Hi, this is Ray.

I want to start by describing a specific pattern I fell into for years without recognizing what was happening. I would make a mistake on something (a wrong answer on a practice test, a bug in code that broke my project, a strategic misjudgment in a decision that mattered) and I would notice the mistake, acknowledge it, feel bad for a moment, and then... move on. The mistake happened. I saw that it happened. I told myself I would do better next time. I moved to the next task.

The result of this pattern, sustained across years, was that I kept making the same categories of mistakes. Different specific instances, same underlying patterns. I would think I had learned from an error because I had noticed it and moved on, but the "learning" was so shallow that it produced no actual change in my subsequent behavior. Six months later, in a slightly different situation, I would make functionally the same mistake again. Notice it. Feel bad. Move on. Repeat.

The turning point came when I encountered the medical education literature on diagnostic error analysis. According to a recent study of medical students, a cohort of 65 final-year medical students participated in a structured three-phase educational intervention comprising preparation, case study analysis, and reflection. Students examined 20 diagnostic error case studies to identify contributory factors, such as cognitive biases, atypical presentations, and systemic barriers. The medical education field has been forced to take error analysis seriously because in medicine, the cost of not learning from mistakes is measured in patient outcomes. What they've figured out about structured error analysis is genuinely useful for anyone whose learning involves making mistakes… which is, if we're honest, every learner.

The pattern I'd been in wasn't unique to me. It's the default pattern for most learners, and it produces exactly the outcome I was experiencing: the appearance of learning from mistakes without the substance of actually learning from them. Today's newsletter is about the alternative. A specific framework for analyzing mistakes deeply enough to produce genuine learning, why surface analysis leaves most of the value on the table, and how to actually apply this in your own study practice. This is the second newsletter I'm writing on mistakes… the first covered the neuroscience of error signaling. This one is the practical framework for what to do when the neuroscience fires. Let's get into it.

Scheduling meetings isn't why you get out of bed. Filing expenses isn't your calling. So hand it all to Catch — the AI admin that takes the tedious work off your plate and gets it done.

Just ask. Catch handles the back-and-forth on scheduling, flights, restaurants, follow-ups, vendors, clients and more — then comes back with it sorted, in moments.

No guesswork. No dropped balls. What's yours stays yours, private by default and reliable by design.

Nobody became a leader to book their own flights. Meet your admin savior and go from swamped to sorted.

Get started free at catchagent.ai — and start spending your time on the things only you can do.

Get Started Now

Let me start with what the research shows about the depth of error analysis, because this is where most learners fall short.

According to research on error analysis quality, deep error detection may involve a thorough awareness of error-related aspects, such as identifying the location, type, and reason for errors. It is deeper if students can connect the current errors with those past but similar. The quality of learning processes related to error detection, such as metacognitive monitoring and comparing, is more likely to be low if students merely write superficial error reasons, such as calculation mistakes. Read that carefully. The dividing line the researchers draw is between surface descriptions ("calculation mistake") and deep analysis (location + type + reason + connection to past similar errors). Surface analysis produces almost no learning. Deep analysis produces the actual change in future behavior that we call "learning from mistakes."

The mechanism worth understanding involves what the research calls the SOLO taxonomy… Structure of the Observed Learning Outcome. In this framework, the depth of your engagement with a mistake corresponds to the depth of the learning you extract from it. Someone who just notes "I got it wrong" has done shallow error processing. Someone who identifies the specific type of error, traces its cause, connects it to their broader knowledge, and modifies their mental model has done deep error processing. Same mistake, different processing, different learning outcomes.

This connects to a foundational finding from learning-from-errors research. According to a comprehensive framework, our understanding of learning from errors includes a detailed analysis of the error causes in order to identify and explain potential misconceptions, a self-evaluation of the underlying knowledge and its modification, as well as the correction of the error in question. The key phrase is "underlying misconceptions." Real learning from errors requires identifying not just what went wrong but WHY it went wrong at the level of your mental model. The mistake is a symptom. The misconception is the disease. Surface analysis treats the symptom. Deep analysis addresses the disease.

Based on the research and my own experience refining this over years, here's the framework I now use for analyzing any significant mistake. I think of it as an autopsy… a systematic examination of what happened and why, before moving on. Each layer goes deeper than the previous one, and skipping layers costs you learning.

Layer 1: What specifically happened?

