nodes

Every node, every connection, and the claim on each one. The same data the field draws, addressed as a document.

Press to start reading at the node on screen. Arrow step through that node's connections, follows one, goes back. Enter sets a path origin, Esc clears.

principle25

A small fixed palette beats infinite options

id
constrained-palette
kind
principle
domains
guitar · design · product
when
2016

Fewer choices, made deliberately, outperform unlimited choices made carelessly.

16 connections

  • superseded Nice to haves are must haves

    I used to believe "nice to haves" were table stakes - that a feature not shipped was a product not finished. I've dropped the belief, but not before it padded a lot of MVPs past the point of usefulness.

  • has the special case Optionality is a tax you pay forever

    The cost of a choice isn't making it once. It's carrying it, supporting it, and explaining it forever after.

  • has the special case The default is the product

    If almost nobody changes a setting, then the defaults you shipped are the only ones that exist.

  • is in tension with Change your mind when the information changes

    Staying open to better information argues for keeping options alive. Committing to a small palette argues for closing them. Both are right and they pull against each other constantly.

  • reinforces Taste is mostly subtraction

    Two versions of the same instinct: one about what you offer, one about what you keep.

  • is broken by An unreasonable number of electric guitars

    The clearest counterexample I own. I am certain the principle is right and I am equally certain I need another guitar.

  • was taught by Cherry sunburst Les Paul Standard

    Two pickups and a switch. You can't configure your way out of a bad take, so you learn the instrument instead.

  • is illustrated by Linoleum block relief printing

    You choose the plate, the ink and the edition before the first pull. The constraint is decided up front and then you work within those constraints.

  • is illustrated by Played one guitar for a year

    Everything else went in a case for a year. Somewhere around month four I stopped thinking about the guitar and started thinking about the music.

  • is broken by The parts Telecaster

    I built a guitar with infinite options and then spent a year adjusting it instead of playing it. I know better and I did it anyway.

  • was taught by BA, Art Studio — UC Santa Barbara

    Every assignment came with a restriction: one color, one process, one size. I spent four years learning that the limit is what makes the decision interesting.

  • is illustrated by Bought the Les Paul when I graduated

    One guitar, chosen deliberately, to mark the end of five years. No shortlist, no comparisons needed, and definitely no Epiphones.

  • is illustrated by pauldenni.com

    One accent color, one typeface, dots and lines. Every element I removed made the remaining ones easier to read.

  • is illustrated by 90s hip hop

    An MPC, a crate of records and a small number of bars. The limits were the instrument.

  • is illustrated by Back to the Future

    One town, one street, two eras. It gets enormous range out of deliberately refusing to go anywhere else.

  • is illustrated by The Guild acoustic

    One sound, no controls, no pedals. Whatever you want out of it has to come from your hands.

Optionality is a tax you pay forever

id
optionality-is-a-tax
kind
principle
domains
product · engineering
when
2019

Every configurable thing is a decision you have to keep re-making, and a surface someone has to keep supporting.

5 connections

  • is a special case of A small fixed palette beats infinite options

    The cost of a choice isn't making it once. It's carrying it, supporting it, and explaining it forever after.

  • is illustrated by An unreasonable number of electric guitars

    Every one needs strings, setup, and a decision about whether today is its day. The tax is not the money, it is the deciding.

  • is broken by The CMS's first custom page builder

    We made it almost too flexible. It gets used for many more cases than we designed for, people invent new ones constantly, and there is very little it will refuse to do. That is the blessing and it is exactly the curse.

  • is illustrated by The legacy CMS

    Five ways to build a carousel of movie and show content, where one would have done. The differences between them were small enough that every change started with working out which one you were in.

  • is illustrated by The parts Telecaster

    Every swappable part is a decision I get to re-litigate forever. The bill comes due every time I pick it up.

Nice to haves are must haves

id
nice-to-haves-are-must-haves
kind
principle
domains
product · business
when
2013

An early belief that "nice to haves" were necessary features that made the product better. I no longer think this way, but it left a trail of MVPs more bloated than they needed to be.

5 connections

  • was superseded by A small fixed palette beats infinite options

    I used to believe "nice to haves" were table stakes - that a feature not shipped was a product not finished. I've dropped the belief, but not before it padded a lot of MVPs past the point of usefulness.

  • is illustrated by The legacy CMS

    Every feature in it arrived as a nice to have that somebody made a case for, and every case was reasonable on its own. Years of reasonable cases produced a system nobody could safely change.

  • is illustrated by An unreasonable number of electric guitars

    I retired this belief everywhere in my working life but it is alive and well on the wall to my right. More guitars have never once made me a better player, yet I keep buying them.

  • is illustrated by vw.com

    Every model, trim, package and finance option, all reachable. It was thorough, and I now think it was mostly overwhelming.

  • is illustrated by Hubs and Collections

    Built for the expansion we could see coming. Each one was a complex feature on its own, and together they were the building blocks that made the growth possible — which is the one period where adding surface genuinely was the strategy.

Taste is mostly subtraction

id
taste-is-subtraction
kind
principle
domains
design · writing
when
2007

The work is deciding what doesn't belong. Anyone can add.

10 connections

  • reinforces A small fixed palette beats infinite options

    Two versions of the same instinct: one about what you offer, one about what you keep.

  • reinforces Recombination is invention

    The subtraction happens in the crate, not in the song. You pass ten thousand records to land on four bars, and the passing is where the taste lives.

  • reinforces Thinking time is the cheapest time in the project

    Deciding what does not belong is thinking work. It looks like nothing is happening, and it is the part that pays.

  • was taught by BA, Art Studio — UC Santa Barbara

    Four years of critique, which is a structured argument about what to remove. It is the most directly useful training I have for the job I actually do.

  • is illustrated by Linoleum block relief printing

    Relief printing is literally subtraction. You cut away everything that is not the image, and what is left of the block is the picture. You cannot put a line back.

  • is broken by Entertainment sites and social games

    The brief was always more: more features, more interactions, more things on the screen. Nobody ever asked what could come out.

  • is illustrated by Played one guitar for a year

    The year was entirely about what I put away. Nothing was added and it was the most I ever improved.

  • is illustrated by 90s hip hop

    A sample is a decision about which four bars of someone else's record matter. Everything else on that record gets left behind.

  • is illustrated by Back to the Future

    There is no spare scene in it. Whatever else was written got cut.

  • is illustrated by Side navigation across the OTT apps

    It was mostly a removal. Fewer competing entry points, and the number that moved was minutes watched.

The canonical choice is usually right

id
canonical-over-clever
kind
principle
domains
guitar · product · engineering
when
2018

The thing everyone already uses is popular because it works. Novelty is a cost you pay for.

9 connections

  • is in tension with Recombination is invention

    One says take the thing everyone already uses. The other says the work is in chopping finished things into something nobody has heard yet. I believe both and they do not sit still together.

  • reinforces Ship the boring thing to kill the expensive thing

    The unglamorous option is usually the one everyone already knows how to execute, which is most of why it works.

  • reinforces AI is a tool, not a solution

    A tool is judged on whether it does the job, not on how novel it is. That was true of frameworks before it was true of models.

  • is illustrated by Cherry sunburst Les Paul Standard

    The obvious choice, and the right one on the merits. It is the most copied electric design there is and that configuration is still the epitome of what a guitar is to me. Being unremarkable to name has nothing to do with being unremarkable to hear.

  • is illustrated by 90s hip hop

    An MPC and a crate of records everyone else also had access to. The source material being unremarkable is the point — what you do with it is the whole art.

  • is illustrated by Adobe Experience Manager migration

    Forty properties onto one platform everyone already knew. The value was standardization, not sophistication.

  • is illustrated by Side navigation across the OTT apps

    Side navigation was not a new idea. It was the pattern the rest of the category had already settled on, and adopting it late still paid.

  • is broken by The parts Telecaster

    Assembled from chosen parts rather than bought whole. The catalog version is the one I reach for, which is the argument against building your own made in the least satisfying way possible.

  • is illustrated by The Stratocaster

    The other guitar everyone has. Boring to name and still the right answer often enough that it has outlasted every design meant to replace it.

Ship the boring thing to kill the expensive thing

id
boring-thing-kills-expensive-thing
kind
principle
domains
product · business
when
2022

An unglamorous migration that retires a costly system beats an elegant one that doesn't.

5 connections

The default is the product

id
defaults-are-the-product
kind
principle
domains
product · design
when
2021

Almost nobody changes a setting. Whatever ships turned on is what your product actually is.

7 connections

  • is a special case of A small fixed palette beats infinite options

    If almost nobody changes a setting, then the defaults you shipped are the only ones that exist.

  • is illustrated by ML personalization inside the CMS

    Personalization that requires configuration is personalization nobody gets. It had to happen without being asked for.

  • is illustrated by Side navigation across the OTT apps

    Nobody asked for different navigation. Changing what was there by default moved minutes watched per user by 10.6%.

  • is illustrated by The CMS's first custom page builder

    Whatever template loaded first became the house style. We shipped a layout choice and got an editorial standard we never wrote down.

  • is illustrated by The enterprise content management platform

    Whatever the CMS makes easy is what gets published across two streaming services. The default template is the editorial policy, whether or not anyone wrote one.

