Technology, Systems, & The Ideas Shaping Them
A venture partner conducts due diligence on a high-flying enterprise AI startup, only to find the entire platform runs on five daisy-chained M4 Mac Studios inside a copper-lined closet cooled by a box fan from a big-box store.
A small local model is fast, cheap and wrong often enough to matter. How to build systems that check probabilistic output before acting on it: deterministic validation firewalls, escalation when models disagree, and knowing which answers deserve a second opinion.
Most AI traffic is routine and does not need a frontier model. A local-first, cloud-resilient architecture for small and mid-sized businesses: everyday queries on local hardware at near-zero marginal cost, rare complex ones sent to Cloud Run behind a Semantic Firewall.
Treating intelligence as a homogeneous cloud resource creates ruinous latency and cost. An architectural blueprint for tiered local SLMs, with FunctionGemma as an L0 router and SQLite context stores on Apple Silicon.