MassVentures

Subconscious has raised $5.1 million to commercialize an MIT-derived inference platform that dramatically reduces the cost and latency of long-running AI agents. The platform uses dynamic context compression to cut token consumption by up to 82% in production, with no model or application changes required. For engineering teams hitting AI spending ceilings and regulated-sector buyers needing air-gapped deployments, the timing is pointed.