AI and Spaced Repetition: Surface What You've Learned Before You Forget It

In 1885, Hermann Ebbinghaus mapped the forgetting curve and proved something uncomfortable: within 24 hours of learning something new, most people retain less than half of it. The science of spaced repetition was built to fight this curve. AI is now extending that fight into your everyday work knowledge — not just the things you study deliberately, but everything you learn on the job.

The Forgetting Curve and Why It Matters for Knowledge Workers

Ebbinghaus spent years running memory experiments on himself, memorizing and testing nonsense syllables under controlled conditions. His findings were stark: memory decays exponentially without reinforcement. The curve drops steeply in the first 24 hours and then levels off, eventually reaching a stable floor — but only if you review the material again before it hits zero.

The insight that emerged from his research: timing your reviews strategically is far more efficient than simply repeating material more often. Review something once right as it's starting to fade, and the next forgetting curve resets at a higher baseline. Each subsequent review extends the interval before the next one is needed. This is spaced repetition — and it's one of the most well-validated findings in cognitive science.

What's less often discussed is what this means for knowledge workers. The Ebbinghaus research was mostly conducted on deliberate, isolated facts — the kind you might study for an exam. But knowledge workers aren't memorizing syllables. They're absorbing a continuous stream of: meeting outcomes, client preferences, project decisions, technical details, feedback from stakeholders, lessons from mistakes. None of it is flashcard-sized. All of it is relevant to future work. And almost none of it gets reviewed systematically.

The result is predictable. You have a useful conversation with a potential partner, take a few notes, and move on. Three months later, when a relevant opportunity appears, you have a vague sense that you talked to someone about something like this — but the specific detail you need is buried somewhere in a Notion page you haven't opened since.

Traditional Spaced Repetition: What Anki and RemNote Actually Do

Apps like Anki, RemNote, and Mochi are built specifically for deliberate spaced repetition. They use algorithms — most commonly a variant of SM-2, developed by Piotr Wozniak in the 1980s — to schedule when you should see each card again based on how well you recalled it last time. Easy recall means a longer interval. Difficult recall means you see the card again sooner.

These tools are remarkably effective for the things they're designed for:

• Learning a new language (vocabulary, grammar rules, example sentences)
• Medical education (anatomy, drug interactions, diagnostic criteria)
• Standardized exam preparation (bar exam concepts, CFA formulas, coding patterns)
• Any domain where facts need to be reliably retrievable under pressure

The limitation is that these are tools for deliberate study. They require you to create the cards yourself — deciding what's worth remembering, phrasing it correctly, and consistently showing up to review sessions. For students and professionals in highly testable domains, this investment pays off enormously. For everyone else, it often doesn't get done.

More importantly, the information that matters most in professional life often doesn't fit into a card. "The Q2 roadmap discussion where Sarah pushed back on the timeline for reasons we should account for in the next planning cycle" is not something you can reduce to a question-and-answer pair. It's contextual, multi-threaded, and only becomes relevant when a specific set of future circumstances brings it back into view.

How AI Extends Spaced Repetition to Work Knowledge

This is where a different approach becomes valuable. Instead of asking you to create and review cards, AI can watch what you're working on and surface relevant past information when it's contextually timely — which is, in effect, a passive form of spaced review.

When REM Labs reads your last 90 days of Gmail, Notion, and Calendar data, it builds a model of what you know, what you're working on, and what's at stake. Each morning brief is curated around that model. If you had a conversation three weeks ago that's relevant to something on your calendar today, the brief surfaces it. If a note you wrote six weeks ago connects to a thread that's active in your inbox, you see that connection.

This isn't exactly the same as spaced repetition — you're not being tested, you're not producing a rating, and the intervals aren't algorithmically determined by your recall accuracy. But the functional effect is similar: information you might otherwise forget is brought back to your attention at a moment when it's useful.

Deliberate Study vs. Passive AI Surfacing

It's worth being clear about what AI surfacing can and can't do compared to deliberate spaced repetition, because conflating them leads to unrealistic expectations in both directions.

What deliberate spaced repetition does better

If you need to reliably recall something under pressure — the kind of recall a doctor needs for drug dosages or a lawyer needs for case citations — there's no substitute for deliberate, tested practice. Spaced repetition apps force active retrieval, which is a significantly more powerful memory consolidation technique than passive review. Reading your notes is not the same as being tested on them.

For any domain where you need fast, accurate retrieval of explicit facts, Anki-style spaced repetition is still the gold standard. AI surfacing is not a replacement for this kind of deliberate practice.

What AI surfacing does that deliberate study can't

Deliberate study requires you to decide in advance what's worth remembering. Professional work is full of information that you don't know is important until later. You can't create an Anki card for something you don't yet know will matter. AI that reads your context doesn't have this limitation — it can surface something you wrote without any intention of preserving it, because a new situation makes it relevant.

AI surfacing also works on the full complexity of professional knowledge — messy, narrative, multi-party, context-dependent information that doesn't fit into cards. The note you took in a stakeholder meeting isn't a fact to be drilled. It's a piece of context that needs to surface at the right moment. That's a job for contextual AI, not flashcard algorithms.

A Practical Workflow for Maximizing Learning Retention at Work

The most effective approach combines both techniques, applied to what each does well.

Layer 1: Deliberate spaced repetition for durable facts

If you're learning a new technical skill, studying for a certification, or trying to internalize a body of knowledge that has explicit, testable components — use Anki, RemNote, or a similar tool. Make cards while the material is fresh. Show up for reviews. The time investment is worth it for durable knowledge you need on demand.

Layer 2: Consistent capture for professional context

The hardest part of AI-assisted surfacing is that it can only work with what's been captured. Good capture habits dramatically increase what's available to surface later. Practical habits that compound over time:

• Write a brief note after important meetings — even just two or three sentences about decisions made and next steps
• When you learn something worth remembering on a call or in a document, paste a quick note into Notion
• Reply to important emails with a summary line so the thread captures the decision, not just the discussion
• Let your calendar be honest — add a note to meeting entries when context matters

None of this requires a formal system. It just requires the habit of writing things down somewhere that REM Labs (or any AI with access to your tools) can find them later.

Layer 3: Let the morning brief do the retrieval for you

Once your context is being captured across Gmail, Notion, and Calendar, the morning brief handles the retrieval layer. You don't need to remember to review notes from six weeks ago — the system surfaces them when they're relevant to what's happening today. This is the passive spaced review loop, and over time it meaningfully changes how much of your past work informs your current decisions.

Layer 4: Use Q&A to actively probe your own knowledge

Beyond the morning brief, asking direct questions to your AI memory layer acts like a more active retrieval practice. "What did we decide about the API design in February?" or "What's my history with this client?" forces the system to pull specific context and forces you to engage with the answer. It's not the same as Anki-style testing, but it's a stronger form of review than passive reading.

The Long Game: Memory That Compounds

Here's what changes when you take learning retention seriously, across both deliberate practice and AI-assisted surfacing: your work compounds in ways it otherwise wouldn't.

Most professionals are, without realizing it, relearning the same lessons repeatedly — discovering the same client preferences, making the same planning mistakes, reinventing the same solutions — because the knowledge gained in past work isn't available when they need it. The cost of this is invisible because it's a counterfactual: you never see the decision you would have made differently.

Spaced repetition, in its original Ebbinghaus form, was about making study efficient. AI-assisted surfacing is about making professional experience efficient — turning the knowledge locked in your past 90 days of work into something that's actually available to you when the next relevant situation appears.

The forgetting curve is real, and it applies to everything you learn on the job, not just the things you study deliberately. The most effective knowledge workers find ways to fight it on both fronts.