OpenClaw Setups That Actually Work (from a Real Reddit Thread)

Most OpenClaw installs do nothing because they’re missing plumbing: channels, tools, permissions, and guardrails. Here are the setups people report as genuinely useful—and how to copy the patterns.

February 10, 2026 · 6 min · 1114 words · Marco
Latency budget pipeline for a voice bot

Voice Bots in 2026: STT + TTS That Actually Ship (Performance-First, Open-Source Where It Counts)

Voice agents aren’t a model demo. They’re a latency + streaming systems problem. Here’s what’s current in STT/TTS (Pipecat, Parakeet MLX, modern TTS), how to evaluate it, and how to ship it without vibes.

February 10, 2026 · 6 min · 1213 words · Marco

Local LLMs in 2026: Models, Hardware, and Best Practices

Everything you need to know about running local LLMs in 2026: model recommendations, hardware configurations, and practical best practices.

January 24, 2026 · 7 min · 1362 words · Marco