We’re releasing Jev 1.14.0 along with experimental features for dreamers
Points and comments are a snapshot, not live.
TypeSafe releases Jev 1.14.0 and experimental features like images and longer context.
TypeSafe released Jev 1.14.0 with incremental improvements, strongest in multilingual support. Experimental features include composable image support via `TypeSafe-Danger-Zone-Images`, a `box` type returning single bounding boxes, and extended context via `TypeSafe-Danger-Zone-Long-Context`. These are flagged as 'danger zone' features not production-ready. The team shared internal benchmark comparisons against open-weight models, noting bias toward their own models. Jev-latest will update to 1.14.0 in one week; 1.13.0 will remain available with 60 days deprecation warning.
What commenters are saying
Commenters split on whether TypeSafe is genuinely cautious or strategically holding back features to mask lack of moat. Some argued the incremental update seems like a response to competitors like SubQ's longer context and image processing. Many shared practical use cases: automatic image tagging, classification in agentic pipelines, inlining as SQL functions, and routing effort levels for reasoning models. Several noted Jev is cheaper and faster than LLM alternatives for structured decisions. Others cautioned that public benchmarks are unreliable and urged running personal evals.