• HumanAI
  • Posts
  • 🪬 HumanAI #020 – The F*ck-Up Library

🪬 HumanAI #020 – The F*ck-Up Library

NASA has one. And after reading this, you might want one too.

Sponsored by

⏱️ Today’s Estimated Read Time: 5 min 12 sec.
🎧 Perfect soundtrack? Play “Once In A Lifetime (by Talking Heads)” in the background while you scroll to set the vibe.

The Dawn of HumanAI

AI is inevitable. It’s happening. 
Betting against evolution and progress is usually a bad decision. 
So, we embrace it and use it to our advantage. 
But we’ll always remember who we are—
HUMANS.

Lucas “Mencho” Menzano

Hey!

It's been a while, hasn't it?

13 months since my last proper HumanAI edition.

THIRTEEN.

I could tell you I've been incredibly busy with work, family, life, etc etc.

Which is true.

But I also know myself well enough to recognize a good old procrastination problem when I see one.

Funny thing is, I actually wrote about this exact problem in edition #010.

Apparently I didn't learn much from it.

Which brings me to today's topic.


👀 In today’s:

  • The dumbest $125 million mistake in space history.

  • Why NASA has spent 50 years collecting other people's screw-ups.

  • An AI experiment that made me rethink how we use ChatGPT.

  • How to create your own F*ck-Up Library (yes, that's what we're calling it).

Let’s go!

đź§  HUMAN INSPIRATION: Houston, we have a measurement problem.

Mars Climate Orbiter (Artist’s Concept)

September 23, 1999.

After travelling through space for more than 9 months, NASA's Mars Climate Orbiter was finally approaching Mars.

The spacecraft had launched in December 1998 with a pretty cool mission: study the Martian atmosphere, track its weather, and help communicate with another spacecraft.

Hundreds of scientists, engineers and technicians had been working on it.

Years of development.

Around $125 million spent on the mission.

And then…

Poof.

Gone.

NASA lost contact with the spacecraft as it attempted to enter orbit around Mars.

What happened?

A solar storm? An alien attack? Someone accidentally unplugged the WiFi?

Nope.

Someone forgot to convert the units.

One team had supplied spacecraft thruster data using imperial units. The navigation software expected metric units.

Pound-force seconds vs. newton-seconds, to be precise.

That mismatch helped send the spacecraft onto the wrong trajectory.

Nine months flying through space, only to get screwed by something you probably learned in school.

I mean… IMAGINE being in that meeting.

Anyway, the investigation also found problems with the processes that should have caught the error. It wasn't just one person making a silly mistake.

And here's what I found fascinating while reading about this disaster.

NASA actually published the lessons from the mission in a system called the Lessons Learned Information System.

It's a real database.

And you can browse it yourself.

Each entry documents what happened and what future NASA teams should learn from it.

But it gets better.

NASA has also been running something called the Aviation Safety Reporting System since 1976.

Pilots, air traffic controllers and other aviation professionals can anonymously report mistakes, close calls and dangerous situations.

As of 2026, NASA has received more than 2.3 MILLION reports.

That's 50 years of people documenting things that went wrong (or almost went wrong) so someone else doesn't have to learn the hard way.

2.3 million.

A gigantic collection of human f*ck-ups.

Now, I have no intention of becoming an astronaut anytime soon.

But this got me thinking…

Why don't we have something like this for our own lives?

Think about it.

You screw up at work.

You figure out what happened.

You learn your lesson.

A few years later, you make another decision, under different circumstances, and somehow manage to screw up in almost the exact same way.

Congratulations!

You've unlocked the same achievement twice.

And it's not always because we're stupid.

Sometimes we simply forget. Sometimes the circumstances look different enough that we don't recognize the pattern.

I can barely remember what I had for dinner on Tuesday.

Expecting my brain to accurately recall every lesson from the past 20 years seems a little optimistic.

Until recently, I wouldn't have known what to do about that.

But we now have a pretty interesting tool at our disposal.

And there's some actual research behind this idea.

⚡ AI REAL-LIFE USE: Teach your AI to learn from your mistakes

In 2023, a group of researchers from Princeton and other institutions published a paper with a very nerdy name:

(Don't worry, I won't make you read it.)

The researchers were exploring a problem with AI agents.

When an AI fails at something, how can we help it avoid making the same mistake again?

Their approach was surprisingly human.

The agent tries something.

It gets feedback.

It reflects on what went wrong.

It writes down a lesson.

And the next time it tries, it has access to those notes.

The interesting part?

They didn't have to retrain the AI model.

They gave it a way to remember what happened and use that information in its next attempt.

On one coding benchmark, called HumanEval, this approach reached a 91% success rate, compared with 80% reported for GPT-4 in the same research.

Now, those numbers come from a specific experiment. It doesn't mean your ChatGPT suddenly gets 91% smarter by keeping a diary.

But the idea is pretty damn interesting.

And I thought…

What if we used this same principle with our own mistakes?

I came up with a little experiment.

I'm calling it:

đź§  THE F*CK-UP LIBRARY

Your personal database of mistakes, bad decisions, expensive lessons and things you'd rather not repeat. PRO TIP: This is a GREAT tool for teams too!! A shared f*ck-up database!

Let's build it!!

🔹 STEP 1: Create your library

Open ChatGPT.

Create a new Project and name it:

The F*ck-Up Library đź’©

Yes. With the poop emoji. I think it's appropriate.

