "inputTokens": 8,
   "outputTokens": 2386,
   "cacheReadInputTokens": 70632,
   "cacheCreationInputTokens": 10744,
   "webSearchRequests": 0,
   "costUSD": 0.0739152,
   "contextWindow": 1000000,
   "maxOutputTokens": 128000,
   "thinkingTokens": 255,
   "canonicalModel": "claude-sonnet-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 `slides.md` in the work directory. It has 8 slides separated by `---` lines: title, problem, five use cases, and the ask.\n\nThe five use cases are:\n1. **GitHub Actions debugging, Grok vs Opus.** Opus cost \"~$10 worth of tokens (did not find cause)\". Grok cost \"$1.2 worth of tokens\".\n2. **107 e-commerce tasks across three agents.** Accio cost \"about $3.69\", Codex \"about $9.27\" and Claude Code \"about $9.51\".\n3. **Sonnet 5.5 max vs Opus 5.5 xhigh.** Per-task cost was 7.67 for Sonnet and 3.46 for Opus.\n4. **Minecraft clone, Sol vs Fable.** Sol in Devin took \"70 minutes\". Fable in Claude Code took \"90 minutes\".\n5. **Full DeepSWE run on a local 27B model.** Score 26.5, \"45 hours\", and 820000000 total tokens.\n\nEach use-case slide gives its source URL and copies the stated metrics from the notes exactly. I picked cases that compare models, agents or settings on cost or time, because that is what a ranking site sells. I left out the single-agent demos and the failure case.\n\nThe slides call the evidence single-user reports, not controlled experiments. Slide 3's \"smaller model cost more\" point and slide 2's \"roughly 40%\" figure are my own readings of the numbers. Please check they say what you want.",
 "ttft_ms": 2119,
 "type": "result",
 "duration_ms": 21329,
 "uuid": "[REDACTED]",
 "ttft_stream_ms": 1391,
 "time_to_request_ms": 323,
 "first_content_frame_ms": 1392,
 "queued_turn_count": 0,
 "result_index": 0
}