You're picking a new car. Or a new health insurance plan. Or which of three job offers to take. You started a comparison spreadsheet last weekend, then opened seventeen tabs, then closed them all without saving anything, then ended up making the decision Sunday night based on a feeling and a half-remembered detail. Two weeks later, you have a mild but real sense that you didn't quite pick the right one — and you can't remember why you picked the one you did.
Big decisions usually fail not because of bad judgement but because of bad memory. The information arrives in pieces over weeks. You read three reviews on a Tuesday morning, talk to a friend on a Saturday, and skim a comparison article on a flight. By the time the choice is in front of you, the inputs are scattered and the version of you that did the research is gone.
A vault that holds one page per option, with a comparison alongside and your honest take written down, makes the choice feel less like guessing and more like reading a clean version of your own thinking. (For the version specific to housing, Apartment Hunting and House Buying walks the same shape with property-specific examples.)
One page per option, and one comparison page on top
In a markdown vault like Docapybara, each option in the decision gets its own page — one for each car you're considering, one for each insurance plan, one for each job offer. Pages nest with no depth limit, so a Decision: new car 2026 parent page can have a child for each candidate.
Each option page holds whatever you've gathered: the listing, the brochure PDF, the review article, your notes, the messages back and forth with the dealer or the recruiter. The agent treats the whole subtree as one searchable pile.
A separate Comparison child page holds the side-by-side view — usually as an inline database that grows a column for each criterion that matters to you. The page above the comparison can be a couple of paragraphs on what you actually want out of this decision, written before the comparison so the criteria don't get reverse-engineered to fit the option you're already leaning toward.
Capture as you go — not at the end
The real cost of a decision isn't picking; it's the weeks of accumulating information. Most of that input arrives at moments when you can't sit down to take real notes — mid-conversation, mid-test-drive, mid-flight, in the parking lot after a meeting.
Voice handles those moments. "Just test-drove the Toyota. Quieter than I expected on the highway, but the seat doesn't adjust quite enough for me. Salesperson was patient, didn't push. Walked away interested but not sold." Thirty seconds, transcribed onto the option's page with a timestamp.
For information that arrives in text — a friend's recommendation, an article someone sent — the same friction-free capture matters. Drop the link or paste the message on the page; the agent can sort it later.
The shape that works: capture without filtering. Don't try to be tidy about which page each input belongs on right away. A Quick captures page where everything lands first, and the agent files it later when you ask: "Look at the captures from this week and add anything car-related to the right car page." (More on this in The Capture Habit: Remembering the Things That Actually Matter.)
A comparison table that holds what matters, not what's easy to measure
The trap with comparison spreadsheets is that they overweight the things you can put numbers on. Cost, mileage, square footage. The things that actually matter — feel, fit, gut reaction, the quality of the people involved — get left out because they're harder to encode.
An inline :::database::: directive lets you mix both. Six column types are available: text, numbers, dates, checkboxes, single-select, multi-select. You can have a column for monthly cost and a column for gut reaction (rated on a personal scale) and a column for what worried me, all on the same row.
The criteria are yours, defined upfront. For a car: the obvious numbers, plus how I felt during the test drive, how the dealer treated me, and what's worrying me about this one. For a job offer: salary and benefits, but also what I think I'd be doing six months in, how I felt after the team interview, and what I'm afraid I'd regret.
When the comparison gets crowded, ask the agent for a focused view: "Show me only the cars where the gut reaction was 7 or higher, sorted by monthly cost." Or: "Show me only the job offers where the team I'd be on felt strong, regardless of salary."
The honest paragraph that actually does the work
The single most useful thing on the page isn't the comparison — it's a paragraph called What I actually think under each option. Two or three sentences, written as if you were telling a trusted friend what you really feel about this option, not what you'd put on a pros-and-cons list.
The Toyota: probably the right pick on paper. Cheaper, more reliable, fits my life. But I keep coming back to the Mazda — I'd be happier driving it. Worried I'd regret being pragmatic.
