Subconscious raises $5.1M to cut the cost of long-running AI agents
What's the deal? SubconsciousDealroom has a profile for this one. Try Dealroom → has raised $5.1 million in a seed round to launch an inference platform built for long-running AI agents. MassVenturesDealroom has a profile for this one. Try Dealroom → led the round, with participation from Foothill VenturesDealroom has a profile for this one. Try Dealroom →, Underscore VCDealroom has a profile for this one. Try Dealroom →, E14 FundDealroom has a profile for this one. Try Dealroom →, and the Agent FundDealroom has a profile for this one. Try Dealroom →. The platform is available today for developers, with an on-premise package for enterprises.
What's the endgame? Born out of MIT research, the system uses dynamic context compression and caching to run agents faster, longer, and cheaper — with no changes to underlying hardware, models, or apps. For agents that consume beyond 200,000 tokens, it can extend the effective context window past five million tokens and cut costs by up to 80%.
Why now? Agents are the most computationally intensive — and expensive — way to use AI models, processing thousands of times as many tokens as a chatbot. Subconscious believes agents will make up virtually all inference in the near future.
Does it work? One 20-person engineering team switched from ClaudeDealroom has a profile for this one. Try Dealroom → to the GLM 5.2 model hosted on Subconscious in July, cutting monthly AI spend from $40,000 to $6,000 while reporting faster throughput and no loss in model capability. On the DeepSWE coding benchmark, GLM 5.2 on Subconscious solved 46% of problems at an average cost of $2.79, versus 44% and $3.92 on standard infrastructure.
What could go wrong? The pitch rests on aggressive compression not degrading output — a hard promise to keep across every workload. Early customers report no capability loss, but broader adoption will test that claim.
"Open models finally got good enough this summer that their quality vs closed source models stopped being a compromise," said Jack O'Brien, chief executive officer of Subconscious. "Teams need to spend less, but engineers are addicted and there's no going back."
The signal: The round sits in the top 5% of all-time US telecom seed deals by size, a sign of investor appetite for infrastructure that makes agents economical. As AI costs climb and teams cap per-developer spend, the fight is shifting from model quality to the cost of running them at scale.
Read more: gamesbeat.com
Image credit: NeoSpire