  • is illustrated by Hubs and Collections

    A hub is a merchandising default, and we had lively debates about what one even was. Whatever it surfaces first is what most subscribers will ever see of the catalog.

  • is illustrated by vw.com

    A configurator where the trim that loads first is the trim most people price out.

You can't fix a bad signal at the end of the chain

id
no-fixing-downstream
kind
principle
domains
guitar · engineering
when
2014

Quality is set at the source. Everything after is damage control.

8 connections

  • reinforces Practice makes permanent, not perfect

    Both are the same observation about time: what happens early decides what is still true later, and neither effort nor tooling reverses it.

  • has the special case QA, measurement and security are not phases

    QA at the end is the definition of trying to fix the signal downstream. So is measurement bolted on after launch, and so is security.

  • is illustrated by Content operations tooling

    Bad metadata does not get better as it travels. It becomes a wrong title on a hundred surfaces, and the only real fix is upstream of all of them.

  • is illustrated by Learning solos by ear

    Tabs let you play the notes without ever hearing them. You cannot add the hearing back afterward.

  • is illustrated by Linoleum block relief printing

    A bad plate prints badly a hundred times. Nothing you do at the press recovers what wasn't cut properly.

  • is illustrated by Fifty incremental releases, not one cutover

    Fifty releases means fifty chances to catch it early. One cutover means finding out everything at once, on a night when publishing cannot stop.

  • is illustrated by Live sports integration

    There is no fixing a live event after the fact. Whatever the pipeline does wrong at kickoff is what the audience saw.

  • is illustrated by Mid-quarter checkpoints

    Drift found at the quarter boundary is a quarter you cannot get back. Six weeks in, it is a reallocation.

QA, measurement and security are not phases

id
quality-is-not-a-phase
kind
principle
domains
engineering · product
when
2020

They are properties of how a thing is built, not steps you append once it works. Bolted on at the end, all three are theater.

5 connections

  • is a special case of You can't fix a bad signal at the end of the chain

    QA at the end is the definition of trying to fix the signal downstream. So is measurement bolted on after launch, and so is security.

  • is in tension with Move as fast as you can without making rework

    Building quality in is the first thing cut when a date is at risk, and cutting it is how you buy a week now and pay a month later. I am not a fan of "move fast and break things".

  • is illustrated by Fifty incremental releases, not one cutover

    Shipping continuously into a live 24/7 operation removes the option of a testing phase at the end. There is no end to put it at.

  • is illustrated by Live sports integration

    Live sports is a constant moving target with an enormous amount of operational support behind it — many moving parts across many interconnected systems, all of which have to be right before anyone kicks off.

  • is illustrated by Content operations tooling

    QA throughput improved by half not because we tested more, but because the tooling stopped producing the errors upstream.

Practice makes permanent, not perfect

id
practice-makes-permanent
kind
principle
domains
guitar · engineering
when
2002

Whatever you rehearse is what sticks, including the wrong version. Unlearning costs more than learning did.

11 connections

  • has the special case Learn on the version that won't let you cheat

    If whatever you rehearse is what sticks, then the tool that lets you rehearse a shortcut is the dangerous one.

  • reinforces You can't fix a bad signal at the end of the chain

    Both are the same observation about time: what happens early decides what is still true later, and neither effort nor tooling reverses it.

  • reinforces Thinking time is the cheapest time in the project

    If the wrong version is the one that sticks, then the hour spent deciding which version to build is the cheapest hour in the project.

  • was taught by Learning solos by ear

    Learn a phrase slightly wrong by ear and you will play it slightly wrong for a decade. I still play things wrong that I learned at fifteen.

  • was taught by Piano

    Thirty years of playing and I am not as good as that should make me, because for a lot of those years I practiced without paying attention.

  • is illustrated by Cherry sunburst Les Paul Standard

    One instrument, one scale length, one set of habits. Everything I built that year still works because it was built on something that did not move.

  • is broken by Gaming launch sites and mobile apps

    Every one of these was built fast and thrown away, so nothing got done properly twice. I learned habits on those projects that took years to unlearn.

  • is illustrated by Linoleum block relief printing

    You cut the plate the way you have always cut plates. The technique you built badly at twenty is still in your hands at forty.

  • is illustrated by Played one guitar for a year

    One instrument long enough to build real habits on, rather than four instruments' worth of half-habits.

  • is illustrated by Saxophone

    Concert band rehearses the same passage until nobody has to think about it, which is exactly what the phrase warns about, aimed on purpose.

  • is illustrated by The Guild acoustic

    The habits I built on it are the ones I still have. Thirty years later my hands still do what that guitar taught them.

Thinking time is the cheapest time in the project

id
thinking-time-is-the-cheapest-time
kind
principle
domains
product · writing
when
2019

Working a problem through properly before anyone builds pays for itself many times over.

16 connections

  • is in tension with Move as fast as you can without making rework

    These pull against each other and I know which way I lean. Getting it right takes time that shipping fast does not have — and taking that time is still the cheapest time in the project, because doing it twice costs more than doing it slowly once.

  • reinforces Practice makes permanent, not perfect

    If the wrong version is the one that sticks, then the hour spent deciding which version to build is the cheapest hour in the project.

  • reinforces The work you disappear into is the work that teaches you

    Both are a defense of time that produces nothing you can show anyone that afternoon, and returns something you could not have got any faster.

  • is in tension with You can't name the thing you're wrong about

    Thinking a problem all the way through is the best defense I have, and it is not a guarantee. Some errors are only visible from a position you have not reached yet.

  • reinforces Imagination needs a workout

    Both are arguments for time that produces nothing you can show anyone that afternoon.

  • reinforces Taste is mostly subtraction

    Deciding what does not belong is thinking work. It looks like nothing is happening, and it is the part that pays.

  • is illustrated by The calendar is a contract

    Standing meetings pushed into the morning across the whole organization, so the afternoon is a block big enough to think in. Mine is three to five.

  • is illustrated by The ground-up CMS rebuild

    Most of the first year was sequencing rather than shipping. Every month spent deciding the order bought back several months of not building things twice.

  • is illustrated by Back to the Future

    Nothing in it is decoration. Every element planted in the first act is carrying weight in the third, which only happens if the whole thing was worked out before a frame was shot.

  • is illustrated by Backend and platform teams sequence first

    A quarter spent sequencing properly is a quarter nobody spends waiting on a capability that was promised in parallel.

  • is broken by Claude Code executive briefing system

    It hands me a synthesized week in seconds, and the hour I used to spend assembling that picture myself was an hour of thinking. I am not yet sure what I traded away.

  • is illustrated by Drum major, Rose Parade

    Every decision worth making was made in rehearsal. On Colorado Boulevard there is no deciding, only executing what was decided.

  • is illustrated by Linoleum block relief printing

    All the deciding happens before the first cut, because the first cut is permanent. The press is the easy part.

  • is broken by LLM-assisted tooling

    The risk I built into my own week: I used to carry the org in my head between updates, and carrying it there was where the connections got made. Now the system carries it and I read the output, which is not the same as having done the noticing.

  • is illustrated by Learning solos by ear

    It is much slower than reading it off a page, and it is the only version that stays learned.

  • is illustrated by Black holes and spacetime

    Most of the work is sitting with a question you cannot yet run an experiment on. The instruments catch up decades after the thinking does.

Move as fast as you can without making rework

id
move-fast-without-rework
kind
principle
domains
product · engineering
when
2017

Prioritizing speed on its own usually produces a bad result, because rework is expensive. Break stage if you must; don't break prod, and don't make yourself do the job twice.

10 connections

  • is in tension with Thinking time is the cheapest time in the project

    These pull against each other and I know which way I lean. Getting it right takes time that shipping fast does not have — and taking that time is still the cheapest time in the project, because doing it twice costs more than doing it slowly once.

  • reinforces Credibility is delivered, not requested

    Increments are how you avoid rework and how you prove you are worth waiting for. The same practice does both jobs.

  • is in tension with QA, measurement and security are not phases

    Building quality in is the first thing cut when a date is at risk, and cutting it is how you buy a week now and pay a month later. I am not a fan of "move fast and break things".

  • is broken by Froze new development on the legacy CMS

    Two years of deferring near-term requests to avoid throwaway work. It is the no-rework half of the principle honored completely and the go-fast half abandoned, and every deferred request was somebody's real problem in the meantime.

  • was taught by Gaming launch sites and mobile apps

    A game ships on its date and the site ships with it or it is pointless. My first lesson that some deadlines are facts about the world rather than negotiating positions.

  • is illustrated by Live sports integration

    A game ends when it ends. Everything about the publishing path had to work against a clock nobody in the building controlled.

  • is broken by The ground-up CMS rebuild

    A multi-year rebuild is the slowest possible answer to a problem people had that week. It was right, and it cost real credibility while it ran.

  • is broken by Drum major, Rose Parade

    Relaying a signal through a runner is the slowest possible way to move information, and it was the only way that worked at that scale.

  • is broken by Entertainment sites and social games

    Everything shipped against a date that existed before the brief did. Nothing was ever finished, it was due — which is exactly the speed that produces work you throw away and do again.

