Subconscious raises $5.1M for long-running agents

Subconscious raised $5.1 million to build an inference platform for long-running AI agents. GamesBeat reports that its dynamic context compression and caching can compress up to 95% of context and cut costs by up to 80%, claims that remain company-reported. Teams tracking the storage cost of long-running agent history should file this as an inference-layer bet and test quality, cache behavior, and spend on their own traces before switching runtimes.