This is the descriptive layer. What was the mistake, precisely? Not "I got it wrong." Not "I made a calculation error." The specific mistake, described in enough detail that you could show someone else exactly what happened. What was the input? What was your response? What was the correct response? Where specifically did they diverge?

Most learners stop before completing this layer. They notice a wrong answer without actually looking closely at what specifically was wrong. The specificity matters because the subsequent layers depend on it. In Metal Gear Solid, Snake doesn't just note "I got detected." He examines the specific sensor, the specific angle he was at, the specific action that triggered the alert. The specificity is what enables the subsequent analysis.

Layer 2: What type of mistake was this?

Not all errors are the same category. Different types call for different responses. Some common types worth distinguishing:

Knowledge errors: You didn't know something you needed to know.
Application errors: You knew the relevant knowledge but applied it wrong.
Perception errors: You misread or misunderstood what the question was asking.
Procedural errors: You skipped a step or did steps in the wrong order.
Attention errors: You knew and could do the right thing but weren't paying attention.
Judgment errors: You made a reasonable-seeming choice that turned out wrong given information you should have weighted differently.
Overconfidence errors: You were sure you were right when you should have checked.
Framing errors: You approached the problem from the wrong angle entirely.

Different types of mistakes require different interventions. If it was a knowledge error, you need to acquire specific knowledge. If it was an attention error, you need to work on focus during similar tasks. If it was a framing error, you need to work on your approach to problem-solving itself. Without identifying the type, you can't target the right intervention.

Layer 3: What's the underlying cause?

This is the mental model layer, and it's where most of the learning actually happens. Beyond the specific mistake and its type, what's the underlying issue that produced it? What did you believe, understand, or assume that led to this specific error? What's the misconception that, if fixed, would prevent this specific category of mistake in the future?

This layer requires honest introspection. Sometimes the underlying cause is a factual gap you can address. Sometimes it's a habit of thinking that produces reliable errors. Sometimes it's an emotional response… wanting a specific answer to be correct, avoiding the harder analysis because it felt uncomfortable, rushing because of external pressure. The underlying cause isn't always cognitive. It's sometimes affective or motivational, and the intervention has to match.

According to error analysis research, errors initiate explanation and reflection processes in which deficient concepts are contrasted with correct concepts in order to establish accurate mental models. The mechanism is contrast. You have to see what you actually think next to what you should think, and let the contrast do its work. Without articulating your actual belief clearly, you can't see the contrast, and the correction doesn't take.

Layer 4: What patterns does this connect to?

This is the meta-layer, and it's the one that produces long-term learning across many similar mistakes. Have you made this specific type of error before? Does this connect to other patterns you've noticed in your work? What does this tell you about your broader approach, your recurring blind spots, your systematic weaknesses?

The research is explicit that this connection layer is where deep learning lives. As one summary put it, it is deeper if students can connect the current errors with those past but similar. The connection turns individual mistakes into pattern recognition. One mistake, analyzed shallowly, gives you information about that mistake. One mistake, connected to past similar mistakes, gives you information about your patterns… which is dramatically more valuable because patterns are what produce future mistakes.

In Attack on Titan, the successful analysts don't just note "Titan broke through the wall." They analyze the specific breach, categorize it against previous breaches, identify the pattern in Titan behavior across events, and update their broader model of what they're dealing with. The single event is data. The pattern across events is understanding. Your mistakes work the same way.

Okay, how do you actually do this in practice without spending three hours per mistake? Here's what works.

For minor mistakes: Just Layer 1 and Layer 2. Take 60 seconds. Note what specifically happened and what type of mistake it was. That's it. The 60 seconds is dramatically more than most learners spend, and it produces meaningful learning over time.

For significant mistakes: All four layers. Take 5-10 minutes. Write it down. Actually write it, because articulation is where the analysis becomes concrete. The writing forces the specificity that mental reflection tends to skip.

For recurring mistakes: All four layers, plus a specific plan for the intervention. If you've noticed this pattern before, the analysis stops being about this specific instance and starts being about the pattern itself. What's the systematic change you need to make? What's the specific practice that would address this?

Keep an error log. This is the practical tool that makes the analysis compound. A running document (physical notebook or digital, whichever fits your practice) where you record your mistakes and their analysis. Over time, the log becomes a map of your patterns. You can look back and see which types of errors you make most often, which categories you've genuinely improved on, and which categories keep recurring. This meta-information is what actually produces long-term change.

