A decade ago, just after moving from Australia to the US, I sat across from my accountant as he explained — for the third time — why my particular mix of freelance income, US-based employment, and international investment dividends made my taxes “complicated.” His face had that expression professionals get when they’re about to charge you a shit-ton. (I know you know the look).
“Can we simplify this?” I asked. He shrugged. “Not really. This is just how it is when you’re between different countries.”
I accepted it. Filed it away as “the way things are.” Paid the eye-watering fees every quarter. Spent entire weekends gathering documents I didn’t understand for reasons I couldn’t grasp. And before long, I stopped thinking of it as a problem. It became like gravity, or the price of avocados — an immutable force of nature you just work around.
Then a few days back, I’m reading Ethan Mollick’s latest post (if you’re not reading “One Useful Thing,” what are you even doing with your life?), and he drops this observation about people not noticing AI breakthroughs (i.e. recent LLM models) because “they do not have a store of very hard problems they are waiting for AI to solve.”
And I just sat there, staring at my screen, thinking: “Wait, I have tons of hard problems. Like… uh… there’s that thing with the… um… and the other thing that always… uh… ah fuck.”
I had a meta-problem: I couldn’t actually name my problems. Not off the top of my head. I knew they existed. I’d spent countless hours whinging about them, working around them, losing sleep over them. But ask me to list them? To actually articulate what problems I’d love to solve if I had the right tools?
Blank space, baby. (Not in a sassy Taylor Swift kinda way.)
I have obstacles standing between me and a better life/career/world. I know I do. Yet I’d somehow never thought to, you know, write them down?
The Graveyard of “I Give Up” Problems
We all have them, don’t we? Those genuinely difficult problems we’ve abandoned. Not every day annoyances like your neighbour’s leaf blower at 7am, but the real mind-melters that could change things if only we could crack them.
Maybe it’s that career pivot you can’t figure out or the mysterious health thing that three specialists couldn’t diagnose. Or maybe it’s something bigger.
Think about climate scientists trying to model ocean ecosystems with millions of variables and feedback loops, but using simplified models they know are wrong. Or real-time, large-scale disaster response coordinators aiming to integrate data from satellites, sensors, ambulances, and shelters to optimise relief deployment. The kind of problems where smart people just shrug and say “we’re doing our best with what we’ve got.”
These aren’t “I wish my coffee maker was faster” problems. These are the problems that keep very smart people up at night, some because they affect their lives, others because they might affect everyone’s lives.
A friend runs a small print publishing company. She’s wanted to expand internationally. But managing foreign rights, translations, distribution, and payments across dozens of countries? It’s like solving a Rubik’s cube soaked in baby oil. On a unicycle. In a cyclone.
“That’s for the big publishers,” she told me over a beer last month. “Companies like mine just don’t go global.”
But then something like GPT-5 drops and suddenly she could feed it all her contracts, distribution requirements, and market research and ask: “Help me build a step-by-step plan to take my publishing company global without hiring a team”? What if she could talk to an AI that understands international publishing law, currency exchanges, and distribution logistics well enough to walk her through it?
My friend wouldn’t even think to try because she’s buried that dream in the graveyard of impossible things, right next to “learning to play guitar,” “understanding cryptocurrency,” and “scaling your business without becoming the stressed-out asshole boss you swore you’d never be.” (Do I seem like I’m projecting? Because I’m definitely projecting.)
Building Your “When AI Gets Good Enough” List
The question is, when given the tooling to help, will you even remember what you wanted to solve? Or will you end up being that person using GPT-5 to write better emails while your actual hard problems stay buried in the graveyard, forgotten?
How do you resurrect these problems from the graveyard? How do you go from “that’s just how it is, let’s forget about it” to “I’m gonna remember this and solve it as soon as I have the right tools”?
First, you need to recognise them. They’re usually hiding behind phrases like:
“It is what it is” (No, it isn’t.) “That’s above my pay grade” (Not anymore, it isn’t.) “I’ve tried everything” (You haven’t tried GPT-5, have you?) “Maybe when I retire” (Or maybe tomorrow?) “It would be nice, but…” (But what? But you don’t have an AI assistant that could process complexity like a superhuman intern?) I’ve just started keeping what I call my “Waiting for the Robots” list. (My wife thinks this name is stupid. She’s probably right. I’m keeping it anyway.)
There are so many ways to recognise and record these things. I find voice notes on my phone easiest (and recently became a Limitless Pendant user… still figuring out best ways to utilise it). Plus most meetings we attend these days are transcribed. You can drop transcriptions into ChatGPT and ask it to recognise those phrases and pull out overlooked solvable problems.
After a month of actually paying attention, you’ll build a list that’ll make you either excited or depressed. Maybe both. You’ll have personal stuff, work challenges, and if you’re feeling ambitious, some genuine save-the-world problems. Your list might end up looking something like this:
The Personal Stuff:
“Figure out my completely screwed up international tax situation” “Finally diagnose why I’m exhausted all the time in a way that doesn’t involve someone saying ‘have you tried yoga?’” “Plan the career transition that doesn’t bankrupt me or destroy my marriage”
The “This Could Actually Matter” Stuff:
“Model ocean ecosystem collapse accurately enough to know when we’re truly fucked instead of just kinda guessing.” “Detect AI-generated misinformation in real-time across multiple languages” “Create an early warning system for the next pandemic that actually works”
Your list might look different. (I truly hope it does or you’ll be real busy and in demand). But I bet whether you’re figuring out your personal finances or preventing ecological collapse, you’ve got problems that, five years back, felt like trying to nail jello to a wall but now feel more solvable.