  • is illustrated by Fifty incremental releases, not one cutover

    Fifty releases means value lands continuously instead of in one distant event. It is the slow strategy that feels fast to everyone waiting.

People over process, every time

id
people-over-process
kind
principle
domains
product · business
when
2015

No process gets you as far as knowing the specific people you're working with. Teams are made of individuals, not roles.

18 connections

  • reinforces Sit next to the person doing the work

    You can't do any of this from a distance. Get your hands dirty with the people doing the work, or you're guessing from a status report.

  • reinforces Say the thing plainly

    Process exists to carry expectations between people. Say them out loud and you need less of it.

  • is illustrated by Developed four PMs into independent owners

    Prioritization stopped escalating not because we wrote a better framework, but because four people were trusted to decide and knew what they were being measured on.

  • is illustrated by Divided the platform into owned product areas

    Ownership by platform area, not by project, is what made prioritization stop escalating. The structure worked because it matched how the people already thought about the system.

  • was taught by Drum major, Rose Parade

    The drill chart was written months before. What actually held it together on the day was three of us knowing exactly which commands to give and how to coordinate them, because we had practiced it hundreds of times.

  • was taught by Gaming launch sites and mobile apps

    Eight years across four agencies came down to one skill: taking a client's idea and getting it to delivery, through the thousand small steps in between. No process ever walked those steps for me.

  • was taught by Playing with other people

    You can play a piece note for note, perfectly, following the process exactly — and now a model can too. What is missing is several people improvising off each other and finding a groove. No notation holds it and no perfect rhythm reproduces it.

  • is illustrated by “You have more power (to influence) than you think you do”

    No framework grants you standing. It comes from the people in the room deciding you are worth listening to.

  • is illustrated by A channel for each kind of message

    It is not a policy anyone has to follow. It is me telling people how to reach me so they do not have to work it out.

  • is broken by A weekly cross-team product review board

    I fixed a people problem by adding a recurring meeting. It worked, and it is still process where a few direct conversations might have done.

  • was taught by Adobe Experience Manager migration

    Forty properties meant forty teams with their own habits. The platform was the easy part; getting each of them to give up a workflow they liked was the job.

  • is illustrated by Built a two-tier product organization

    Two leads managing their own PMs rather than five people reporting to me. The org chart is a statement about who gets to grow, and a flat one says nobody.

  • is illustrated by Built the team out with high performers

    I could have arranged the team a dozen ways on paper. Who was actually on it mattered more than any of those arrangements.

  • is illustrated by Grew the product function from a single practitioner

    One person covering four workstreams is not a prioritization problem. No framework fixes not having enough people.

  • is broken by SAFe and Scrum certifications

    Four process certifications, and they were worth getting. No framework tells you what the person doing the work already knows. Most teams follow the process instead of asking them, and call it alignment.

  • is illustrated by The CMS's first custom page builder

    The ticket queue was a process standing in for a conversation. Putting the tool in editorial's hands deleted the queue rather than optimizing it, and gave engineering back the work they actually wanted to do.

  • is illustrated by The uncle who taught me guitar

    No method, no curriculum, no book. One person who played, sitting next to me, correcting the specific thing I was doing wrong that week.

  • is illustrated by The calendar is a contract

    Half of it exists because we have people in eastern, central and Pacific time. The rule follows the humans, not the other way round.

Sit next to the person doing the work

id
proximity-to-the-work
kind
principle
domains
product · design
when
2017

An hour watching someone struggle with the product teaches you more than three months of requirements gathering.

17 connections

  • reinforces People over process, every time

    You can't do any of this from a distance. Get your hands dirty with the people doing the work, or you're guessing from a status report.

  • reinforces Product and engineering are two sides of one coin

    The reason to sit next to engineering is the same as the reason to sit next to the operator: the decision improves when you can see what it costs.

  • reinforces The work you disappear into is the work that teaches you

    Disappearing into the work is the most complete form of sitting next to it.

  • is broken by Built a two-tier product organization

    Two tiers means most of the work now reaches me through someone else's account of it. That is the part of going from IC to manager I have had to keep working at — further from the making, but a higher vantage on what is coming for us. Still the right call for the team.

  • was taught by The CMS's first custom page builder

    Product brought me the problem, we brainstormed it, and the front-end engineers wanted it more than anyone — it saved them building every page from scratch and let them start templatizing. Nobody had to be convinced it was needed.

  • is illustrated by The enterprise content management platform

    Six hundred internal users is a public product with the marketing budget removed. The people using it badly are down the hall, which is the entire advantage.

  • is illustrated by The uncle who taught me guitar

    The person who taught me was in the room, playing the thing, and answering the question I actually had rather than the one a book anticipated.

  • was taught by VW Dealer iOS application

    Built for the people selling the cars, not the people buying them. The first product I worked on where the user was somebody doing a job, and I have mostly built for that user ever since.

  • is illustrated by Content operations tooling

    Ninety percent of the curation time was manual work nobody outside the team knew was happening. You only find that by sitting with them while they do it.

  • is illustrated by Divided the platform into owned product areas

    An owner per area means somebody is close enough to each part of the platform to notice when it is going wrong.

  • was taught by Entertainment sites and social games

    Project management, which means being the person who knows where every one of the thousand steps currently is. You are close to the work and a long way from the person who ends up using it.

  • is illustrated by Grew the product function from a single practitioner

    Covering four workstreams myself meant I was nominally on all of them and actually next to none of them. Splitting the work was the only way any of it got done.

  • is broken by Learning to work at a different altitude

    I am further from the thing being made, and that distance is real. What I get for it is altitude: I can see things coming for us that I could not see as an individual contributor. The work changes rather than shrinks — everything I do now is in support of the team and the product, instead of just the product I owned.

  • is illustrated by ML personalization inside the CMS

    Personalization stopped being an engineering request the moment it became a field a merchandiser could fill in themselves.

  • is illustrated by VW Dealer iOS application

    Dealers used it standing on a lot with a customer waiting. You cannot design for that from a conference room, and we did not, for a while.

  • is illustrated by Hubs and Collections

    These are still key features of the app, and the building blocks survived all the way into the CMS we are building now. Getting the shape right the first time is why they outlived the platform they were built on.

  • is illustrated by The Stratocaster

    Purchased from my uncle for my fifteenth birthday — a Strat he was selling anyway to upgrade. The person teaching me had owned it, so every question about it had an answer a phone call away.

Say the thing plainly

id
explicit-over-assumed
kind
principle
domains
writing · business
when
2016

I would rather ask an obvious question than carry an assumption. Unstated expectations are the ones that get missed.

18 connections

  • reinforces Change your mind when the information changes

    You cannot be corrected on a position you never stated. Saying the thing plainly is what makes changing your mind possible.

  • reinforces Credibility is delivered, not requested

    Saying plainly what you will deliver and when is the only version of trust anyone can check.

  • reinforces People over process, every time

    Process exists to carry expectations between people. Say them out loud and you need less of it.

  • reinforces You can't name the thing you're wrong about

    If I cannot audit my own blind spot, the next best thing is stating the position plainly enough that somebody else can find it for me.

  • was taught by “You have more power (to influence) than you think you do”

    Said to all of us, somewhat crudely, and it landed. We were trusted experts at a product-led company that wanted our ideas — so not speaking up in the meeting wastes both the moment and the position.

  • is illustrated by A channel for each kind of message

    Matching the channel to the message is the same instinct as asking the obvious question. It removes a guess about how urgent something is before anyone has to make one.

  • is illustrated by A weekly cross-team product review board

    Prioritization arguments were happening anyway, privately, as escalations. The board did not create the conflict, it just made everyone have it in the same room.

  • was taught by Drum major, Rose Parade

    You cannot hint at a tempo to four hundred people in motion. Every signal has to be one everyone already agreed the meaning of, months earlier, because at that distance nothing can be clarified.

  • is illustrated by The calendar is a contract

    A calendar kept honestly, including the unavailable blocks, is a public statement of where my time actually goes.

  • is illustrated by Backend and platform teams sequence first

    The whole mechanism is publishing what will exist and when, in writing, before anyone plans around it. Most dependency failures are just an assumption nobody said out loud.

  • is illustrated by Claude Code executive briefing system

    Most of what it does is force things into writing on a schedule. Risks and goals that live only in someone's head are the ones that surprise you.

  • is illustrated by Developed four PMs into independent owners

    Ownership without a stated success measure is just blame with extra steps. The measures were the part that made it real.

  • is illustrated by Divided the platform into owned product areas

    Naming an owner per area is the least ambiguous thing you can do. Shared ownership is the polite version of nobody deciding.

  • is illustrated by Froze new development on the legacy CMS

    The freeze only held because it was stated the same way to every group, in public, with no quiet exceptions. The moment one team gets a carve-out it is not a policy, it is a queue.

  • is illustrated by LLM-assisted tooling

    Every time it produced something wrong, the cause was a rule I had assumed instead of written down. The fix was never a cleverer prompt, it was saying the obvious thing out loud in a file.

  • is illustrated by Brian Cox

    The whole method is saying a genuinely difficult thing plainly, without pretending it is simpler than it is.