Projects let you keep conversations and reference files together, so your past lessons can be used in future chats.

You can also do this with Claude, Gemini, or whatever AI you prefer. A simple document works too.

🔹 STEP 2: Tell AI about your biggest screw-ups

Here's my favourite part.

Open Voice Mode and start talking.

Tell it about something that went wrong.

A business decision.
A failed project.
A terrible hire.
An argument.
A missed opportunity.
That investment you made because some random guy on YouTube said it would go to the moon.

Whatever.

Don't prepare a speech. Just tell the story.

Then give it this prompt:

📌 PROMPT #1 — THE F*CK-UP COLLECTOR

I want to build a personal library of mistakes and lessons learned.

Interview me about situations that went wrong in my life or work.

Ask me one question at a time to understand what happened, what I was thinking, what information I had available, and what I could have done differently.

Don't just agree with my version of events. Challenge me when necessary.

After each story, create a short entry with:

  1. What happened

  2. What went wrong (and what was outside my control)

  3. Warning signs I missed

  4. What I learned

  5. A simple rule to remember next time

Keep each entry short, specific and useful.

Help me identify recurring patterns across my mistakes.

For example, imagine you once launched a marketing campaign without checking if tracking was working.

Your library might end up with an entry like this:

F*CK-UP #001

The mistake: Launched a paid campaign without testing conversion tracking.

The damage: Spent $2,000 with unreliable conversion data.

What went wrong: Assumed the setup was correct because everything looked fine in the dashboard.

Lesson: Looking good and working properly are two very different things.

New rule: NEVER launch a campaign before testing the full conversion journey yourself.

Easy, right?

🔹 STEP 3: Put your library to work

Now imagine you've collected 20, 50 or even 100 of these.

You have a record of lessons learned from YOUR life.

Next time you're about to make an important decision, instead of asking AI what you should do, ask it what you've already learned.

📌 PROMPT #2 — THE F*CK-UP DETECTOR

I'm considering this decision:

[Explain the situation]

Review my F*ck-Up Library and look for relevant past experiences.

Have I made similar mistakes before?

What warning signs should I pay attention to?

Which lessons from my past could help me here?

Reference the specific entries you used.

If there's no relevant lesson in my library, tell me. Don't force a connection.

Finish with three questions I should answer before making this decision.

Imagine the response:

"Mencho, remember that time you launched a project before validating whether anyone wanted it?"

Oh, right.

That.

Again.

It's almost like having a personal advisor who has witnessed all your stupid decisions.

And unlike your friends, it doesn't get tired of hearing the same story 15 times.

ONE IMPORTANT DETAIL: Keep your library in a file within the Project and update that file as you add lessons. Don't rely on the AI remembering everything perfectly. And be careful with sensitive personal or company information.

🔹 BONUS: F*ck-Up Fridays

Want to take it one step further?

Every Friday, spend five minutes telling AI:

"Here are the things that went wrong this week. Help me figure out if there's anything worth adding to my library."

Maybe some weeks there's nothing.

Great!

Maybe other weeks you'll have enough material for an entire Netflix documentary.

At least you'll get something useful out of it.

🚀 FOR THE BOLD ONES… You're sitting on something nobody else has.

There's something about this whole experiment that I find particularly interesting.

We're all getting access to similar AI tools.

Same models. Same features. Same ridiculously powerful technology.

And people are spending HOURS looking for the perfect prompt.

Which is fine, I guess.

But here's a thought.

I could give you every single prompt I've ever written.

You could give me yours.

But I couldn't give you 38 years of my life.

My experiences.
My mistakes.
The people I've met.
The things I've tried.
The stuff that worked.
And the things that went horribly wrong.

That's information you can't just Google.

It's yours.

And maybe we're overlooking how valuable that information could be when combined with AI.

I don't know if The F*ck-Up Library will make you a better decision-maker.

I haven't tested it long enough to claim that.

But I know one thing: I wish I had started documenting my mistakes 10 years ago.

There are quite a few I'd love to avoid repeating.

Including, apparently, abandoning a newsletter I really enjoy writing for more than a year.

So yeah.

I guess this edition is my first entry.

F*CK-UP #001: Waiting for the perfect moment to do something you love.

Lesson learned (again).

Weekly HumanAI Challenge:

This one's simple.
Think about a mistake you made in the past year.
A real one. Not something like "I work too hard because I care too much". Please.

Create your first F*ck-Up Library entry using the prompt above.

And if you feel like sharing… HIT REPLY and tell me about your biggest f*ck-up!
I'd love to hear it.

Who knows, maybe I'll share a few of my own in the next edition.

Stay active, stay sharp.

Mencho
🪬

P.S. NASA lost a spacecraft because two teams weren't speaking the same measurement language. I lost 13 months of newsletters because I kept telling myself I'd write the next one "when I have more time".

At least mine was cheaper.

SORRY! AD SPACE… ATLASSIAN IS SPONSORING THIS NEWSLETTER; IF YOU WANT TO SUPPORT ME JUST CLICK THE BUTTON BELOW:

State of Product reveals what AI still hasn’t solved

80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.

What do you think of this edition? Tap your choice below and let me hear it 👇

🪬🪬🪬🪬🪬 AMAZING - LOVED IT!

🪬🪬🪬 GOOD / OKAY

🪬 NOT GREAT

OR… HIT REPLY and share your thoughts on what you want to have more of in this Newsletter!