These are the things Mack has held long enough to have tested, written in his own voice rather than described. How he works, what he believes a team runs on, and where his attention goes when nobody is paying for it.
How I try to make things, whether that is a screen, a codebase or a page of writing.
Simplicity is the ultimate sophistication. Anyone can make a thing complicated; making it simple is the hardest work in any craft, and the version worth aiming at is the one where nobody notices what was taken away. If it cannot be explained to the person deciding on it, the understanding is not finished. A screen, a contract, a diagram, a sentence.
Start by naming the single thing you want someone to do, then make that the one obvious item in front of them. If they should send a question, build the interface around sending a question. If they should scroll, let nothing compete with scrolling. If they should learn a hotkey, show the cause and the effect until it sticks. Everything after that is taking away what gets in the way. What is left is the part nobody can describe, which is how it makes someone feel.
Even a plain website is design. You are deciding what goes where, how the information ranks, whether a stranger can scan it and leave with the point. Done well it says something you never wrote down: the work is careful and the people behind it can be trusted. Most of what I know came from studying products that already work and asking what is doing the job. The messaging, the color, the layout, how fast it explains itself.
Once the fundamentals and the data structures are there, the rest transfers. Each step made the next one shorter: HTML and CSS, then JavaScript, then TypeScript, then Dart. What carries more weight is the framework around the language, judged against what the software has to do and where it runs. Agentic coding has taken more of the syntax burden off again, which leaves the real work where it always was: what the program should do, and everything sitting around the code rather than in it.
The same scrutiny on a one-line fix as on a filing, a contract or a deck. Small errors are rarely just small errors. They are what makes everything near them get doubted. In design the tells are obvious: crowded elements, a typo, type sizes that jump too far, colors that do not sit together, a component behaving differently in two places. None of it is acceptable in something being handed over. And not in bursts: 1,405 active days across seven years, six per cent of them on a weekend.
How I decide whether a thing is worth building at all.
I meet most ideas with skepticism. Not pessimism, just an unwillingness to be enthusiastic before I understand something. It resolves into real optimism when the idea holds up, or realism when it does not and somebody needs to say so. The test is the same either way: what are the steps to build this, and what will it take to keep it running? An idea without that is a conversation, not a plan.
Apple is unusually disciplined about what it declines to make. That is the underrated half of the job. My own version is a reliable instinct for what people will not want and a habit of steering away early. The useful part is what happens when the instinct stops answering quickly. When it becomes hard to say whether people would want a thing, that is often the signal there is something there.
How I try to operate around other people: in a meeting, on a call, at home, and in whatever the work happens to be that week.
Vague status is expensive. It reads as kindness and costs somebody else a week. An uncomfortable number delivered early beats a comfortable one late. Every project has a real shape: what is done, what is next, what the timeline honestly supports. Holding that picture is most of the job. Keeping it current once it stops being flattering is the rest. Send the number, not a meeting about how to frame the number.
I can talk with almost anyone. An engineer, a lawyer, an investor, a customer having a bad day, someone whose work has nothing to do with software. That is not a social nicety, it is what lets a team trust you quickly. Most of what looks like leadership is being someone people will tell the truth to.
Support queues, compliance filings, payroll runs. Assembling the office furniture when we moved in, and taking the same office apart at the end. Someone has to carry the parts nobody puts on a slide. I would rather be the person doing it than the person who notices afterward that it got done.
What I think a team runs on. Most of it traces back to one place: people are people before they are roles.
Culture starts with the human condition. People have far more in common than not and are still radically different day to day, so this begins with enough empathy to imagine what someone is carrying. Teams go wrong standing on opposite sides of a fence, arguing to settle who was right. That breeds resentment. The alternative comes from sport: same side, looking at the goal rather than at each other, treating it as a third thing you march toward together. Perspectives sharpen each other instead of competing. The Patriots, the Red Wings, the Penguins of those eras all read each other's gaps, filled them, and took the accolade as a team.
Trust goes when words and actions drift apart. People talk far more than they act and promise more than they deliver, usually without meaning harm by it. Someone who commits to one thing and does it builds more trust than someone who commits to three and delivers one. Say less and land it. That gap, repeated quietly over months, is what hollows out an organization.
One difficult person sends ripples through an organization, and difficult takes many forms: a steadily negative outlook, no faith in where the thing is going, personal chaos carried into the work. Almost nobody sets out to be that, and most cannot see it in themselves. What fixes it is feedback that is expected, welcomed and early, rather than held close until it is a grievance. Withheld feedback becomes resentment, then things left unsaid, then work that suffers for reasons nobody will name.
The idea that people run at peak eighteen hours a day, seven days a week is nonsense. Pretending otherwise just moves the cost somewhere less visible: fatigue arrives, health slips, and worry about health or family takes the room the work used to have. Someone holding one clear objective out-performs someone splitting the same head across six, because attention, judgment and care divide with it. That is arithmetic, not philosophy. Give someone one clear objective, real freedom to fail, and not much else in the way.
Work becomes fun when a team goes through something genuinely hard together, comes out the other side, and can say it was worth it. The rest is liking the people you are with, and enough humility to accept that everyone fails. What matters is what gets done with the failure: turned over, learned from, carried into the next attempt. That is what makes trial and error feel survivable rather than dangerous.
Where my attention goes when it is not on a deliverable, and the problems I would happily spend the next several years inside.
The other side of the table. I have raised across eight rounds and lived with the diligence, the reporting and the cap table afterwards. What does an operator notice that a purely financial read misses? And why does so little software sit between a firm and its portfolio? Reporting, audits, updates: most of it still runs on spreadsheets and email.
The administrative layer of a business is still people doing rote work across ten disconnected tools: accounting, tax, vendor management, state registrations, compliance filings. I ran all of it by hand for six years. That gives me a concrete sense of which parts could be automated end to end and which genuinely need a person. Most of it is the first kind.
As more of what people read, watch and hear is generated, the premium on whatever demonstrably is not goes up. At consumer scale my guess is the market quietly rejects a good deal of what is being built, in-home robots and face-worn devices among them. Not because the technology fails, but because people do not want it in the room.
A product surface is right for one person and wrong for the next. The same capability through a CLI, an API or an MCP server can be a better product than any screen. Design is a multimodal problem now, much less a question of making one interface look good. What a coherent product feels like across surfaces that share no pixels is the part I find genuinely open.
AI-native software. The distance between the frontier and everyone else. Where model value settles. What makes an organization hold. Purpose-built software for niche markets. Government technology. What AI takes while it is helping. Sports technology. Ask about any of them.