SiliconPulse 1.0.0

The first release of SiliconPulse, a native macOS monitor for Apple Silicon and local AI workloads.

• Overview: the Mac's state at a glance. CPU, GPU and unified memory with live charts; memory pressure, thermal state and swap; the active local model; and the busiest processes.
• Local AI: detects oMLX, Ollama, LM Studio, llama.cpp, MLX and OpenAI-compatible servers on this Mac. It shows the loaded model, its memory, whether it is generating, and the runner's own CPU, memory and GPU time. Tokens/sec appears only when a runner actually reports it, and system GPU is never attributed to a model.
• Processes: a native table with filters, search, high-usage highlighting, AI-process marking and an inspector.
• Memory, Compute, Disk & Network and History screens, with a 60 s, 5 min or 30 min window.
• Menu bar: a live summary plus a popover with the machine's health, CPU, GPU and memory, and the active model.
• Honest numbers: every value has a quality class (exact, reported, calculated, estimated or unavailable) and a scope (whole Mac, runner process or model), shown when you hover it.
• Lightweight: one sampling loop that slows down when nothing is on screen; under 1% CPU while it sits in the menu bar.
• Private: no telemetry, accounts or uploads. The only network requests go to AI runners on 127.0.0.1.
• Opens at login if you want it to, starting quietly in the menu bar.

Install

Requires an Apple Silicon Mac with macOS 14 or later. Open SiliconPulse.dmg and drag SiliconPulse to Applications.

This build is not notarized yet, so macOS blocks it the first time. Open it once from Applications, then go to System Settings › Privacy & Security and click Open Anyway (or run xattr -dr com.apple.quarantine /Applications/SiliconPulse.app).

Checksums are in SHA256SUMS.txt.