Review the log periodically. Weekly is often enough for most learners. Look at the errors you've logged. Notice patterns. Ask what interventions are working and what aren't. The review is when the log becomes actionable rather than just archival.

Let me give some concrete examples of what shifts when you actually do the deep analysis.

Example one: Getting math problems wrong. Surface analysis: "calculation error." Deep analysis: "I consistently make errors in problems involving negative numbers under time pressure, and specifically when the problem has more than one negative in it. The underlying issue is that I never fully internalized negative number rules; I've been mechanically applying them but under pressure I revert to intuitive but wrong operations." That deep analysis points at a specific intervention (spend time actually understanding negatives, not just practicing them) that surface analysis wouldn't have identified.

Example two: A wrong medical diagnosis in the study I cited. Surface analysis: "missed the diagnosis." Deep analysis: "I anchored on the most obvious presenting symptom, didn't consider atypical presentations of a common disease, and my search for information stopped once I had a plausible explanation. The underlying issue is confirmation bias in diagnostic reasoning, which happens especially under time pressure." That deep analysis produces training in specifically resisting confirmation bias, which addresses many future potential errors.

Example three: A relationship conflict. Surface analysis: "I said the wrong thing." Deep analysis: "I responded to what she said based on what I would have meant if I'd said the same words, rather than what she actually meant given her context. This is a pattern I've noticed before… I default to interpreting others' words through my own framework rather than checking what they actually mean. The underlying issue is insufficient perspective-taking under emotional pressure." That deep analysis points at a specific practice (checking my interpretation before responding) that addresses the pattern rather than the instance.

In each case, the deep analysis takes maybe five minutes more than the surface analysis. But the deep analysis produces actual behavior change, while the surface analysis produces the illusion of learning.

Some patterns to watch for that prevent the framework from working.

Emotional overwhelm during analysis. If you're too upset about the mistake to think clearly, wait. Give yourself a few hours or a day. Analyze when you can actually think, not when you're in the initial emotional response. The analysis requires clarity that emotional intensity blocks.

Self-flagellation disguised as analysis. "I always mess this up, I'm just bad at this" isn't analysis. It's self-criticism. Real analysis is about the specific mistake, its type, its cause, and its patterns… not about your worth as a person. If your "analysis" turns into a general indictment, you've stopped analyzing.

Skipping to solutions before completing the analysis. The temptation is to jump immediately to "what should I do differently." But the intervention only works if you've correctly diagnosed the mistake. Rushing to solutions before completing the analysis produces solutions to the wrong problems.

Analyzing only failures, not successes. Some of your most useful information comes from analyzing what went right when things went right. Why did the successful attempt work? What was different from the times it didn't? This positive analysis complements error analysis and produces balanced learning.

Never actually reading the error log. A log you never look at is just documentation. The value comes from the review. Schedule the review. Actually do it.

Here's what I want you to take from all this. The default pattern most learners fall into with mistakes (notice, feel bad, move on) produces the illusion of learning from errors without the substance. The behavior stays the same. The mistakes recur. The frustration builds. Over years, this pattern is one of the most reliable ways to spend a lot of effort without producing much actual improvement.

The alternative isn't dramatic. It's just deeper. Instead of stopping at surface awareness of a mistake, spend a few minutes on structured analysis. Identify what specifically happened. Categorize the type. Investigate the underlying cause. Connect to broader patterns. Log the analysis. Review periodically. This modest addition to your practice produces dramatically different learning outcomes over months and years.

The medical education literature figured this out because they had to. When mistakes have serious consequences, you can't afford to keep making the same ones. The rest of us have been treating our mistakes more casually because the consequences of any single mistake are usually smaller. But the cumulative cost of not learning from our mistakes (across years of study, projects, and skill development) is enormous. The learners who systematically analyze their errors compound their learning at a rate that non-analyzing learners can't match.

If you've been in the notice-and-move-on pattern, please try the framework this week. Pick one significant mistake you make. Actually run the four layers on it. Notice what surfaces. Notice how different your understanding of the mistake becomes when you engage with it deeply rather than glancing at it. Most learners who try this once are struck by how much information they've been leaving on the table.

In Kingdom Hearts, when Sora fails a boss battle, he doesn't just retry with slightly different button-mashing. He examines what went wrong, changes his equipment, adjusts his approach, and tries again with actual learning behind the retry. Your mistakes are boss battles. Analyze them accordingly. The next attempt will be different because you've actually processed what happened, not just noted that it happened.

Keep learning (and keep analyzing),

Ray