Life in D/P: Your Entire Life Is a Mix of Deterministic and Probabilistic Steps

Last week I published The D/P Framework — a way to map any AI product as a sequence of deterministic (D) and probabilistic (P) steps. The idea was simple: every workflow is a mix of “same input, same output” steps and “same input, who knows what output” steps. And the pattern you choose determines your entire product strategy.

That was the professional version.

Here’s what actually happened after I published it: I couldn’t stop seeing D/P patterns everywhere. Not in products. In life.

Making coffee. Driving to work. Texting a friend. Raising a kid. Job hunting. All of it — every single daily activity — maps to a sequence of D’s and P’s. And once you see it, you can’t unsee it.

This is either a useful insight or a sign I need to go outside more. Probably both.

Quick Recap: D and P

If you read Part 2, skip this. If you didn’t, here’s the two-sentence version:

Deterministic (D): Same input, same output. Every time. Flipping a light switch. Adding 2 + 2. No surprises.

Probabilistic (P): Same input, different output. An LLM call. A first date. Asking your toddler to put on shoes. You can influence the odds, but you can’t guarantee the result.

Now let’s apply this to everything.

Your Morning Routine Is D→D→D

Think about what you did this morning.

Alarm goes off at 6:30 → you turn it off → you get out of bed → you walk to the bathroom → you brush your teeth → you make coffee.

Every single one of those steps is deterministic. Same input, same output. The alarm makes a sound, you press the button. Toothpaste goes on brush, brush goes in mouth. Water goes in kettle, heat is applied, coffee is made. There’s no uncertainty here. No probabilistic step. No confidence threshold.

Your morning routine is a D→D→D→D→D pipeline. It’s the most reliable system you run every day, and you engineered it that way on purpose. You don’t think about it because that’s the whole point — you turned every step into something automatic and predictable.

Now here’s what’s interesting: it wasn’t always like this. When you first moved into your apartment, your morning was full of P steps. Where are the mugs? Which burner works? Is the water pressure going to scald me? Over time, you converted every P into a D by learning, repeating, and building habits.

You did exactly what a good PM does: you reduced probabilistic steps to deterministic ones wherever possible.

Cooking Is Where D Meets P

Following a recipe is D. Measure 2 cups of flour. Set the oven to 375°F. Bake for 25 minutes. Same input, same output. A recipe is literally a deterministic algorithm written for humans.

But seasoning to taste is P. “Add salt until it tastes right.” Same starting point, different result every time, based on your palate, the humidity, the brand of tomatoes, whether you’re in a good mood. No two batches of soup are identical because the final step is probabilistic.

A home cook following a recipe exactly is running D→D→D→D. A chef who adjusts seasoning by instinct, swaps ingredients based on what looks fresh, and plates based on visual feel is running D→D→P→P.

The leap from “cook” to “chef” is the leap from all-D to a workflow where you’re comfortable with P steps. You’ve built enough experience to trust your probabilistic judgment. Same pattern in products — junior PMs want deterministic rules for every decision. Senior PMs are comfortable making judgment calls with incomplete information.

And baking? Baking is pure D. You deviate from the recipe, you get a brick. There’s a reason people say “cooking is an art, baking is a science.” They’re describing D/P without knowing it.

Driving Is D Until It Isn’t

Your daily commute on a clear day with no traffic? D→D→D→D. You know every turn, every lane change, every light. You could probably drive it blindfolded. (Don’t.)

Now add rain. That’s a P modifier on every step. The road is the same, but traction is probabilistic. Braking distance is probabilistic. Whether the car next to you is going to hydroplane is probabilistic. Same route, entirely different risk profile.

Add a construction detour you’ve never seen before. Now your navigation step went from D to P. You’re relying on instinct, road signs, or Google Maps — and Google Maps is running its own D→P→D pipeline to give you that new route.

Traffic itself is P→P→P. Everyone is making independent probabilistic decisions — lane changes, speed adjustments, merge behavior — and the system’s behavior is emergent. No one designed it. There’s no product manager for highway traffic. And it shows.

This is why autonomous driving is so hard. The car needs to operate in a D→D→D mode in an environment that’s fundamentally P→P→P. Every other driver is a probabilistic agent. Pedestrians are probabilistic agents. Weather is probabilistic. The autonomous vehicle is trying to run a deterministic algorithm through a probabilistic world. Sound familiar? That’s the same challenge every AI product faces.

Job Searching Is P→D→P→D

Here’s one that hits close to home for anyone who’s been on the market.

Applying: D. You fill out the form. You upload the resume. Same input, same output. The application either gets submitted or it doesn’t.

Resume screening (by a human or an ATS): P. Same resume gets different results at different companies, different days, different screeners. You have no control over who reads it, what mood they’re in, or whether the algorithm was updated last Tuesday.

The phone screen: P. You say roughly the same things about your experience, but the interviewer’s reaction is probabilistic. Chemistry is probabilistic. Whether they had a bad morning is probabilistic.

