Essays on biologically-inspired memory architecture, cognitive AI, and the future of agent intelligence.
We ran our engine on LoCoMo, scored below the published leaders, and published the number anyway. The diagnosis — answering was already at parity; retrieval coverage was the gap — led to fact extraction, dual-channel retrieval, a fabricated-dates bug, and the negation problem nobody in memory retrieval talks about.
Read the engineering story →Recall benchmarks test whether AI memory remembers. TWIST tests whether it intervenes correctly when what was said stops being true — and stays silent on look-alike safe cases. Now a published paper (arXiv:2609.28575) with a human-validated 161-item key (κ = 0.85) and 13 baseline configurations. None pass — including our own product.
Read the release →Every AI assistant you've ever used has the same fatal flaw: it doesn't know how to forget. The solution to better AI memory isn't better storage — it's better forgetting. An introduction to the four cognitive layers of a digital mind.
Read essay →A response to Forbes Technology Council's argument that institutional memory is the missing piece in enterprise AI. The diagnosis is right. The prescription — better documentation discipline — isn't. The real answer is memory as infrastructure, not habit.
Read essay →Two major publications — Forbes and Avataar Ventures — argued in the same month that organizational memory is the bottleneck for enterprise AI. Neither proposed a technical solution. This is the case for cognitive memory infrastructure as the next trillion-dollar category.
Read essay →The largest data platform in the world just published its thinking on AI agent memory. It's a great article about storage. It's also proof that the industry is still missing the point — memory isn't a storage problem. It's a cognition problem.
Read essay →In a six-week window, Cloudflare, Google, LinkedIn, and xAI all shipped memory products. Kapex's response correctly names three of the four dimensions that matter — salience, decay, safety. This is the case for the fourth: metacognition.
Read essay →