  • is illustrated by Paramount+ and Pluto TV convergence

    The hard part was never the platform. It was getting several leaders to agree, out loud and in the same words, on what would not come across.

  • is illustrated by SAFe and Scrum certifications

    What the frameworks are actually good at is giving a room shared words, so people stop meaning different things by the same term.

Change your mind when the information changes

id
willing-to-be-wrong
kind
principle
domains
product · writing
when
2018

I don't have all the answers and the people around me usually hold the missing piece. The catch is that it only works in retrospect — you cannot name the belief you are wrong about until something proves it.

12 connections

  • is in tension with You can't name the thing you're wrong about

    One says stay ready to change your mind. The other says you cannot run ahead of the information — it is hard to know what you are wrong about until you are actually proven wrong about it. Being willing is not the same as being able, and the willingness is still the only part I control.

  • is in tension with A small fixed palette beats infinite options

    Staying open to better information argues for keeping options alive. Committing to a small palette argues for closing them. Both are right and they pull against each other constantly.

  • reinforces Say the thing plainly

    You cannot be corrected on a position you never stated. Saying the thing plainly is what makes changing your mind possible.

  • reinforces What counts as contributing changes when the job does

    Each time the job changes, the version of yourself that was good at the last one is the thing in the way.

  • is illustrated by Mid-quarter checkpoints

    The checkpoint exists to make changing the plan a normal event on a calendar rather than an admission of failure.

  • is illustrated by “You have more power (to influence) than you think you do”

    You only find out whether the idea was any good by saying it out loud to people who might disagree.

  • is illustrated by A weekly cross-team product review board

    Half the value was hearing why someone else's item mattered more than mine. That only works if you turn up prepared to lose the argument.

  • was taught by BA, Art Studio — UC Santa Barbara

    In a critique you defend the work and then watch someone find the thing you could not see. You learn quickly that being right matters less than the piece getting better.

  • is illustrated by Developed four PMs into independent owners

    Giving someone real ownership means their call stands even when it is not the one I would have made. That is the whole transaction and it is uncomfortable in exactly the way it should be.

  • is illustrated by Learned on acoustic before electric

    I wanted a Gibson 'The Paul' and was told no by the person teaching me. Being overruled by someone who knew more was the whole value of the transaction.

  • is illustrated by Black holes and spacetime

    A field that has rewritten its own foundations repeatedly and treats that as the good outcome rather than an embarrassment.

  • is illustrated by SAFe and Scrum certifications

    I went and learned the frameworks properly before deciding what I thought of them. Most of my objections are around assuming a certification will teach you how to work as a team. That part still has to be figured out the hard way.

Product and engineering are two sides of one coin

id
product-and-engineering-are-one-job
kind
principle
domains
product · engineering
when
2021

Each makes the other better. Doing either without the other produces something nobody asked for.

6 connections

  • reinforces AI is a tool, not a solution

    AI collapses the distance between deciding and building, which only makes the point sharper: skip the product work and you get a fast, confident answer to the wrong question.

  • reinforces Sit next to the person doing the work

    The reason to sit next to engineering is the same as the reason to sit next to the operator: the decision improves when you can see what it costs.

  • is illustrated by Backend and platform teams sequence first

    Sequencing the platform first is not deference to engineering, it is arithmetic. Dependent teams cannot commit to what does not exist yet.

  • is illustrated by ML personalization inside the CMS

    The ML team had the models and no route to an audience. We had the surface and no models. Neither half was a product on its own.

  • is illustrated by pauldenni.com

    I wrote the data model and I built it. The model held, but seeing it as dots and lines made the tweaks obvious in a way reading the schema never did.

  • is illustrated by The enterprise content management platform

    A platform for internal users has no marketing to hide behind. If the product call is wrong the engineering is wasted, and everyone involved finds out in the same standup.

AI is a tool, not a solution

id
ai-is-a-tool-not-a-solution
kind
principle
domains
engineering · product
when
2024

It is trained on the human experience and still only simulates it. Treat it as the answer and you discount the thing that makes a product worth using.

8 connections

  • reinforces Recombination is invention

    A model recombines too, and faster than anyone. What it does not bring is a reason for the particular choice, which is the only part I care about.

  • reinforces Product and engineering are two sides of one coin

    AI collapses the distance between deciding and building, which only makes the point sharper: skip the product work and you get a fast, confident answer to the wrong question.

  • reinforces The canonical choice is usually right

    A tool is judged on whether it does the job, not on how novel it is. That was true of frameworks before it was true of models.

  • was taught by LLM-assisted tooling

    It holds context I cannot hold and surfaces things I would have missed for another week. What it does not do is decide which of them is worth spending capital on, and that was always the actual job.

  • is illustrated by pauldenni.com

    Built with Claude Code, and the part that decides whether it is any good is the prose on the edges, which no model can write for me because it is a claim about how I think.

  • is illustrated by Playing with other people

    The clearest case I know of the human component a model cannot simulate. It is not accuracy that is missing; perfect accuracy is exactly the problem.

  • is illustrated by 90s hip hop

    A model recombines at a scale no producer could. What it cannot supply is why these four bars, from this record, under this verse — and that choice is the entire difference between a beat and a search result.

  • is illustrated by Claude Code executive briefing system

    It synthesizes the signal and I still decide what matters. The judgement is the job; the synthesis was never the job.

Recombination is invention

id
recombination-is-invention
kind
principle
domains
writing · design · product
when
1995

Finished work is raw material. You chop it into a palette and improvise across the top, and what comes out is not imitation. It is one of the oldest things people do with what they inherit.

8 connections

  • reinforces AI is a tool, not a solution

    A model recombines too, and faster than anyone. What it does not bring is a reason for the particular choice, which is the only part I care about.

  • reinforces Taste is mostly subtraction

    The subtraction happens in the crate, not in the song. You pass ten thousand records to land on four bars, and the passing is where the taste lives.

  • is in tension with The canonical choice is usually right

    One says take the thing everyone already uses. The other says the work is in chopping finished things into something nobody has heard yet. I believe both and they do not sit still together.

  • was taught by 90s hip hop

    Take several fully formed ideas, chop them up, improvise another perfect idea on top, and end up with something completely unique. That is not borrowing, it is one of the most human things there is, and it shows how infinitely creative we are.

  • is illustrated by Learning solos by ear

    Working a part out by listening is the same muscle a producer uses digging for a sample: hearing what is in there, and deciding which part of it matters.

  • is illustrated by Playing with other people

    Improvising off what someone else just played is recombination happening live, with no chance to revise.

  • is illustrated by Piano

    Every instrument I picked up afterwards got understood in terms of the keyboard. Starting somewhere and translating outward is how I have learned most things since.

  • is illustrated by Saxophone

    A single line inside an arrangement, meaningless alone and load-bearing in context. Six years of learning that a part only means something next to the other parts.

Imagination needs a workout

id
imagination-needs-a-workout
kind
principle
domains
writing · design
when
1985

The questions too big to answer are the ones worth sitting with. Awe is a working state, not a distraction from one.

4 connections

  • reinforces Thinking time is the cheapest time in the project

    Both are arguments for time that produces nothing you can show anyone that afternoon.

  • was taught by Back to the Future

    It captured my imagination as a kid in the eighties and never let go. A film does not have to teach you anything to be load-bearing; sometimes it just has to show you a time-traveling, flying car.

  • is illustrated by Black holes and spacetime

    Black holes, spacetime, where we fit inside it. It gives my imagination a workout no work problem ever has, and the data coming back now makes it hard not to sit in awe of the thing.

  • is illustrated by Brian Cox

    He makes it easy to get excited about physics, which is a rarer skill than the physics. Getting someone to care is the hard half of any explanation.

Learn on the version that won't let you cheat

id
learn-on-the-unforgiving-version
kind
principle
domains
guitar · engineering
when
1995

Start on the forgiving tool and you learn the shortcuts. Start on the one that punishes you and you learn the instrument.

5 connections

  • is a special case of Practice makes permanent, not perfect

    If whatever you rehearse is what sticks, then the tool that lets you rehearse a shortcut is the dangerous one.

  • was taught by Learned on acoustic before electric

    An electric will let you fake a barre chord on two strings and sound fine. An acoustic will not, so you learn to actually fret it. That decision paid dividends for the next thirty years but I resented it for about 15 minutes in Guitar Center.

  • is illustrated by The Guild acoustic

    Heavier strings, higher action, and no amplifier to hide behind. Everything wrong with your left hand is audible.

  • was taught by The uncle who taught me guitar

    He made the call I would not have made for myself, and he was right. Most of what I know about the instrument traces back to being told no once.

  • is illustrated by Piano

    Every note laid out in front of you with nowhere to hide a wrong one. Starting there at four is why I can hear intervals on any instrument I have picked up since.

Credibility is delivered, not requested

id
credibility-is-delivered-not-requested
kind
principle
domains
product · business
when
2023

Nobody owes you the benefit of the doubt on a long project. 'Trust me, it will be better' is not an argument — shipped increments are.

7 connections

  • reinforces Move as fast as you can without making rework

    Increments are how you avoid rework and how you prove you are worth waiting for. The same practice does both jobs.

  • reinforces Say the thing plainly

    Saying plainly what you will deliver and when is the only version of trust anyone can check.

