Clide vs Gemma 4 Multimodal Fine-Tuner

Clide is a native macOS terminal app that rethinks the terminal experience for the agent era. Instead of bolting AI onto an existing terminal, Clide builds around it: an AI pair-developer lives in a side panel alongside a customizable grid of up to 6×6 terminal panes. The AI can read terminal scrollback, preview files, and execute commands into any pane—with user confirmation—making it a genuine collaborator rather than a glorified autocomplete. Built with SwiftTerm, AppKit, and SwiftUI (explicitly not Electron), Clide is genuinely native—fast, memory-efficient, and system-integrated. Drag files from Finder into the AI chat, use the screenshot HUD to share visual context, speak commands via voice input, and rely on workspace memory that persists across sessions. Zero telemetry. Free. What separates Clide from tools like Claude Code or Cursor is its terminal-centric model: rather than AI owning the editor and calling a shell, Clide keeps the shell primary and lets the AI reach into it. For server-side developers, sysadmins, and anyone who actually lives in a terminal, this architecture is more natural and less footprint-heavy than spinning up a full IDE for AI assistance.

Gemma 4 Multimodal Fine-Tuner is an open-source toolkit that lets developers fine-tune Google's Gemma 4 and 3n models across all three modalities — text, images, and audio — using only Apple Silicon hardware. It runs natively on PyTorch with Metal Performance Shaders (MPS), bypassing the NVIDIA requirement that has historically blocked Mac users from serious local fine-tuning work. The toolkit handles the full training pipeline including dataset prep, LoRA adapters, and multi-modal data collation. It ships with working example notebooks, a validation suite, and clean abstractions that don't require deep familiarity with the underlying MPS stack. Apple Silicon's unified memory architecture actually helps here — large multimodal batches fit in memory that would otherwise require GPU VRAM splitting on CUDA setups. Posted to Hacker News on April 7 as a Show HN, it pulled 109 upvotes and 165 GitHub stars within hours. The timing is sharp: Gemma 4 just dropped days ago with new multimodal capabilities, and the community immediately wanted local fine-tuning. This fills that gap faster than Google's own tooling.