50+ Years of Cognitive Research. Now Powering AI Memory.
Eternal Recall isn't built on buzzwords or Silicon Valley trends. It's built on ACT-R—one of the most thoroughly validated theories of human cognition ever developed.
While other AI memory systems use simple database timers ("delete after 30 days") or opaque algorithms you can't inspect, Eternal Recall implements the actual mathematical models that cognitive scientists use to understand human memory.
The result? An AI memory system that behaves like human memory—because it's based on the same principles.
ACT-R (Adaptive Control of Thought—Rational) is a cognitive architecture developed principally by Dr. John R. Anderson at Carnegie Mellon University. It represents over 50 years of research into how humans think, learn, remember, and forget.
ACT-R models how humans recall "chunks" of information from memory and how they solve problems. The key insight is that memory strength isn't binary (remembered vs. forgotten)—it exists on a continuous spectrum of "activation" that changes over time based on use.
ACT-R has been used to model hundreds of psychological experiments and has been validated across domains including:
ACT-R divides human knowledge into two fundamental types:
"Knowing That"
Facts you can consciously recall and articulate. Stored as "chunks" of information.
"Knowing How"
Skills and processes you execute without conscious thought. Stored as "production rules."
The central concept in ACT-R is activation—a numerical measure of how accessible a memory is at any given moment. Higher activation means the memory is more likely to be recalled.
A(t) = Total activation at time t
B(t) = Base-level activation (recency + frequency)
Σ wⱼsᵢⱼ = Spreading activation from related memories
This equation captures two fundamental truths about human memory:
Other AI memory systems treat memories as binary: either you have access or you don't. ACT-R—and Eternal Recall—treats memory as a spectrum.
Important, frequently-used memories stay accessible. Rarely-used memories fade but are never deleted. Just like your brain.
The base-level activation of a memory reflects its history of use. Every time a memory is accessed, it gets a boost. Over time, that boost decays—but never completely disappears.
tₖ = Time since each access of the memory
d = Decay rate (typically 0.3-0.5)
β = Base constant
Memory accessibility over time follows a predictable curve:
Memories fade but never fully disappear. Each use boosts them back up.
When you think about one thing, related things become easier to recall. ACT-R models this with spreading activation—activation flows through connections between memories.
Imagine you're thinking about "Python." Activation spreads to connected concepts:
If you've often discussed Python debugging together, that connection is strong. Concepts you rarely connect have weaker associations.
Our Knowledge Graph implements spreading activation:
This is why Eternal Recall feels intuitive—it retrieves context the same way your brain does.
ACT-R was developed to model human cognition—to understand and predict how humans think, learn, and remember. It's been used for:
I'm the first to apply ACT-R's memory equations to AI persistent memory.
Instead of modeling how a human remembers, I used these equations to give an AI a human-like memory system. The activation, decay, and spreading activation that cognitive scientists validated over 50 years—now powering how your AI remembers you.
ACT-R equations were optimized for human memory. They capture:
These are exactly the properties you want in an AI memory system.
Eternal Recall launches on Kickstarter in January 2026