Building with Jev: Workshop

Your software has a state. It needs a decision—not another chat response.

What is Jev?

It's a new AI model. TypeSafe AI describes Jev as a System One model: pass in software state and questions, receive typed values, probabilities, and confidence scores that software can consume directly.

That changes the integration question: instead of asking only what a model should say, consider what your software needs to decide. Classification, routing, and workflow branching are useful contexts for understanding the distinction—not a claim that Jev replaces general-purpose LLMs.

Why Jev?

TypeSafe reports 40–200× lower latency and up to 400× lower cost on System One-style tasks. LangChain’s early evaluation also found Jev substantially cheaper and faster, with much lower variance on continuous scoring. So the reason to use Jev isn’t that an LLM can’t make the decision.

It’s that you may not want to fire up a language generator every time your software needs an if-statement with judgment.

📅 Date｜ October 3, 2026 (Saturday)
⏰ Duration｜ 4-hour workshop, including a pizza lunch break
📍 Location｜ Second Space + Online
💻 Format｜ Hybrid, join in person or online
🧑‍💻 Level｜ Medium to advanced development knowledge required
🛠️ Workshop｜ Hands-on session, bring your laptop
🍕 Lunch｜ Pizza provided
🗣️ Language｜ English
🎟️ Registration｜ Advance registration required

Read the technical context

• TypeSafe AI: Introducing System One Models & Jev — the model’s design and state-to-decision framing.

• LangChain: Building a Harness with Jev — how structured decisions fit inside an agent loop.

• MindStudio: Jev System One Model Launch — an overview of the contrast with text-generating models.

Register for the workshop and bring your laptop to follow along.

合作單位：高雄市政府青年局