All the latest feature releases, model additions, Apple Silicon optimizations, and studio improvements.
v0.4.0
LATEST RELEASE
Major workstation milestone introducing Plan Mode for autonomous multi-step execution, the persistent 3-tier Brain Engine and neural visualizer, Apple Silicon Metal P-core scheduling optimizations, compatibility patches for Liquid Foundation Models and LoRA adapters, the iRun-Link remote CUDA host bridge, and an intriguing glimpse into the future.
✨ Key Features & Improvements
Plan Mode & Roadmap Engine: Automated milestone planning, dynamic task dependency graphs, recursive sub-task execution, and granular effort metrics with interactive status cards.
The Brain Engine & Neural Visualizer: 3-tier memory system (Short-term, Long-term, Archival) with contextual recall, live node-link neural visualizer, and intelligent web ingestion without prompt bloat.
Apple Silicon Metal P-Core Scheduling: Optimized Metal dispatch routines prioritizing high-performance P-cores, smart prefix KV caching, and zero-crash multi-tab collections.
Compatibility Patches & Tool Protocols: First-class support for Liquid Foundation Models (LFM), Pythonic tool calling protocol, dynamic LoRA adapter hot-swapping, and refined Markdown rendering.
iRun-Link CUDA Host Bridge: Connect your iPhone, iPad, or Mac to a Windows or Linux CUDA host over LAN with dynamic random-port pairing and 7-digit ephemeral key verification.
What's Next:Custom LLM on the way, built fully from the ground up.
v0.3.0
PREVIOUS RELEASE
Comprehensive workstation update adding nested subagents, thinking and functioning loops, code/chat completion, expanded context architecture, font customization, light themes, and major stability patches.
🚀 Highlights & Additions
Nested Subagents: Multi-level subagent orchestration and real-time execution telemetry on the 120Hz ProMotion Subagent Bar.
Thinking & Functioning Loops: Autonomous execution loops alternating between plan, build, and bughunt.
Code & Chat Completion: Fast on-device autocompletion, streaming token generation, and syntax-aware formatting.
Expanded Context Architecture: Context preservation techniques that deliver deep memory and recall even with smaller context windows.
Font Customization: Full typography selector across SF Pro, JetBrains Mono, New York (Serif), Inter, and Menlo.
Light Themes Added: New light glassmorphic palettes including Sakura Frost and Ocean Deep alongside dark themes.
Categorized Model Browser: Dedicated Language, Vision, and Logic tabs with 1-click Hugging Face GGUF downloads and real-time web_search.
Major Patches & Bug Fixes: Fixed multi-turn conversation memory leaks and optimized Apple Silicon Unified Memory kernel speeds.
v0.2.0
Introduced the React harness, improved tool calling protocols, initial Design tab with 2D and 3D spatial generation, and enhanced lightweight system prompts.
🎨 Key Features
React Harness Integration: Live component preview harness and interactive UI sandboxing.