Thinking in New Boxes
An AI went rogue for four days and nobody noticed. Here's what that says about the boxes you've stopped questioning too
This sounds like some bad cyberpunk novel. Unfortunately, it’s real.
Just a few days ago, an OpenAI AI agent escaped its test environment, hacked into Hugging Face for four days, and stole credentials. OpenAI didn’t realize it was its own AI until Hugging Face had already called the FBI. Four days! That’s longer than I’ve kept a New Year’s resolution.
This news triggered all sorts of alarm bells in me. It also triggered a memory of a previous conversation.
Two years ago, an AI expert came to my organisation to give a talk on AI safety — how AI might end up destroying the world faster than climate change ever could (great, one more thing for the group chat to catastrophise about).
I asked him a simple question: what’s the possibility that one clever person, working alone, invents a rogue AI agent in his basement? He was very confident in his answer. Not possible, he said. AI requires too much compute power. One person, working alone, wouldn’t be able to do anything that disruptive without being noticed.
Lo and behold, in November 2025, Peter Steinberger built OpenClaw in a one-hour personal weekend project. Annoyed that a fluid, messaging-based local AI assistant didn’t yet exist, he connected WhatsApp to a command-line interface so he could text his own computer. The project became a viral, record-breaking open-source movement — built by exactly the kind of lone person the expert told me, with great confidence, couldn’t do it.
One hour! I have spent longer than that deciding what to make for dinner.
The AI expert wasn’t stupid, and he wasn’t lying to me. He was doing what all of us do: answering from inside a box he didn’t know he was standing in. The difference is his box got publicly disproven by a guy with a WhatsApp habit.
Here’s the thing, Uncommoners — you have boxes too. Yours probably don’t involve rogue AI agents (if they do, please write in, I have questions). But they’re there.
That’s what sent me back to a book I’d read a while ago — Thinking in New Boxes, by Luc de Brabandere and Alan Iny, two strategy and innovation people from BCG, which is a fancy way of saying “people who get paid a lot to tell companies the thing they already suspected.”
Its core premise: your brain cannot actually “think outside the box,” because it’s hardwired to use mental models — boxes — to make sense of a complicated world. You can’t escape the box. What you can do is get better at noticing the one you’re in, and build a new one that fits who you’re actually becoming.
The book lays out five steps. I’ve been sitting with what they’d look like turned on your own life instead of a company’s strategy.
1. Doubt everything
Start by noticing the “seductive comfort” of your current box — the things you’ve decided about yourself that you stopped questioning years ago, not because they’re true, but because questioning them sounded exhausting and there was laundry to do. A few ways in:
Run a beliefs audit — on yourself. Write down what you assume to be true about your own life: “I’m too old to start over.” “This is just who I am now.” “This is what a woman my age is supposed to be doing.” Half of it has been running the show without ever being said out loud, like a houseguest who moved in during 2019 and never left.
Try the five-word challenge. Describe yourself in exactly five words. Then describe yourself again without reusing any of them. (If “mother, wife, manager, tired, busy” is your first five, what’s left when you’re not allowed to use them?)
Ask yourself a provocative question. “What have I never been afraid of that could actually reroute the next ten years of my life?” Sit with that one. Maybe with a cup of tea. Maybe with something stronger.
Borrow someone else’s eyes. How would the 25-year-old version of you handle your current problem? How would the woman you want to be in ten years handle it? (She’s judging you a little. Let her.)
Notice you’re deducing, not inducing. Deduction is applying your old rules to a new situation. Induction — building a new rule out of details you’re only just noticing about your actual life right now — is where the real room is. Also it’s the one they never explain properly and everyone just nods along in the meeting.
2. Probe the possible
Look outside your usual circle for fresh material. Read something you wouldn’t normally read. Talk to someone living a life that looks nothing like yours. Notice what’s shifting around you — technology, norms, what “career” or “family” even means now — that might be quietly opening doors you’ve assumed were shut, or at least locked and possibly haunted.
3. Diverge
Generate ideas for your own life, as many as you can, including the ones that sound absurd. Especially the ones that sound absurd. Start from a real question (”How could I...?”) and suspend judgment completely, for now. Quantity over quality. Your inner critic does not get a vote in this section.
4. Converge
Bring order back. Out of everything you generated, pick the one or two that actually feel worth trying. A new box you could really live in, not just an interesting pile of what-ifs sitting in your Notes app next to a grocery list from March.
5. Reevaluate relentlessly
No plan for your life is good forever. You’ll keep changing, the world will keep changing, and the box that fits you now has an expiry date already ticking, quietly, like milk. Stay close enough to your own life to notice when it’s time to open it back up.
I’ve been in strategic planning for twenty years, and I’m no stranger to this framework. And still — I watch people be slow to do step one. Slow to doubt. Slow to notice that things they assumed were settled are now openly up for grabs. I’ve been just as guilty of it myself, in quieter ways — clinging to a version of my life that used to be true, mostly because rebuilding it felt too exhausting.
The AI expert in that room wasn’t an outlier. He was just the version of “slow to doubt” that happened to have a microphone that day. Most of us are doing our own version of it, just without an audience, and without the FBI getting involved (hopefully).
You don’t need OpenAI’s mistake or a five-year strategy offsite to start. You need to pick one assumption about your own life — the one you’re most confident about, the one you’d bet your last dollar on being permanent — and question it today. Genuinely. Then plan accordingly.
What’s the box you’re most sure you’ll never need to open? Tell me. I’m collecting them. I promise to be less smug about it than that AI expert was.
Until next time,
Brinny
Brinny is a mid-lifer and a mum, writing from inside a demanding corporate career and a full life in Singapore. Subscribe for bite-sized wisdom to help you build a happier life and a career you love.