  • is illustrated by Built the team out with high performers

    Nobody was going to take a multi-year rebuild on faith. What made it credible was releases actually landing, and that came down to who was building it.

  • was taught by Fifty incremental releases, not one cutover

    Rolling out feature by feature instead of arriving three years later let people see real progress, and see that the new system actually did what I had promised. That is what turned it around — not the argument, the evidence.

  • is broken by Froze new development on the legacy CMS

    The freeze cost me politically and it should have. Stakeholders had no way to know whether a project that size would land, and I was asking for trust while offering nothing to check it against. I knew that going in, which is why the new CMS went out in increments rather than all at once.

  • is illustrated by Paramount+ and Pluto TV convergence

    A program that size is asking several organizations to believe a claim about the future. The only currency is what has already shipped.

  • is illustrated by The ground-up CMS rebuild

    Fifty releases into a live publishing operation, and the thing that changed people's minds was not the roadmap. It was watching the new system quietly get better than the old one.

What counts as contributing changes when the job does

id
contribution-changes-shape
kind
principle
domains
business · product
when
2022

Every step away from the work redefines what your contribution even is, and each time you have to relearn it rather than keep doing the last version faster.

4 connections

  • reinforces Change your mind when the information changes

    Each time the job changes, the version of yourself that was good at the last one is the thing in the way.

  • is illustrated by Built a two-tier product organization

    Managing five people directly was a job I knew how to do and could still do today. Choosing not to is what moved my contribution to the structure itself: who owns which pillar, who is being grown into what, and what the team can absorb next.

  • was taught by Learning to work at a different altitude

    I did the work and got promoted, and then had to work out what contributing even meant from there. Doing the old job faster was not one of the available answers.

  • is illustrated by Developed four PMs into independent owners

    The output stopped being decisions I made and became decisions other people were empowered to make. Not being able to point to a feature and say "I did that" took some getting used to.

The work you disappear into is the work that teaches you

id
making-is-meditative
kind
principle
domains
guitar · design · writing
when
2015

Playing music lets me focus completely and, in the same motion, find things I could not see from outside it. Product and engineering do the same thing when they force you into someone else's position.

8 connections

  • reinforces Sit next to the person doing the work

    Disappearing into the work is the most complete form of sitting next to it.

  • reinforces Thinking time is the cheapest time in the project

    Both are a defense of time that produces nothing you can show anyone that afternoon, and returns something you could not have got any faster.

  • was taught by Piano

    Thirty years of it, and the reason is not that I am good. Playing lets me focus completely and, in the same motion, notice things about myself I cannot get at from outside.

  • is illustrated by Learning solos by ear

    You cannot do it while thinking about anything else, which is most of the appeal.

  • is illustrated by Linoleum block relief printing

    Cutting a block is slow, quiet and completely absorbing. Most of what I have worked out about my own taste, I worked out with a knife in my hand.

  • is illustrated by pauldenni.com

    The same absorption as learning a piece. Somewhere in the middle of it I stopped building a website and started finding out what I actually think, and how these ideas connect to each other

  • is illustrated by Playing with other people

    The most complete version of it. You cannot be anywhere else while it is happening.

  • is illustrated by The enterprise content management platform

    Product and engineering have the same quality when you let them: they force you outside your own position and into someone else's, which is where the useful perspective is.

You can't name the thing you're wrong about

id
cant-audit-your-own-blind-spot
kind
principle
domains
writing · product
when
2026

Holding your opinions loosely does not let you find the wrong one in advance. That arrives with the information, not before it.

5 connections

  • is in tension with Change your mind when the information changes

    One says stay ready to change your mind. The other says you cannot run ahead of the information — it is hard to know what you are wrong about until you are actually proven wrong about it. Being willing is not the same as being able, and the willingness is still the only part I control.

  • reinforces Say the thing plainly

    If I cannot audit my own blind spot, the next best thing is stating the position plainly enough that somebody else can find it for me.

  • is in tension with Thinking time is the cheapest time in the project

    Thinking a problem all the way through is the best defense I have, and it is not a guarantee. Some errors are only visible from a position you have not reached yet.

  • is illustrated by “You have more power (to influence) than you think you do”

    Saying the idea out loud is how it gets tested. The ones I keep to myself stay unexamined and I go on assuming they are fine.

  • is illustrated by Black holes and spacetime

    A field that keeps discovering the thing it could not have known to look for. The frontier moves and the previous certainty was never flagged as provisional at the time.

role9

Director, Technical Product Management

id
role-director-tpm
kind
role
domains
product · business
when
2024-01

Own product strategy for the enterprise CMS behind Paramount+ and Pluto TV, and a five-person product organization through two leads.

5 connections

  • contains Built a two-tier product organization

    The shape the team was built toward rather than the one it fell into. Three pillars, each with a lead who owns it well enough to advise other teams on it later, and I intend to promote into the third.

  • contains ML personalization inside the CMS

    A partnership with the ML platform team, run from the CMS side.

  • came after Senior Manager, Technical Product Management

    Same platform, wider remit: from running the operating rhythm to owning the portfolio and the people running it.

  • contains The enterprise content management platform

    The platform is the job. Everything else in the role exists to move it.

  • contains LLM-assisted tooling

    The job is holding twenty threads across programs, people and risk, and continuity is the first thing to go when the week gets loud. There was no chief of staff coming, so I built the function rather than ask for it.

Senior Manager, Technical Product Management

id
role-senior-manager-tpm
kind
role
domains
product · business
when
2022-10

Built the operating rhythm with the PMO and developed four PMs into independent owners of their platform areas.

6 connections

Manager, Technical Product Management

id
role-manager-tpm
kind
role
domains
product · business
when
2022-04

Grew the CMS product function from one practitioner to four and split the platform into owned areas.

5 connections

Senior Technical Product Manager

id
role-senior-tpm
kind
role
domains
product · engineering
when
2020-10

Launch and expansion features for Paramount+ while it went from 8 million to 40 million subscribers.

4 connections

Technical Product Manager

id
role-tpm
kind
role
domains
product · engineering
when
2017-06

The CMS's first page builder, the content operations tooling, and side navigation across the OTT apps.

5 connections

Senior Digital Producer

id
role-senior-digital-producer
kind
role
domains
product · design
when
2015-12

Migrated 40+ B2B and B2C properties onto Adobe Experience Manager across a decentralized portfolio of brands.

3 connections

Integrated Digital Producer

id
role-integrated-producer
kind
role
domains
design · business
when
2012-12

Volkswagen's digital platform, the VW Dealer iOS app, and integrated paid media campaigns.

3 connections

  • came after Interactive Producer

    Bigger accounts, longer engagements, the first work that outlived its launch.

  • came before Senior Digital Producer

    From campaigns to platforms, which turned out to be the direction the whole career was heading.

  • contains vw.com

    The account the role was built around.

Interactive Producer

id
role-interactive-producer
kind
role
domains
design · engineering
when
2011-06

Entertainment websites and mobile and social games.

3 connections

Project Manager

id
role-project-manager
kind
role
domains
business · design
when
2009-03

Gaming and entertainment websites and mobile apps. The first job in the industry.

2 connections

project18

The enterprise content management platform

id
enterprise-cms
kind
project
domains
product · engineering
when
2017

The system through which content operations, merchandising, editorial and marketing publish every title, collection and page across Paramount+ and Pluto TV. 600+ internal users.

10 connections

The ground-up CMS rebuild

id
cms-ground-up-rebuild
kind
project
domains
product · engineering
when
2022

A multi-year replacement of the legacy platform, delivered as 50 incremental releases without interrupting a 24/7 live publishing operation.

11 connections

Paramount+ and Pluto TV convergence

id
paramount-pluto-convergence
kind
project
domains
product · business
when
2024

Company-wide initiative to run two independently operated services on a single platform and backend stack. I own the CMS workstream.

4 connections

The legacy CMS

id
legacy-cms
kind
project
domains
engineering · product
when
2017

The system the rebuild replaces. Frozen to all but essential development for two years while its successor was built.

3 connections

  • illustrates Nice to haves are must haves

    Every feature in it arrived as a nice to have that somebody made a case for, and every case was reasonable on its own. Years of reasonable cases produced a system nobody could safely change.

  • illustrates Optionality is a tax you pay forever

    Five ways to build a carousel of movie and show content, where one would have done. The differences between them were small enough that every change started with working out which one you were in.

  • came before The ground-up CMS rebuild

    The old system is the reason the new one has the shape it does, including the parts we deliberately did not carry across.

The CMS's first custom page builder

id
custom-page-builder
kind
project
domains
product · design
when
2019

Let editorial compose and deploy pages themselves. Cut developer dependency for page creation by 84%.

6 connections

  • breaks Optionality is a tax you pay forever

    We made it almost too flexible. It gets used for many more cases than we designed for, people invent new ones constantly, and there is very little it will refuse to do. That is the blessing and it is exactly the curse.

  • taught me Sit next to the person doing the work

    Product brought me the problem, we brainstormed it, and the front-end engineers wanted it more than anyone — it saved them building every page from scratch and let them start templatizing. Nobody had to be convinced it was needed.