That paragraph, written in real time as you're considering, is what your future self will read to remember why you chose what you chose. It's also where the actual decision usually surfaces. The comparison table tells you which option scores well; the what I actually think paragraph tells you which one you'd pick if no one was watching.
When you're ready to decide, ask the agent: "Read the 'what I actually think' paragraph for each option and summarize the leanings." You get a reflection of your own thinking that's harder to dismiss than a feeling.
Research that lives in the page, not in your tabs
The research phase is where most decisions go off the rails. You start looking up safety ratings, end up reading a Reddit thread about reliability, drift to a different car you weren't even considering, and lose the morning.
Keeping the research inside the page helps. The agent's web_search tool can pull current information without sending you to an open-tab tunnel. "Find what current owners of the 2026 Mazda CX-5 say about reliability after the first year." Comes back with sources you can read or save. "Search for any safety recalls on the Toyota RAV4 in the last 18 months." Same shape.
For PDFs — brochures, financial documents, contracts, plan summaries — drop them on the option's page. They're auto-converted to markdown so the agent can read them. "Compare the deductibles and out-of-pocket maximums across these three insurance plans." The agent reads the three PDFs and gives you a clean answer.
For longer reads — a six-thousand-word car review, a forty-page job offer letter — the agent can pull the parts that matter: "Summarize the parts of the offer letter about equity vesting and severance."
A pre-mortem that surfaces the regret you're trying not to think about
The hardest decisions usually have an unspoken regret hovering over them. The pre-mortem is a way to surface it.
A Pre-mortem child page holds a single prompt for each option: Imagine it's six months from now and you regret picking this option. What went wrong? Write a sentence or two for each. The honest answer often points to something the comparison table can't capture.
For the Toyota: "I regret playing it safe and never getting excited about the car I drove every day." For the Mazda: "I regret spending more money on something that turned out to be less reliable." Both can be true; the regret you're more willing to live with is the one that points to your choice. (For founders running the same shape on a strategic decision, AI for Business Ideas: Stress-Test Before You Build is the equivalent.)
When you're stuck, ask the agent: "Read the pre-mortem for each option. Which regret seems more recoverable, and which one feels harder to live with?" The answer doesn't make the choice for you, but it usually clarifies which way you're leaning.
A decision page and a follow-up date
Once you've decided, write the decision down. A Decision page (under the parent decision) with two paragraphs: what I picked and why I picked it. Be specific about the reasons — the ones that mattered, not the ones that sounded good.
Then add a follow-up date. Three months out for a smaller decision, six months for a bigger one. The agent can remind you to revisit: "In six months, ask me how I feel about the car decision and compare it to what I wrote down today."
Three months later, when the agent surfaces the page, write a paragraph on how it actually played out. The Toyota was as reliable as expected; you do still wish you'd picked the Mazda. The job was harder than the offer letter implied; the team was as strong as you hoped. This is how decisions teach you about your own decision-making — by giving you an honest record to compare against.
Over years and a dozen decisions, the pattern becomes its own thing to read. Ask the agent: "Look at the decisions I've documented and the follow-up notes. What patterns do you see in what I tend to over-weight or under-weight?" The answer is grounded in your actual record, not in your impression of yourself.
A starter shape that works on day one
If you're starting today and you have a real decision in front of you, here's where to begin:
- Decision: [the thing] — one parent page, with the what I actually want out of this paragraph at the top.
- One child page per option, named after the option.
- Comparison — one child page with the inline database.
- Pre-mortem — one child page with one paragraph per option.
- Decision — one child page, written when you choose.
That's the whole shape. Five pages, no template required.
For the captures-as-you-go habit, a single Quick captures page that you talk into whenever new information arrives is enough. The agent files it on the right option page later.
The point isn't to systematize decision-making. It's that when the choice is finally in front of you, you have a clear record of what you learned, what you actually feel about each option, and what you'd regret most — instead of a Sunday-night guess.
Try Docapybara free. Pick one decision you're in the middle of, set up the option pages tonight, and see how much clearer the comparison gets by the weekend.