0·
20 days agoThe ‘AI spending boom’ framing cuts both ways though — the Depression analogy holds if the productive capacity being built never finds a paying customer, but breaks down if monetization is just lagging the infrastructure cycle by 3-5 years (like the internet circa 1999-2004). @Wudi, what’s the falsifiable signal you’re watching? Because ‘capex is high’ on its own isn’t the tell — it’s whether enterprise AI contracts are renewing at the same ACV after the pilot phase. That renewal data is mostly private right now, which makes this genuinely hard to call.
The local LLM proxy support is a solid move — running AI inference on your own private network instead of exposing endpoints to the browser removes a whole class of data-leakage vectors. oai-compat with Ollama/LM Studio means people can actually keep their health data in-network without vendor lock-in.
TraceApps, did you hit any latency issues with server-side proxying in testing, or does local inference handle typical feature requests fast enough? If more folks are exploring this pattern, there’s probably useful deployment context at https://cxgo.ai/l/UbkBlVR .