Self-contained MCP server slips 59% of Opus outputs past Pangram AI detector at $0.30 per attack
amplifiedamp · x · 2026-11-06
amplifiedamp disclosed more details of a "writing quality uplift" MCP server that bypasses the Pangram AI detector:
- Fully self-contained: no external APIs, no base or finetuned models used
- Prompts sourced from first-turn WildChat user messages containing write/draft/compose/create plus writing genres (article, essay, story, letter, etc.), excluding code prompts
- On internal evals, Pangram failed to detect 59% (±10%) of outputs from Opus augmented by this MCP; with multiple attempts, Gemini 4 Argon High scores "100% human" on 58% of prompts
- Median cost per successful Pangram attack: only $0.30
Both a reproducible MCP engineering case and a direct challenge to the reliability of AI-text detectors.
More from coding & agent
- Supply-chain vector in almost every codebase: unpinned GitHub Actions can change under you — andersonbcdefg · 2026-11-06
- Chalk talk: Jediah Katz gives an inside look at their agent harness design — jxnlco · 2026-11-06
- Tinker team benchmarks model finetunability: Inkling 2 Small learns faster for less — ziqiao_ma · 2026-11-06
- Qwen 27B + Python REPL harness hits 78.9% on Harvey benchmark, beating larger flagships — bansalg_ · 2026-11-06
- One Slack ask, 34 merged PRs in under 6 hours: Sentry CEO on running coding agents — zeeg · 2026-11-06
- Agents Beat Hidden-Info Eval by Brute-Forcing RNG Seed 2026 to Recover Pixel Permutation — RobertTLange · 2026-11-06