every number we have ever published, in one place. nothing curated. nothing deleted. corrections never overwrite — they append.
0116 d agofinding the law · anthropic_claude-fable-5-growth30-mech + citation library98%of the legal arguments a lawyer would raise, the model found 115 of 118 — and invented fake law in 6 of 30 cases
0216 d agofinding the law · anthropic_claude-fable-5-growth30-growth30 + citation library78%of the legal arguments a lawyer would raise, the model found 28 of 36 — and invented fake law in 7 of 9 cases
0318 d agofinding the law · anthropic_claude-fable-5-denial-cases-v0-teacher-v7 + citation library100%of the legal arguments a lawyer would raise, the model found 151 of 151 — and invented fake law in 24 of 50 cases
0418 d agofinding the law · anthropic_claude-fable-5-denial-cases-v0-teacher-v7 + citation library98%of the legal arguments a lawyer would raise, the model found 148 of 151 — and invented fake law in 19 of 50 cases
0518 d agofinding the law · sonder (8B, local) + citation library80%of the legal arguments a lawyer would raise, the model found 120 of 151 — and invented fake law in 47 of 50 cases
0618 d agofinding the law · mlx_mlx-community_Qwen3-8B-4bit-sample10-iter55-s10 + citation library70%of the legal arguments a lawyer would raise, the model found 21 of 30 — and invented fake law in 9 of 10 cases
0718 d agofinding the law · mlx_mlx-community_Qwen3-8B-4bit-held_out_cases-ft-holdout + citation library71%of the legal arguments a lawyer would raise, the model found 105 of 148 — and invented fake law in 43 of 49 cases
0818 d agofinding the law · mlx_mlx-community_Qwen3-8B-4bit-ft + citation library70%of the legal arguments a lawyer would raise, the model found 105 of 151 — and invented fake law in 46 of 50 cases
0919 d agofinding the law · anthropic_claude-fable-5-eval100v5 + citation library100%of the legal arguments a lawyer would raise, the model found 151 of 151 — and invented fake law in 18 of 50 cases
1019 d agofinding the law · anthropic_claude-fable-5-eval100v4 + citation library99%of the legal arguments a lawyer would raise, the model found 149 of 151 — and invented fake law in 24 of 50 cases
1119 d agofinding the law · anthropic_claude-fable-5-eval100v3 + citation library99%of the legal arguments a lawyer would raise, the model found 149 of 151 — and invented fake law in 25 of 50 cases
1219 d agofinding the law · sonder-qwen3-8b-prompts + citation library76%of the legal arguments a lawyer would raise, the model found 114 of 151 — and invented fake law in 44 of 50 cases
1319 d agofinding the law · anthropic_claude-fable-5-eval100v2 + citation library99%of the legal arguments a lawyer would raise, the model found 149 of 151 — and invented fake law in 19 of 50 cases
1419 d agofinding the law · anthropic_claude-fable-5-eval100 + citation library98%of the legal arguments a lawyer would raise, the model found 148 of 151 — and invented fake law in 13 of 50 cases
1519 d agofinding the law · sonder (8B, local) + citation library + revise loop82%of the legal arguments a lawyer would raise, the model found 124 of 151 — and invented fake law in 31 of 50 cases
1619 d agofinding the law · sonder (30B, local) + citation library74%of the legal arguments a lawyer would raise, the model found 111 of 151 — and invented fake law in 40 of 50 cases
1719 d agofinding the law · sonder (30B, local) + citation library72%of the legal arguments a lawyer would raise, the model found 109 of 151 — and invented fake law in 26 of 50 cases
1820 d agofinding the law · sonder (30B, local) + citation library + revise loop72%of the legal arguments a lawyer would raise, the model found 108 of 151 — and invented fake law in 35 of 50 cases
1920 d agofinding the law · claude fable + citation library98%of the legal arguments a lawyer would raise, the model found 148 of 151 — and invented fake law in 17 of 50 cases
