Engram exits stealth, nabs $98M

Engram is tackling LLM memory at the architecture level rather than relying on brittle RAG or context window stuffing. While $98M is a massive war chest for a stealth exit, the startup’s promise of 100x token reduction could redefine agentic developer workflows if it scales.

• –By separating reasoning from memory, Engram's continuously updating models can adapt in real-time, avoiding the high cost and latency of traditional fine-tuning.
• –The startup claims up to a 100x reduction in token usage by preparing personalized contexts in advance, addressing a major bottleneck in production-level AI agents.
• –Securing early partnerships with Microsoft, Notion, and Harvey signals deep corporate demand for reliable, secure, and persistent organizational memory.
• –A $98M round backed by Sequoia, Kleiner Perkins, and Andrej Karpathy underscores the industry's belief in the team's ability to solve the "genius stranger" dilemma in AI.