  • illustrates People over process, every time

    The ticket queue was a process standing in for a conversation. Putting the tool in editorial's hands deleted the queue rather than optimizing it, and gave engineering back the work they actually wanted to do.

  • illustrates The default is the product

    Whatever template loaded first became the house style. We shipped a layout choice and got an editorial standard we never wrote down.

  • is part of Technical Product Manager

    Shipped as an individual contributor, and still the thing I would point at first.

  • is part of The enterprise content management platform

    The first capability that made the CMS something operators used directly rather than filed tickets against.

Content operations tooling

id
content-ops-tooling
kind
project
domains
product · engineering
when
2019

Cut manual curation time for metadata and assets by 90% and improved QA throughput by 50%.

5 connections

Side navigation across the OTT apps

id
side-navigation
kind
project
domains
design · product
when
2020

Replaced the existing navigation across CBS All Access platforms. Drove a 10.6% increase in minutes watched per user.

5 connections

  • illustrates The default is the product

    Nobody asked for different navigation. Changing what was there by default moved minutes watched per user by 10.6%.

  • illustrates The canonical choice is usually right

    Side navigation was not a new idea. It was the pattern the rest of the category had already settled on, and adopting it late still paid.

  • illustrates Taste is mostly subtraction

    It was mostly a removal. Fewer competing entry points, and the number that moved was minutes watched.

  • came before Hubs and Collections

    Navigation came first. The surfaces it had to point at came after.

  • is part of Technical Product Manager

    The one piece of consumer-facing work in an otherwise internal-platform career.

Hubs and Collections

id
hubs-and-collections
kind
project
domains
product · design
when
2021

Content discovery surfaces introduced to Paramount+ during the subscriber expansion, alongside premium bundles.

7 connections

  • illustrates Nice to haves are must haves

    Built for the expansion we could see coming. Each one was a complex feature on its own, and together they were the building blocks that made the growth possible — which is the one period where adding surface genuinely was the strategy.

  • illustrates Sit next to the person doing the work

    These are still key features of the app, and the building blocks survived all the way into the CMS we are building now. Getting the shape right the first time is why they outlived the platform they were built on.

  • illustrates The default is the product

    A hub is a merchandising default, and we had lively debates about what one even was. Whatever it surfaces first is what most subscribers will ever see of the catalog.

  • is part of Senior Technical Product Manager

    Delivered during the subscriber expansion, when discovery surfaces were the strategy.

  • came before The ground-up CMS rebuild

    Built years earlier on the old platform, and carried across intact. The blocks outlived the system they were made in.

  • is depended on by Live sports integration

    Live events needed somewhere to live in the app before they could be published to it.

  • came after Side navigation across the OTT apps

    Navigation came first. The surfaces it had to point at came after.

Live sports integration

id
live-sports-integration
kind
project
domains
product · engineering
when
2021

Brought live sports into the Paramount+ application, which meant the publishing path had to work against a clock nobody controls.

5 connections

ML personalization inside the CMS

id
ml-personalization-in-cms
kind
project
domains
product · engineering
when
2024

Personalization and recommendation services integrated directly into the CMS, so merchandisers deploy personalized carousels without engineering involvement.

5 connections

LLM-assisted tooling

id
llm-assisted-tooling
kind
project
domains
product · writing
when
2025

An operating system I built for the parts of the job that do not fit in my head. It synthesizes signal across programs and people, holds continuity between weeks, and surfaces risk while it is still cheap to act on.

5 connections

  • taught me AI is a tool, not a solution

    It holds context I cannot hold and surfaces things I would have missed for another week. What it does not do is decide which of them is worth spending capital on, and that was always the actual job.

  • illustrates Say the thing plainly

    Every time it produced something wrong, the cause was a rule I had assumed instead of written down. The fix was never a cleverer prompt, it was saying the obvious thing out loud in a file.

  • breaks Thinking time is the cheapest time in the project

    The risk I built into my own week: I used to carry the org in my head between updates, and carrying it there was where the connections got made. Now the system carries it and I read the output, which is not the same as having done the noticing.

  • came before Claude Code executive briefing system

    The LLM assisted tooling came first and taught me what the model was actually good at. The briefing system is the same idea aimed at my own week.

  • is part of Director, Technical Product Management

    The job is holding twenty threads across programs, people and risk, and continuity is the first thing to go when the week gets loud. There was no chief of staff coming, so I built the function rather than ask for it.

Claude Code executive briefing system

id
claude-code-exec-briefs
kind
project
domains
product · writing
when
2026

Synthesizes meeting and program signal into executive briefs, risk and goal tracking, and 1:1 prep.

5 connections

  • illustrates AI is a tool, not a solution

    It synthesizes the signal and I still decide what matters. The judgement is the job; the synthesis was never the job.

  • illustrates Say the thing plainly

    Most of what it does is force things into writing on a schedule. Risks and goals that live only in someone's head are the ones that surprise you.

  • breaks Thinking time is the cheapest time in the project

    It hands me a synthesized week in seconds, and the hour I used to spend assembling that picture myself was an hour of thinking. I am not yet sure what I traded away.

  • came after LLM-assisted tooling

    The LLM assisted tooling came first and taught me what the model was actually good at. The briefing system is the same idea aimed at my own week.

  • came after pauldenni.com

    Both came out of the same year of finding out what this class of tool can and cannot do.

pauldenni.com

id
pauldenni-com
kind
project
domains
design · engineering · writing
when
2026

This site. A knowledge graph rather than a portfolio, where objects connect through principles and a path between two of them reads as a paragraph.

5 connections

Adobe Experience Manager migration

id
aem-migration
kind
project
domains
engineering · business
when
2016

40+ B2B and B2C properties moved onto one platform, standardizing content operations across a decentralized portfolio of brands.

4 connections

VW Dealer iOS application

id
vw-dealer-ios
kind
project
domains
design · engineering
when
2014

An application built for the people selling the cars rather than the people buying them.

3 connections

  • taught me Sit next to the person doing the work

    Built for the people selling the cars, not the people buying them. The first product I worked on where the user was somebody doing a job, and I have mostly built for that user ever since.

  • illustrates Sit next to the person doing the work

    Dealers used it standing on a lot with a customer waiting. You cannot design for that from a conference room, and we did not, for a while.

  • is part of vw.com

    Two halves of the same account: the public platform and the tool the dealers used.

vw.com

id
vw-dot-com
kind
project
domains
design · business
when
2013

Volkswagen's digital platform, run alongside integrated paid media campaigns.

4 connections

Entertainment sites and social games

id
social-and-entertainment-games
kind
project
domains
design · engineering
when
2012

Promotional sites and mobile and social games for entertainment properties, on schedules set by release dates rather than by the work.

5 connections

  • breaks Move as fast as you can without making rework

    Everything shipped against a date that existed before the brief did. Nothing was ever finished, it was due — which is exactly the speed that produces work you throw away and do again.

  • taught me Sit next to the person doing the work

    Project management, which means being the person who knows where every one of the thousand steps currently is. You are close to the work and a long way from the person who ends up using it.

  • breaks Taste is mostly subtraction

    The brief was always more: more features, more interactions, more things on the screen. Nobody ever asked what could come out.

  • is part of Interactive Producer

    The whole of that role: whatever the studio was releasing that quarter.

  • came after Gaming launch sites and mobile apps

    Same category of work, one agency later, with slightly more say in how it got made.

Gaming launch sites and mobile apps

id
gaming-launch-sites
kind
project
domains
design · business
when
2010

Launch campaigns for game titles. My first work on the delivery side, and where I learned what a date actually costs.

5 connections

  • taught me Move as fast as you can without making rework

    A game ships on its date and the site ships with it or it is pointless. My first lesson that some deadlines are facts about the world rather than negotiating positions.

  • taught me People over process, every time

    Eight years across four agencies came down to one skill: taking a client's idea and getting it to delivery, through the thousand small steps in between. No process ever walked those steps for me.

  • breaks Practice makes permanent, not perfect

    Every one of these was built fast and thrown away, so nothing got done properly twice. I learned habits on those projects that took years to unlearn.

  • is part of Project Manager

    The first project management job, and the work it consisted of.

  • came before Entertainment sites and social games

    Same category of work, one agency later, with slightly more say in how it got made.

decision16

Froze new development on the legacy CMS

id
froze-legacy-cms
kind
decision
domains
product · business
when
2022

Two years of saying no to every stakeholder group, uniformly, so the replacement could be resourced properly.

6 connections

  • breaks Credibility is delivered, not requested

    The freeze cost me politically and it should have. Stakeholders had no way to know whether a project that size would land, and I was asking for trust while offering nothing to check it against. I knew that going in, which is why the new CMS went out in increments rather than all at once.

  • breaks Move as fast as you can without making rework

    Two years of deferring near-term requests to avoid throwaway work. It is the no-rework half of the principle honored completely and the go-fast half abandoned, and every deferred request was somebody's real problem in the meantime.

  • illustrates Ship the boring thing to kill the expensive thing

    Two years of saying no to the old system bought one year of not rebuilding the new one twice.

  • illustrates Say the thing plainly

    The freeze only held because it was stated the same way to every group, in public, with no quiet exceptions. The moment one team gets a carve-out it is not a policy, it is a queue.