The Three Kinds of “Holy Shit, This Is Hard” Problems
After going through my own graveyard of abandoned problems (and talking to way smarter people than me with their own graveyards), I’ve realised they fall into three categories:
The “Too Many Spinning Plates” Problems: These involve so many variables that your brain kinda just nopes out. On the personal side, it’s like planning a major life change affecting everything simultaneously. Or maybe in your work life it’s understanding how a new drug will interact with the bazillion variables in human biology. Your brain can’t hold all these pieces at once — but guess what can? An AI that doesn’t have the computational limitations of meat’n’bone hardware.
The “I Don’t Know What I Don’t Know” Problems: These are hard because the information needed is scattered across domains. Yes, my tax situation falls here. But so does designing sustainable cities that actually work — which requires understanding urban planning, environmental science, human psychology, economic systems, and about seventeen other fields. You’d need to be an expert in everything. Or have access to something that effectively is.
The “Herding Cats” Problems: These require coordinating multiple systems that don’t play nice together, like my friend’s publishing challenge. Or like working on rural healthcare access and coordinating between federal regulations, state systems, insurance companies, local hospitals, and transportation networks.
These are EXACTLY the kinds of problems LLMs are getting scary good at helping with. Complexity problems? Give me an AI that can hold a thousand variables in context. Information problems? These models have ingested more information than you could read in a million lifetimes. Coordination problems? What if you had an AI assistant that could draft communications, project manage, and keep track of all the moving pieces while you focus on the human elements?
Before you think I’ve drunk all the AI Kool-Aid let’s pump the brakes for a second. I still consider these LLM models tools, not oracles. They can still hallucinate like they’re at Burning Man. So don’t perform surgery based on a raw ChatGPT output without, you know, checking. Anything medical, legal, or financial, you still want actual humans in the loop. AI can help you understand and navigate these things, sure, but some solutions require humans doing human things, for better or worse.
The Magic of Admitting You Have Problems
Something weird happens when you start cataloguing your hard problems: you stop feeling like such a failure for not solving them. (Projecting again, I know.)
When these problems are just floating around in your subconscious, occasionally surfacing at 3am to ruin your sleep, they feel like personal failures. Why can’t I figure this out? Why is everyone else managing this better? What’s wrong with me?
When you write them down and look at them objectively, you realise, holy shit, these are genuinely hard problems. Of course you couldn’t solve them with a spreadsheet and a positive attitude. Maybe they just require computational power, knowledge synthesis, and strategic thinking at a level that human brains just aren’t built for.
It’s like beating yourself up for not being able to fly. Sure, humans can’t fly — but we built aeroplanes, didn’t we?
Your Homework (Yeah, I’m Giving You Homework, Deal With It)
Start your list. Today. Right now. I don’t care if it’s on your phone, a napkin, or carved into your desk with a butter knife.
But this is crucial: don’t list the little stuff. Don’t put “inbox zero” or “organise closet” or whatever basic adulting you’re avoiding. List the real ones. The problems that would change everything if solved — whether “everything” means your life, your field, or literally everyone’s future.
Then pick something off your list and do this:
- Go to ChatGPT. In case you missed it, they released their GPT5 model last week (even for free users) which is arguably the most powerful model most of the general population (i.e. not a geek like me) will have experienced so far.
- Run a 30-minute experiment. Right now. Don’t wait. Cut and paste the prompt below and replace with the relevant details for the problem you’ve picked. Don’t overthink it, just fill in the [blanks] best you can and see what happens:
You are a [domain] analyst. Context: [brief description of the situation or challenge]. Goal: [desired outcome]. Constraints: [time/money/compliance/etc.]. Knowns: [Relevant details]. Unknowns: [Relevant details]. Step 1 - Think deeply and reason step-by-step about the context, goal, constraints, knowns, and unknowns.| Step 2 - Produce:
- A first-pass plan.
- Key risks and assumptions.
- The 3 fastest, lowest-cost ways to invalidate or stress-test this plan. Step 3 - Build relevant models or visualisations with code and display them in canvas. Step 4 - Output any additional assets you deem necessary. Rule: Any legal/medical/financial advice must be flagged: "Validate with human expert".
- Just observe and review what happens. In most cases, I think you’ll be blown away. Fair warning: GPT5 is great at understanding the context and suggesting many different actions… you may well go down a ChatGPT rabbit hole for hours! But you may also unlock something amazing.
And then? Keep paying attention. Experiment. Learn. When the next more-advanced LLM model drops, and everyone’s sharing their cute chatbot conversations and making it write limericks, you’ll be the one thinking bigger. You’ll be the one who sees a new capability and immediately knows fourteen ways to use it.
The future isn’t about AI solving our problems, but about AI making us capable of solving problems we never thought we could tackle. The tools are here. Your abandoned problems are waiting. What happens next is on you.
(And seriously, if you’re working on something that could prevent the next pandemic, fix mental health challenges, or solve an energy crisis, definitely share. But also, if you’ve figured out the whole “maintaining friendships as an adult” thing, please DM me immediately. C’mon, both kinds of problems matter.)