Benchmarks, papers and independent reproductions. Read these when you need to understand behaviour rather than adopt a library.
537 active projects in this collection. All collections →
A streaming event bus whose routing, subscription and consumption are decided by a probabilistic judge. The reference judge is TypeSafe AI's Jev (System One) model: send it a payload and a set of typed questions, get back calibrated probabilities instead of prose.
MCP server exposing TypeSafe's Jev model as a typed evaluate tool.
Text/PDF - File categorization and sorting with Typesafe AI Jev or local calibrated decision model
Chess moves, evaluations, persona opponents, and game classification with TypeSafe AI System One
Behavioral contracts for TypeSafe Jev — pin production expectations, eval model upgrades, catch flips and confidence regressions.
Chain-of-thought and self-refinement for TypeSafe's Jev: feed its typed answers back as state and ask again. Benchmarks vs TypeSafe's own cookbook numbers.
FUn little experiment with Typesafe AI Jev Model playing chess against stockfish :)
What your last session knew, scored against what this one is doing. MCP server: a per-project ledger written as things happen, recalled per task with TypeSafe's Jev evaluation model via Vercel AI Gateway.
I kept watching coding agents burn context on decisions that aren't hard - triage 400 tickets, tag 600 files, route to one of six teams. jev-mode moves those verdicts to a typed-judgment model. I A/B'd it: 78% fewer tokens, 16x less work-attributable input, accuracy 96.1% vs 93.7%. Python, no deps, MIT.
Independent benchmark data for TypeSafe's Jev (System One model) vs LLMs: accuracy, calibration, cost. Boards + per-decision logs, CC-BY-4.0
Train lightweight language backbones for typed decisions and candidate probabilities. CE/Brier training, evaluation, and an offline end-to-end demo.
Jev test bench — Texas Hold'em edition: live-fire testing of TypeSafe's Jev evaluation model through a React poker game (Vercel AI Gateway). MIT.
Agente de trading para o mercado BTC Up/Down de 5 minutos da Polymarket: modelo em código, Jev (TypeSafe System One) como portão, ordens maker, calibração e shadows em paper
TypeSafe Jev demonstration for new analyzation — experimenting with Jev for fast analysis of news and tickers
Open auto mode for AI agents — a calibrated tool-call firewall powered by TypeSafe Jev. Ships as a Claude Code hook
A Jev-inspired decision interface for existing LLMs. Explicit choices, scores, calibration, and review thresholds.
Jev (TypeSafe) 性能評価プロジェクト — 日本郵便 KEN_ALL をマスタに、AI SDK 経由の Jev が住所のあいまい一致にどこまで使えるかを検証
A small reproducible MuJoCo pilot comparing Jev, Claude Haiku, and reactive rules for pick-and-place.
Typed, policy-driven decision workflows on top of TypeSafe AI Jev: confidence routing, fallbacks, evaluation, and RAG patterns for TypeScript apps.
Typed-decision benchmark from PadFlow (land development SaaS): schemas, anonymized labeled rows, and a runner for confidence-calibrated models like TypeSafe Jev.
I tortured Jev into being a RISC-V CPU.
Jev (TypeSafe System One) × ASReview SYNERGY abstract screening demo — Choice/Noul vs gold labels
Self-hostable managed agents harness. Author YAML manifests; Felix compiles them into governed agents with durable fibers, memory, skills, eval, approvals, and sandboxes — over OpenAI, A2A, MCP, and SSE. Fork, rewind, and steer runs. Docker, Helm, AWS, or GCP.
A hands-on educational lab for DeepEval + Jev: inspect real recorded judgments, explore release gates and drift playback, and run fresh evaluations on your own workloads.