{"test":"proofread","app":"docs","excluded":"harness fault: the Gemini panel did not open and the prompt was typed into the document"}
{"test":"proofread","app":"revise","excluded":"harness fault: opened the run 1 document (with its pending suggestions) instead of the new import"}
{"config":"revise-argon","app":"revise","model":"Gemini 4 Argon","plan":"Pro","run":1,"session":"2026-10-17 first session","corrected":72,"changed_otherwise":4,"not_addressed":6,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"config":"revise-argon","app":"revise","model":"Gemini 4 Argon","plan":"Pro","run":2,"session":"2026-10-17 first session","corrected":73,"changed_otherwise":4,"not_addressed":5,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"config":"revise-argon","app":"revise","model":"Gemini 4 Argon","plan":"Pro","run":3,"session":"2026-10-17 first session","corrected":73,"changed_otherwise":2,"not_addressed":7,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"config":"docs","app":"docs","model":"Gemini in Docs (model not shown)","run":1,"session":"2026-10-17 first session","corrected":29,"changed_otherwise":1,"not_addressed":52,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)","note":"scored after pressing Gemini's Retry once"}
{"config":"docs","app":"docs","model":"Gemini in Docs (model not shown)","run":2,"session":"2026-10-17 first session","corrected":0,"changed_otherwise":0,"not_addressed":82,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)","failed":true}
{"config":"docs","app":"docs","model":"Gemini in Docs (model not shown)","run":3,"session":"2026-10-17 first session","corrected":51,"changed_otherwise":2,"not_addressed":29,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"config":"docs","app":"docs","model":"Gemini in Docs (model not shown)","run":4,"session":"2026-10-17 first session","corrected":29,"changed_otherwise":2,"not_addressed":51,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"config":"revise-luna","app":"revise","model":"GPT-6 Luna","plan":"Free","run":1,"session":"2026-10-17 first session","corrected":53,"changed_otherwise":4,"not_addressed":25,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"config":"revise-luna","app":"revise","model":"GPT-6 Luna","plan":"Free","run":2,"session":"2026-10-17 first session","corrected":53,"changed_otherwise":6,"not_addressed":23,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"config":"revise-luna","app":"revise","model":"GPT-6 Luna","plan":"Free","run":3,"session":"2026-10-17 first session","corrected":56,"changed_otherwise":5,"not_addressed":21,"test":"proofread","excluded":"superseded: first-session run, replaced by five runs per configuration on 2026-10-18 under a stricter protocol (script timing only, no retries, rotating order)"}
{"test":"proofread","config":"word","app":"word","session":"2026-10-18 rerun","excluded":"quota: Copilot replied 'Oops! Something happened. Can you try again? Please try again tomorrow or upgrade for priority access.' within 3 s; the account's daily AI limit was used up, so this is not a proofread result"}
{"test":"proofread","config":"word","app":"word","session":"2026-10-18 rerun","excluded":"quota: Copilot replied 'Oops! Something happened. Can you try again? Please try again tomorrow or upgrade for priority access.' within 3 s; the account's daily AI limit was used up, so this is not a proofread result"}
{"test":"proofread","config":"word","app":"word","session":"2026-10-18 rerun","excluded":"quota: Copilot replied 'Oops! Something happened. Can you try again? Please try again tomorrow or upgrade for priority access.' within 3 s; the account's daily AI limit was used up, so this is not a proofread result"}
{"test":"proofread","config":"word","app":"word","session":"2026-10-18 rerun","excluded":"quota: Copilot replied 'Oops! Something happened. Can you try again? Please try again tomorrow or upgrade for priority access.' within 3 s; the account's daily AI limit was used up, so this is not a proofread result"}
{"config":"revise-haiku","app":"revise","model":"Claude Haiku 5.5","plan":"Free","run":1,"session":"2026-10-18, added after the main rounds: five rotating rounds of Claude Haiku 5.5, Gemini 3.8 Flash, and GPT-6 Sol, headless Chrome","corrected":35,"changed_otherwise":2,"not_addressed":45,"ms":44535,"ms_lo":44280,"words_changed_approx":33,"test":"proofread","excluded":"left out of the post by choice: Claude Haiku 5.5 is the lowest-cost model in Revise's picker and not one we'd pick for a full proofread of a long document; replaced in the comparison by Claude Sonnet 5.5. A complete, valid run under the same protocol."}
{"config":"revise-haiku","app":"revise","model":"Claude Haiku 5.5","plan":"Free","run":2,"session":"2026-10-18, added after the main rounds: five rotating rounds of Claude Haiku 5.5, Gemini 3.8 Flash, and GPT-6 Sol, headless Chrome","corrected":36,"changed_otherwise":2,"not_addressed":44,"ms":38415,"ms_lo":38159,"words_changed_approx":37,"test":"proofread","excluded":"left out of the post by choice: Claude Haiku 5.5 is the lowest-cost model in Revise's picker and not one we'd pick for a full proofread of a long document; replaced in the comparison by Claude Sonnet 5.5. A complete, valid run under the same protocol."}
{"config":"revise-haiku","app":"revise","model":"Claude Haiku 5.5","plan":"Free","run":3,"session":"2026-10-18, added after the main rounds: five rotating rounds of Claude Haiku 5.5, Gemini 3.8 Flash, and GPT-6 Sol, headless Chrome","corrected":29,"changed_otherwise":0,"not_addressed":53,"ms":30026,"ms_lo":29749,"words_changed_approx":34,"test":"proofread","excluded":"left out of the post by choice: Claude Haiku 5.5 is the lowest-cost model in Revise's picker and not one we'd pick for a full proofread of a long document; replaced in the comparison by Claude Sonnet 5.5. A complete, valid run under the same protocol."}
{"config":"revise-haiku","app":"revise","model":"Claude Haiku 5.5","plan":"Free","run":4,"session":"2026-10-18, added after the main rounds: five rotating rounds of Claude Haiku 5.5, Gemini 3.8 Flash, and GPT-6 Sol, headless Chrome","corrected":32,"changed_otherwise":1,"not_addressed":49,"ms":38534,"ms_lo":38252,"words_changed_approx":36,"test":"proofread","excluded":"left out of the post by choice: Claude Haiku 5.5 is the lowest-cost model in Revise's picker and not one we'd pick for a full proofread of a long document; replaced in the comparison by Claude Sonnet 5.5. A complete, valid run under the same protocol."}
{"config":"revise-haiku","app":"revise","model":"Claude Haiku 5.5","plan":"Free","run":5,"session":"2026-10-18, added after the main rounds: five rotating rounds of Claude Haiku 5.5, Gemini 3.8 Flash, and GPT-6 Sol, headless Chrome","corrected":21,"changed_otherwise":1,"not_addressed":60,"ms":25070,"ms_lo":24814,"words_changed_approx":21,"test":"proofread","excluded":"left out of the post by choice: Claude Haiku 5.5 is the lowest-cost model in Revise's picker and not one we'd pick for a full proofread of a long document; replaced in the comparison by Claude Sonnet 5.5. A complete, valid run under the same protocol."}