Perplexity open-sources Decider models that audited 1B tokens across 7,000 agent sessions, yielding 14 fixes
inductionheads · x · 2026-11-13
Perplexity CEO Arav Srinivas touts "Decision Models" as the future of agentic security: models that can process billions of tokens of telemetry, spot security issues, and help patch fixes at unprecedented scale and cost. Perplexity internally used Decider to read 1B tokens across 7,000 agent sessions in one day, uncovering flaws that led to 14 fixes to its agent security rules. Both models are Apache-licensed and free on OpenRouter until Nov 19.
More from coding & agent
- GitHub's practical guide: write an agent skill once, run it on any agent — 0xkarasy · 2026-11-13
- Swap the Copilot harness config to build fully custom agents — 0xkarasy · 2026-11-13
- PostTrainBench releases 58.4GB of agent fine-tuning traces, 151k downloads in a month — maksym_andr · 2026-11-13
- How Claude Code handled a 150K-line C# game codebase: a solo dev's five-part setup — dmarshallgaming · 2026-11-13
- AI prototype unreliable? Scope it down: let the agent do less, then grow — brandon_galang · 2026-11-13
- Porting an Unreal game to iPhone and iPad with Claude Code, no manual Xcode work — chongdashu · 2026-11-13