Click any tag below to further narrow down your results
Links
TypeSafe released Jev 1.14.0 with a split approach: stable improvements for production use and experimental "danger zone" features (image support, bounding box type, longer context) for developers willing to accept less reliability. The company is also sharing internal intelligence metrics while acknowledging their limitations.
- Image support is now composable across Jev's JSON interface (state, instructions, criteria), though multimodal performance is weaker than text-only and requires careful prompt decomposition.
- New "box" type returns bounding boxes without type errors, representing the first Jev type designed beyond binary decisions and signaling exploration into broader System One Model capabilities.
- Jev supports longer context lengths than previously enabled in production; developers can opt in via a danger zone flag, but TypeSafe intentionally limits context by default because reliability degrades at scale.
- Internal benchmarking data is being shared for the first time, limited to open-weight models only, with explicit caveats that the methodology will evolve and developers should run their own evals.
A new category of AI models that pick answers instead of generating them launched in September 2026 and immediately spawned dozens of competitors. The article cuts through the noise by laying out practical criteria for choosing which decision model to use in production rather than tracking which specific model is "best."
- Jev launched September 15 and hit massive adoption (13% of Vercel's paid customers integrated it within a day), but within two weeks Amazon, OpenAI, Cloudflare, and others shipped competing models with different tradeoffs.
- DecideBench, an independent benchmark testing 400 multiple-choice decisions across eight task families, shows Jev at 98% accuracy for $32 per million tasks, but smaller open models like Laya are 35-59% accurate and much faster—accuracy-speed tradeoffs matter depending on your use case.
- The real decision criteria are deployment location (hosted vs. self-hosted), input types (text-only vs. image support), latency requirements, cost at your actual volume, and API compatibility for future switching—not which model has the best benchmark score.
TypeSafe AI's decision model Jev autonomously completed Pokémon Red by reading game memory, evaluating legal moves with game facts, and making every decision—from battle tactics to menu choices—over 37 hours and 16,150 decisions. The entire run cost $1.65 in API tokens, with full source code and setup instructions open on GitHub.
- Jev made every game decision (battles, catching, team swaps, navigation) based on legal options and contextual facts, while the harness only handled emulation and button presses—no cheats or memory writes.
- The project cost $1.65 total at $0.042 per million input tokens, with typical calls around 1,200 tokens and median decision time of 0.4 seconds; the full run consumed ~39.2M input tokens.
- All code is open-source (GPL-2.0) and reproducible locally with your own legally obtained Pokémon Red ROM, Node 20+, and macOS or Linux—no Nintendo assets or ROMs are distributed.
Jev is a new AI model that only outputs structured data instead of human language, making it dramatically faster (70-500ms vs seconds) and enabling real-time applications like playing Doom. The author argues this speed advantage could become a new computational primitive for AI, though he suspects competitors can replicate it using simpler inference tricks on existing models.
- Jev generates all structured output in a single forward pass instead of token-by-token, achieving 70-500ms response times compared to seconds for standard LLMs, fast enough to play real-time video games.
- The speed advantage likely doesn't require novel model architecture—you can achieve similar results by prefilling responses and generating only one constrained token with existing LLMs, suggesting Jev lacks a substantial technical moat.
- Structured output could unlock entirely new use cases beyond chatbots by injecting "100ms worth of dirt-cheap intelligence" at decision points throughout applications, though Jev won't match frontier LLMs in raw capability.