  • depends on The ground-up CMS rebuild

    The freeze is what funded the rebuild. Without it the same people were maintaining two systems.

  • is part of Senior Manager, Technical Product Management

    Held across this role and the next, which is most of why it worked.

Backend and platform teams sequence first

id
backend-sequences-first
kind
decision
domains
product · engineering
when
2022

Quarterly commitment cycles built so dependent teams know what capabilities will exist before they commit to using them.

6 connections

A weekly cross-team product review board

id
weekly-review-board
kind
decision
domains
product · business
when
2022

One recurring forum where feature prioritization happens in the open, instead of in escalations.

4 connections

Mid-quarter checkpoints

id
mid-quarter-checkpoints
kind
decision
domains
product · business
when
2022

A scheduled point to catch drift and reallocate, rather than discovering it at the quarter boundary.

3 connections

Divided the platform into owned product areas

id
platform-areas-with-owners
kind
decision
domains
product · business
when
2022

Distinct areas with dedicated ownership. The structure the organization still runs under.

4 connections

Built a two-tier product organization

id
two-tier-product-org
kind
decision
domains
business · product
when
2024

Five people through leads who each manage their own PMs, rather than a flat team reporting to me. Three pillars in the CMS product, two of them with a lead today, and the third waiting on someone growing into it.

7 connections

  • breaks Sit next to the person doing the work

    Two tiers means most of the work now reaches me through someone else's account of it. That is the part of going from IC to manager I have had to keep working at — further from the making, but a higher vantage on what is coming for us. Still the right call for the team.

  • illustrates What counts as contributing changes when the job does

    Managing five people directly was a job I knew how to do and could still do today. Choosing not to is what moved my contribution to the structure itself: who owns which pillar, who is being grown into what, and what the team can absorb next.

  • illustrates People over process, every time

    Two leads managing their own PMs rather than five people reporting to me. The org chart is a statement about who gets to grow, and a flat one says nobody.

  • is part of Director, Technical Product Management

    The shape the team was built toward rather than the one it fell into. Three pillars, each with a lead who owns it well enough to advise other teams on it later, and I intend to promote into the third.

  • is the same idea as Drum major, Rose Parade

    My first year I was a side drum major: the lead gave a signal and I ran back to the middle of the band to relay it, which is why six miles of parade route came to about nine miles of actual running. That is the cost of a management layer, paid in legs. I now relay through my leads instead, but the distance is still real.

  • came after Grew the product function from a single practitioner

    One person to four, then five through leads. The second step was harder than the first.

  • came after Learning to work at a different altitude

    IC to manager first, then manager to managing managers. The second was the harder relearning.

Grew the product function from a single practitioner

id
grew-team-from-one
kind
decision
domains
business · product
when
2021

I was running every product workstream myself as the company geared up to relaunch CBS All Access as Paramount+. Growing the product function to four PMs is what let it keep pace with the rest of the CMS team — a project manager, forty engineers, QA and design.

4 connections

Developed four PMs into independent owners

id
developed-four-pms
kind
decision
domains
business · product
when
2023

Expanded their scope and set clear success measures, so prioritization happened at the team level instead of escalating.

6 connections

Fifty incremental releases, not one cutover

id
incremental-not-big-bang
kind
decision
domains
engineering · product
when
2022

The rebuild shipped continuously against a live 24/7 publishing operation rather than landing as a single switch.

5 connections

Played one guitar for a year

id
one-guitar-year
kind
decision
domains
guitar
when
2020

Put everything else away and played a single instrument until I stopped thinking about the instrument.

4 connections

Learned on acoustic before electric

id
learned-on-acoustic-first
kind
decision
domains
guitar
when
1995

I wanted an electric — a Gibson SG - 'The Paul'. My uncle said acoustic first, because starting on electric means never really learning the instrument. He was right, and I'm glad I listened.

4 connections

Bought the Les Paul when I graduated

id
bought-the-les-paul-at-graduation
kind
decision
domains
guitar
when
2003

The grail guitar, bought for myself to mark finishing the degree. A milestone in more ways than one.

2 connections

Learning to work at a different altitude

id
ic-to-manager-transition
kind
decision
domains
business · product
when
2021

I did the work and got promoted. Then I had to figure out what contributing meant when the job was no longer the thing I was building myself.

4 connections

Built the team out with high performers

id
hired-high-performers
kind
decision
domains
business · product
when
2023

Every hire had to learn our custom platform. Nobody was starting blind, though: some had done a rebuild and migration before, others had come up on backend systems like a CMS. That is what I hired them for.

3 connections

  • illustrates Credibility is delivered, not requested

    Nobody was going to take a multi-year rebuild on faith. What made it credible was releases actually landing, and that came down to who was building it.

  • illustrates People over process, every time

    I could have arranged the team a dozen ways on paper. Who was actually on it mattered more than any of those arrangements.

  • depends on The ground-up CMS rebuild

    Fifty releases against a live site is not work you hand to whoever is free. The rebuild needed the people before it could pay anything back.

A channel for each kind of message

id
communication-defaults
kind
decision
domains
business · writing
when
2020

Slack for quick updates and questions, Zoom for the regular meetings and the five-minute sidebars, email for longer non-urgent things, phone and text for outside hours.

3 connections

  • illustrates Say the thing plainly

    Matching the channel to the message is the same instinct as asking the obvious question. It removes a guess about how urgent something is before anyone has to make one.

  • illustrates People over process, every time

    It is not a policy anyone has to follow. It is me telling people how to reach me so they do not have to work it out.

  • contains The calendar is a contract

    Two halves of the same thing: how to reach me, and when.

The calendar is a contract

id
calendar-as-a-contract
kind
decision
domains
business · product
when
2021

Kept current including the blocks where I am unavailable, standing meetings pushed to the morning across the org, and invites for a morning meeting sent before 6pm the day before.

4 connections

artifact10

Cherry sunburst Les Paul Standard

id
les-paul-standard
kind
artifact
domains
guitar
when
2003

The grail guitar. Not the fanciest, and unquestionably the most copied electric design there is. In that configuration it is one of the sweetest sounds in rock and blues.

5 connections

  • taught me A small fixed palette beats infinite options

    Two pickups and a switch. You can't configure your way out of a bad take, so you learn the instrument instead.

  • illustrates The canonical choice is usually right

    The obvious choice, and the right one on the merits. It is the most copied electric design there is and that configuration is still the epitome of what a guitar is to me. Being unremarkable to name has nothing to do with being unremarkable to hear.

  • illustrates Practice makes permanent, not perfect

    One instrument, one scale length, one set of habits. Everything I built that year still works because it was built on something that did not move.

  • is depended on by Bought the Les Paul when I graduated

    This is the guitar. The decision and the object are the same event.

  • is depended on by Played one guitar for a year

    It was this guitar. The experiment does not work with an instrument you are still deciding about.

The parts Telecaster

id
tele-partscaster
kind
artifact
domains
guitar · engineering
when
2015

A guitar assembled from chosen components, every one of which remains a decision I can revisit.

4 connections

An unreasonable number of electric guitars

id
too-many-guitars
kind
artifact
domains
guitar

More instruments than anyone needs, and I am confident I need more. They are for me, not for anyone looking at them. I'm not in a band.

4 connections

  • breaks A small fixed palette beats infinite options

    The clearest counterexample I own. I am certain the principle is right and I am equally certain I need another guitar.

  • illustrates Optionality is a tax you pay forever

    Every one needs strings, setup, and a decision about whether today is its day. The tax is not the money, it is the deciding.

  • illustrates Nice to haves are must haves

    I retired this belief everywhere in my working life but it is alive and well on the wall to my right. More guitars have never once made me a better player, yet I keep buying them.

  • is part of The parts Telecaster

    One of the many, and the one that took the most work to end up least played.

Linoleum block relief printing

id
printmaking
kind
artifact
domains
design
when
2003

Relief printing at home on a small press. You cut away everything that is not the image, and whatever survives the knife is what prints.

7 connections

BA, Art Studio — UC Santa Barbara

id
art-studio-degree
kind
artifact
domains
design
when
2003

A studio art degree, which turns out to be a reasonable preparation for arguing about what to remove.

4 connections

  • taught me Taste is mostly subtraction

    Four years of critique, which is a structured argument about what to remove. It is the most directly useful training I have for the job I actually do.

  • taught me A small fixed palette beats infinite options

    Every assignment came with a restriction: one color, one process, one size. I spent four years learning that the limit is what makes the decision interesting.

  • taught me Change your mind when the information changes

    In a critique you defend the work and then watch someone find the thing you could not see. You learn quickly that being right matters less than the piece getting better.

  • contains Linoleum block relief printing

    It started as coursework, which is where the press was. It outlasted the degree by twenty years and the press is at home now.

The Guild acoustic

id
guild-acoustic
kind
artifact
domains
guitar
when
1995

My first guitar, and not the one I wanted at first. An acoustic makes you fret every string properly or it tells on you.

5 connections

The Stratocaster

id
stratocaster
kind
artifact
domains
guitar
when
1996

My first electric. Purchased from my uncle for my fifteenth birthday, a Strat he was selling anyway to upgrade. The other boring, predictable, everyone-has-one guitar.