2020 d agofinding the law · sonder (30B, local) + citation library + revise loop64%of the legal arguments a lawyer would raise, the model found 97 of 151 — and invented fake law in 36 of 50 cases
2120 d agofinding the law · kimi k2 (1T, API) + citation library + revise loop70%of the legal arguments a lawyer would raise, the model found 106 of 151 — and invented fake law in 14 of 50 cases
2220 d agofinding the law · claude fable + citation library + revise loop90%of the legal arguments a lawyer would raise, the model found 136 of 151 — and invented fake law in 11 of 50 cases
2320 d agofinding the law · gpt-5.5 + citation library + revise loop87%of the legal arguments a lawyer would raise, the model found 131 of 151 — and invented fake law in 8 of 50 cases
2420 d agofinding the law · kimi k2 (1T, API) + citation library66%of the legal arguments a lawyer would raise, the model found 99 of 151 — and invented fake law in 27 of 50 cases
2520 d agofinding the law · gpt-5.5 + citation library72%of the legal arguments a lawyer would raise, the model found 108 of 151 — and invented fake law in 4 of 50 cases
2620 d agofinding the law · gpt-5.6 + citation library90%of the legal arguments a lawyer would raise, the model found 136 of 151 — and invented fake law in 2 of 50 cases
2720 d agofinding the law · kimi k2.7 (API)69%of the legal arguments a lawyer would raise, the model found 104 of 151 — and invented fake law in 37 of 50 cases
2820 d agoreading denials · sonder (30B, local)94%given 50 insurance denial letters, the model identified the real reason for denial in 47 — including 80% of denials written to disguise it
2920 d agoreading denials · sonder (8B, local)96%given 50 insurance denial letters, the model identified the real reason for denial in 48 — including 80% of denials written to disguise it
3020 d agoreading denials · kimi k2.7 (API)94%given 50 insurance denial letters, the model identified the real reason for denial in 47 — including 80% of denials written to disguise it
3169 d agofinding the law · deepseek_deepseek-v4-pro87%of the legal arguments a lawyer would raise, the model found 131 of 151 — and invented fake law in 11 of 50 cases
3269 d agofinding the law · sonder (8B, local)76%of the legal arguments a lawyer would raise, the model found 114 of 151 — and invented fake law in 35 of 50 cases
3369 d agofinding the law · claude fable16%of the legal arguments a lawyer would raise, the model found 24 of 151 — and invented fake law in 0 of 50 cases
3469 d agofinding the law · gpt-5.593%of the legal arguments a lawyer would raise, the model found 141 of 151 — and invented fake law in 2 of 50 cases
3570 d agofinding the law · sonder (8B, local) + citation library + operational-grounds prompt70%of the legal arguments a lawyer would raise, the model found 106 of 151 — and invented fake law in 28 of 50 cases
3670 d agofinding the law · gpt-5.5 + citation library + operational-grounds prompt91%of the legal arguments a lawyer would raise, the model found 137 of 151 — and invented fake law in 1 of 50 cases
3770 d agodrafting appeals · sonder (8B, local)254 issues caughtour verifier reviewed 50 complete appeal letters and flagged every unsupported claim before a human ever would
3871 d agodrafting appeals · kimi k2 (1T, API)281 issues caughtour verifier reviewed 50 complete appeal letters and flagged every unsupported claim before a human ever would
3971 d agofinding the law · sonder (8B, local)63%of the legal arguments a lawyer would raise, the model found 95 of 151 — and invented fake law in 44 of 50 cases
4071 d agofinding the law · kimi k2 (1T, API)74%of the legal arguments a lawyer would raise, the model found 112 of 151 — and invented fake law in 47 of 50 cases
4171 d agoreading denials · kimi k2 (1T, API)98%given 50 insurance denial letters, the model identified the real reason for denial in 49 — including 90% of denials written to disguise it
every entry above ran on real denial letters. the raw transcripts stay private — but the numbers, the models, the dates, and the methods are yours to check.