Blogs

• Do macOS agents hijack your screen? Benchmarking focus-free computer use — How we benchmarked three macOS computer-use agents on a real, logged-in Mac, measuring a dimension every public benchmark ignores: whether the agent steals your cursor and foreground while it works.
• Why concurrency didn't speed up our macOS accessibility-tree reads (and how we proved it) — We built a concurrent prefetch to read a Mac's accessibility tree faster. Benchmarking showed it was ~20-26% slower. Here's the full investigation: the traversal benchmark, the parallelism probe, a GIL bracket test, and a multi-app control that pinned the real cause on the target app's single main thread.
• LiveKit vs Vapi, and why you should use LiveKit — Why we moved our real-time voice AI agent from a hosted voice platform to an owned LiveKit loop, and what each layer of the stack actually owns.
• Building Performance-Optimized Knowledge Bases for Real-Time AI Agents — How we designed ingestion and query paths for sub-second knowledge base retrieval on live voice calls.