4 connections

  • illustrates The canonical choice is usually right

    The other guitar everyone has. Boring to name and still the right answer often enough that it has outlasted every design meant to replace it.

  • illustrates Sit next to the person doing the work

    Purchased from my uncle for my fifteenth birthday — a Strat he was selling anyway to upgrade. The person teaching me had owned it, so every question about it had an answer a phone call away.

  • came after The Guild acoustic

    Acoustic first, as instructed, then the electric I had wanted the whole time.

  • depends on The uncle who taught me guitar

    His Strat before it was mine, sold on when he upgraded.

Piano

id
piano
kind
artifact
domains
guitar · writing
when
1986

Started lessons at four and still play. Not as good as thirty-odd years should make me, but it was the first instrument and it shaped how I hear every other one.

6 connections

  • taught me Practice makes permanent, not perfect

    Thirty years of playing and I am not as good as that should make me, because for a lot of those years I practiced without paying attention.

  • taught me The work you disappear into is the work that teaches you

    Thirty years of it, and the reason is not that I am good. Playing lets me focus completely and, in the same motion, notice things about myself I cannot get at from outside.

  • illustrates Learn on the version that won't let you cheat

    Every note laid out in front of you with nowhere to hide a wrong one. Starting there at four is why I can hear intervals on any instrument I have picked up since.

  • illustrates Recombination is invention

    Every instrument I picked up afterwards got understood in terms of the keyboard. Starting somewhere and translating outward is how I have learned most things since.

  • came before Saxophone

    Piano at four, saxophone at twelve. The second was much easier for having done the first.

  • came before Playing with other people

    Lessons first, alone, for years before any of it was played with anybody.

Saxophone

id
saxophone
kind
artifact
domains
guitar
when
1994

Middle school through high school. Six years of reading charts and playing a line inside a much larger arrangement.

5 connections

  • illustrates Practice makes permanent, not perfect

    Concert band rehearses the same passage until nobody has to think about it, which is exactly what the phrase warns about, aimed on purpose.

  • illustrates Recombination is invention

    A single line inside an arrangement, meaningless alone and load-bearing in context. Six years of learning that a part only means something next to the other parts.

  • contains Drum major, Rose Parade

    Same band, different job: playing a part, then keeping four hundred people together.

  • came after Piano

    Piano at four, saxophone at twelve. The second was much easier for having done the first.

  • came before Learning solos by ear

    Years of reading charts first, then learning to work without them.

Drum major, Rose Parade

id
rose-parade-drum-major
kind
artifact
domains
business · guitar
when
1997

Two years leading a 400-person band down Colorado Boulevard, 1997 and 1998. Six miles of route, and closer to nine actually covered.

6 connections

  • taught me People over process, every time

    The drill chart was written months before. What actually held it together on the day was three of us knowing exactly which commands to give and how to coordinate them, because we had practiced it hundreds of times.

  • taught me Say the thing plainly

    You cannot hint at a tempo to four hundred people in motion. Every signal has to be one everyone already agreed the meaning of, months earlier, because at that distance nothing can be clarified.

  • breaks Move as fast as you can without making rework

    Relaying a signal through a runner is the slowest possible way to move information, and it was the only way that worked at that scale.

  • illustrates Thinking time is the cheapest time in the project

    Every decision worth making was made in rehearsal. On Colorado Boulevard there is no deciding, only executing what was decided.

  • is the same idea as Built a two-tier product organization

    My first year I was a side drum major: the lead gave a signal and I ran back to the middle of the band to relay it, which is why six miles of parade route came to about nine miles of actual running. That is the cost of a management layer, paid in legs. I now relay through my leads instead, but the distance is still real.

  • is part of Saxophone

    Same band, different job: playing a part, then keeping four hundred people together.

reference9

Back to the Future

id
back-to-the-future
kind
reference
domains
writing · design
when
1985

My favorite film by far. It captured my imagination as a kid in the eighties and never let go. Also: DeLoreans are rad.

4 connections

  • taught me Imagination needs a workout

    It captured my imagination as a kid in the eighties and never let go. A film does not have to teach you anything to be load-bearing; sometimes it just has to show you a time-traveling, flying car.

  • illustrates Thinking time is the cheapest time in the project

    Nothing in it is decoration. Every element planted in the first act is carrying weight in the third, which only happens if the whole thing was worked out before a frame was shot.

  • illustrates A small fixed palette beats infinite options

    One town, one street, two eras. It gets enormous range out of deliberately refusing to go anywhere else.

  • illustrates Taste is mostly subtraction

    There is no spare scene in it. Whatever else was written got cut.

Black holes and spacetime

id
astrophysics-reading
kind
reference
domains
writing · engineering

The field I would be in if I were any good at math. Black holes, spacetime, and where we sit inside it — the frontier of what we know as a species, barely scratched.

4 connections

  • illustrates Imagination needs a workout

    Black holes, spacetime, where we fit inside it. It gives my imagination a workout no work problem ever has, and the data coming back now makes it hard not to sit in awe of the thing.

  • illustrates You can't name the thing you're wrong about

    A field that keeps discovering the thing it could not have known to look for. The frontier moves and the previous certainty was never flagged as provisional at the time.

  • illustrates Change your mind when the information changes

    A field that has rewritten its own foundations repeatedly and treats that as the good outcome rather than an embarrassment.

  • illustrates Thinking time is the cheapest time in the project

    Most of the work is sitting with a question you cannot yet run an experiment on. The instruments catch up decades after the thinking does.

90s hip hop

id
nineties-hip-hop
kind
reference
domains
guitar · writing

Nas, Tribe, J5, Dre and Snoop, De La, KRS-One, Biggie, The Roots. Finished ideas taken apart and rebuilt into something that did not exist before.

5 connections

  • taught me Recombination is invention

    Take several fully formed ideas, chop them up, improvise another perfect idea on top, and end up with something completely unique. That is not borrowing, it is one of the most human things there is, and it shows how infinitely creative we are.

  • illustrates AI is a tool, not a solution

    A model recombines at a scale no producer could. What it cannot supply is why these four bars, from this record, under this verse — and that choice is the entire difference between a beat and a search result.

  • illustrates The canonical choice is usually right

    An MPC and a crate of records everyone else also had access to. The source material being unremarkable is the point — what you do with it is the whole art.

  • illustrates A small fixed palette beats infinite options

    An MPC, a crate of records and a small number of bars. The limits were the instrument.

  • illustrates Taste is mostly subtraction

    A sample is a decision about which four bars of someone else's record matter. Everything else on that record gets left behind.

Learning solos by ear

id
learning-solos-by-ear
kind
reference
domains
guitar

Working a part out by listening rather than reading it. Slower, and the only way it stays learned.

6 connections

SAFe and Scrum certifications

id
safe-and-scrum-certs
kind
reference
domains
business · product
when
2018

POPM, CSPO, CSM, SAFe Practitioner. Useful vocabulary, and a reminder that a framework is not a plan.

3 connections

  • breaks People over process, every time

    Four process certifications, and they were worth getting. No framework tells you what the person doing the work already knows. Most teams follow the process instead of asking them, and call it alignment.

  • illustrates Change your mind when the information changes

    I went and learned the frameworks properly before deciding what I thought of them. Most of my objections are around assuming a certification will teach you how to work as a team. That part still has to be figured out the hard way.

  • illustrates Say the thing plainly

    What the frameworks are actually good at is giving a room shared words, so people stop meaning different things by the same term.

The uncle who taught me guitar

id
uncle-who-taught-me
kind
reference
domains
guitar

Taught me to play, made me start on acoustic, and sold me his Strat when I was fifteen.

5 connections

Brian Cox

id
brian-cox
kind
reference
domains
writing · engineering

A physicist who makes it easy to get excited about physics, which is a harder skill than the physics.

2 connections

  • illustrates Imagination needs a workout

    He makes it easy to get excited about physics, which is a rarer skill than the physics. Getting someone to care is the hard half of any explanation.

  • illustrates Say the thing plainly

    The whole method is saying a genuinely difficult thing plainly, without pretending it is simpler than it is.

“You have more power (to influence) than you think you do”

id
more-power-than-you-think
kind
reference
domains
business · product
when
2021

Said to our whole product team by a manager at Paramount. Bluntly put, and right: we were trusted experts at a product-led company that wanted us bringing ideas.

5 connections

Playing with other people

id
playing-with-other-people
kind
reference
domains
guitar

The thing no notation contains and no model reproduces: several people improvising off each other until it locks in.

5 connections

  • illustrates AI is a tool, not a solution

    The clearest case I know of the human component a model cannot simulate. It is not accuracy that is missing; perfect accuracy is exactly the problem.

  • taught me People over process, every time

    You can play a piece note for note, perfectly, following the process exactly — and now a model can too. What is missing is several people improvising off each other and finding a groove. No notation holds it and no perfect rhythm reproduces it.

  • illustrates Recombination is invention

    Improvising off what someone else just played is recombination happening live, with no chance to revise.

  • illustrates The work you disappear into is the work that teaches you

    The most complete version of it. You cannot be anywhere else while it is happening.

  • came after Piano

    Lessons first, alone, for years before any of it was played with anybody.