I had three job offers and two weeks to choose. That sounds like a brag, and it is, but it's also a genuinely paralyzing situation, because every single offer was good. Different company, different role, different salary, different city. I'd been canvassing friends and family for advice and collecting beautifully contradictory opinions from people who loved me but had maybe 30 percent of the picture each. I was exhausted, slightly resentful of my own good fortune, and exactly zero percent closer to a decision.
The problem with a hard decision between good options is that your brain flatly refuses to do honest math. It just keeps re-feeling the same feelings in a loop. Monday I wanted Offer A. By Wednesday I'd talked myself into Offer C. By Friday I distrusted my own preferences so completely that I started doubting whether I even knew what I wanted out of a career. Indecision had quietly become a full-time, unpaid, soul-eroding job, and I was bad at it.
Forcing the implicit to become explicit
On a Sunday afternoon I finally sat down with the Decision Matrix — Score Your Options, Find the Real Trade-Off, Hear the Devil's Advocate prompt on Prompt Dock and GPT 5.2. I listed the three offers as options and six criteria: compensation, growth trajectory, remote flexibility, team quality from the interviews, company stability, and alignment with my five-year goals. Then came the part that did the real work — assigning weights. When forced to put numbers on it, I weighted growth and alignment highest, because that's what I genuinely believed once I had to commit to it on paper instead of just feeling it vaguely. Then I let the model score every cell.
The matrix itself was almost beside the point. The genuinely revealing part was the one-sentence rationale per cell — it surfaced things I'd been quietly telling myself didn't matter but very clearly did. The verdict endorsed Offer B, which was not the highest-paying option, and explained the trade-off that win was hiding. And the devil's advocate section for Offer A was the single most useful paragraph of the whole exercise: it pointed out, fairly and without flattery, that I was discounting the financial runway the top salary would buy me, and that 'runway' is really just 'freedom' wearing a spreadsheet costume.
- Assigning weights made my real priorities visible — turns out I cared about growth roughly 3x more than pay
- The 'too close to call' flag on Offers B and C forced me to find the one fact that would actually break the tie
- The devil's advocate gave Offer A the fair hearing my anxious, exhausted brain was never going to give it
- Twenty focused minutes dissolved two full weeks of circular, sleep-stealing deliberation
Why the confidence outlasted the decision
I took Offer B. Three months in, I stand by it without a flicker of doubt — and that durable confidence is the actual product here, more than the choice itself. I didn't just pick; I know precisely what I optimized for and exactly what I traded away to get it. When the inevitable hard-day doubt shows up — and it always does, around week six of any new job — I can reread my own reasoning instead of relitigating the entire decision from scratch at 11pm. The matrix didn't just choose; it gave me a receipt.
Where I now use it beyond job offers
Once I trusted it for the big stuff, I started running it on decisions I'd previously just agonized over alone. Which of two vendors to sign. Whether to rebuild a feature or patch it. Even a genuinely silly one — which apartment to rent, with criteria like commute, light, and 'will I actually be happy here.' The pattern that keeps repeating is that the math is rarely the surprise; the surprise is always in the weights. The act of being forced to say, on a one-to-five scale, that I value commute more than square footage tells me more about myself than any amount of pro-and-con listing ever did. The matrix is really just a machine for making your own priorities confess.
I've also learned to trust the 'too close to call' flag instead of fighting it. Early on, when two options tied, I wanted the model to just pick one and end my suffering. But a true tie is real information — it means the criteria I gave it genuinely don't separate the options, and the honest move is to go find the missing data the prompt names, not to force a false confidence. Twice now that flag has sent me back to ask one more question of one more person, and both times the answer flipped the decision. A model that refuses to fake certainty is worth its weight.
Why I'd pay for this one
I reached for GPT 5.2 deliberately, and it's genuinely not interchangeable with a quick-rewrite model here. This task needs real reasoning: weighing trade-offs across six dimensions, resisting the strong pull to flatter the obvious front-runner, and then arguing convincingly against its own conclusion in the devil's advocate section. That last bit is the hard part — most models, asked to make a case for the option they just rejected, phone it in with something limp and half-hearted. GPT 5.2 didn't; it made a case for Offer A good enough that I genuinely reconsidered for an hour before reaffirming Offer B with eyes open. This is the one prompt in my entire productivity stack I'd happily pay for, and in fact I have. Grab the Decision Matrix — Score Your Options, Find the Real Trade-Off, Hear the Devil's Advocate prompt on Prompt Dock the next time you've been circling a real decision for days. Twenty minutes genuinely beats two weeks, and the confidence outlasts both.