The offer decision: P → D. The decision to extend an offer is probabilistic (committee deliberation, comparison to other candidates, budget cycles). But the offer letter itself is deterministic — salary, start date, equity. Those numbers are fixed.

The entire job search is a long P→D→P→D chain where you control the D steps (polish resume, practice answers, send thank-you emails) and try to influence the P steps (research the company, build rapport, tell compelling stories) without being able to guarantee any of them.

And here’s the PM insight: every piece of job search advice is an attempt to convert a P step into something more D-like. “Use the STAR format” = make your interview answers more deterministic. “Tailor your resume to the job description” = increase the probability of passing the screening step. “Follow up within 24 hours” = a deterministic action designed to influence a probabilistic decision.

You’re engineering guardrails for a probabilistic pipeline. You’re doing product management on your own career.

Parenting Is P→P→P With No Rollback

I have a kid, and I’ve also watched enough parents to map this one confidently.

Giving instructions to a toddler: P. “Please put on your shoes.” The output is genuinely unknowable. They might put on shoes. They might put shoes on their hands. They might ignore you entirely and start singing. Same input, wildly different outputs. No confidence threshold will save you.

Disciplining: P → P. You choose a response (itself a probabilistic decision based on how tired you are and what worked last time) and the child responds probabilistically. You’re chaining two P steps together with zero guardrails and no rollback mechanism.

The long game — raising a human who becomes a functioning adult: P→P→P→P→P→P over 18+ years. The longest probabilistic chain anyone ever runs. No unit tests. No staging environment. No A/B testing. You’re shipping to production every day with a sample size of one.

This is why parenting advice is so contradictory. Everyone’s running the same general pattern (P→P→P) but with completely different inputs (different kids, different contexts, different resources). What worked for one family is a single data point from a probabilistic system. It’s not generalizable. It’s an anecdote, not a benchmark.

PMs who are also parents already understand this intuitively. They know what it’s like to make decisions with incomplete data, no rollback, and a stakeholder who screams if you get it wrong.

The Meta Insight: We Naturally Optimize for D

Here’s the thing that makes this more than a fun analogy.

Humans instinctively try to convert P steps into D steps. That’s what learning is. That’s what habits are. That’s what routines, checklists, recipes, standard operating procedures, and “best practices” are — they’re all attempts to take something that was once probabilistic and make it deterministic.

The first time you drove a car, every step was P. Steering, braking, checking mirrors — all probabilistic, all uncertain, all requiring active judgment. After thousands of hours, most of those steps became D. You don’t think about turning the wheel. You just turn it. You converted P→D through repetition.

Education is a P→D conversion machine. You go in not knowing how to solve differential equations (P — you’d have to guess). You come out knowing the method (D — same problem, same steps, same answer). School literally converts probabilistic ignorance into deterministic competence.

Therapy, if you think about it, is partly the same thing. You walk in with probabilistic emotional responses — you don’t know why you react the way you do, and the same trigger produces different reactions depending on the day. Over time, therapy helps you understand your patterns, which converts some of those P reactions into more D-like responses. You still feel the feelings, but you develop deterministic strategies for handling them.

And here’s the flip side: some of life’s best moments are the P steps we didn’t convert. Falling in love is P. A spontaneous road trip is P. A joke that lands perfectly in the moment is P. If you converted everything to D, you’d have a perfectly optimized, completely predictable, utterly boring life.

The art of living, like the art of product management, is knowing which steps should be D and which should stay P.

Why This Matters If You’re a PM

Okay, let’s bring this back to work.

If you can see D/P in your morning routine, your commute, your cooking, your relationships, and your career — you can see it in any product. That’s the point.

The D/P framework isn’t just an architecture tool for mapping AI workflows. It’s a thinking tool for understanding any system that combines predictable and unpredictable elements. Which is every system that involves humans, data, or decisions. So… every system.

When a stakeholder says “why can’t the AI just get it right every time?” — you now have language for why. The step is P, not D. Same input, different output. That’s not a bug, it’s the nature of the step. The product decision is how to handle the variance, not how to eliminate it.

When you’re designing a workflow and something feels fragile — look for P→P chains. That’s where errors compound. Insert a D step (a human checkpoint, a validation rule, a structured handoff) to break the chain.

When you’re onboarding a new user and the experience feels unpredictable — you probably have too many P steps early in the funnel. Move the D steps forward. Give users deterministic footing before introducing probabilistic magic.

The framework works because it’s not a metaphor. It’s how systems actually work. AI products, life routines, career decisions — they’re all sequences of deterministic and probabilistic steps. The only question is whether you’re designing those sequences intentionally or stumbling through them by accident.

What’s Next

This is Part 3 of the AI-Native PM series. Part 1: What Is an AI-Native PM? covered the mindset shift. Part 2: The D/P Framework introduced the architecture tool.

This post took the framework out of the product world and into the real one — because the best frameworks aren’t the ones that only work in a slide deck. They’re the ones that change how you see everything.

Next up: applying D/P to your career and building leverage. Stay tuned.

Building in public at fullstackpm.tech. Follow along on X @fullstackpmtech.