The author demonstrates compact Jev-style models built on Qwen3.5 (0.8B–9B), with 4B/9B fitting a 32GB Mac, supporting CUDA and ROCm, and answering a support ticket in 495ms with probabilities.
I RAN A DECISION MODEL ON MY LAPTOP TODAY 1. Kev: tiny Jev-style models built on Qwen3.5, 0.8B up to 9B 2. 4B/9B fit a 32GB Mac, also CUDA and ROCm 3. Answered a support ticket in 495ms with probabilities #OpenSource #LocalAI #ML
The author shares a Jev usage insight: it seemed tricky at first, but clicked when framed as 'not letting the AI write text.' Using inquiry handling as an example, the conventional approach dumps everything on GPT or Claude to analyze, while Jev takes a different approach.
The prototype uses AI to estimate hesitation or interest from cursor back-and-forth movements or lingering before clicking, observing customer demeanor and supplementing explanations or suggesting comparisons.
The author built an e-commerce real-time customer service demo using Jev and gpt-live-1, which instantly suggests recommended products based on the user's speech during a conversation, delivering a great experience.
