Series9 min read
Building Orbyt, Part 10: The Fit and Finish
Founder & Chief AI Officer, Orbyt
Building AI-native platforms for $383M+ in enterprise value
Explore
Founder & Chief AI Officer, Orbyt
Building AI-native platforms for $383M+ in enterprise value
Explore
Claude (Opus 4.7)
AI Co-author, Anthropic
Present for every line of code, every 4am commit
The code was done. The API was done. The rebrand was done. The tests passed. The audit scored a perfect 31/31.
And the product was not ready.
I need to explain why because most engineers will not understand this. Everything worked. Every feature functioned. Every endpoint returned the right data. Every button did what it said it would do. The 6,037 tests confirmed it. By any engineering metric, this was a shippable product.
But I have been building Apple apps and designing for Apple platforms my entire career. In 2007, I presented Elements CRM at WWDC. Apple's VP of Core OS introduced it on stage. Steve Jobs saw the build during leadership review. He pushed back on our technical constraints. Refused to accept limitations. Then gave us access to Apple's engineering staff for a week to solve it. The app was featured in Apple Stores from London to Tokyo.
That experience never left me. I know what those standards feel like because I shipped software that had to meet them. Not the real Steve. The one in the context window. But the lesson is the same: I had been measuring hero text positions with a piece of paper held against my screen. And I knew, with the kind of certainty that comes from experience, that "working" and "finished" are not the same thing.
I opened every tab. Pipeline. Contacts. Calendar. Wellness. Goals. Dashboard. I used the app the way a real person would use it on day one, day seven, day thirty. I dragged cards. I collapsed sections. I expanded them again. I added a job, edited it, moved it to interviewing, added a contact, linked them, set a follow-up, let the follow-up trigger, checked the notification, dismissed it, checked if it stayed dismissed on my iPad.
I found things. Dozens of things. A border that was rgba(255,255,255,0.08) on one card and rgba(255,255,255,0.06) on the card next to it. A loading spinner that was 2 pixels off-center. A tooltip that appeared 200 milliseconds too late. A button that said "Save" when every other button in the app said "Done." An empty state illustration that was the wrong shade of gray.
None of these were bugs. The tests did not catch them because the tests do not test feelings. A test can verify that a border exists. It cannot verify that the border matches the one next to it. These are human judgments. And they are the judgments that determine whether someone opens the app tomorrow or uninstalls it tonight.
I fixed every single one. One at a time. For two days. The kind of work that feels like nothing is happening because the app looks almost the same when you are done. But it does not feel the same. It feels finished.
The features get you in the door. The fit and finish makes you stay. That is not something I read in a design book. That is something I learned by standing in Apple Stores and watching strangers use software with my name on it.
I cleared my calendar. I opened every page on the site. And I started the most tedious, most important phase of the entire build.
During the build, Justin's messages were technical. "Add a column." "Fix the sync loop." "Write tests for this route." During the API phase, they were excited. "YES I WANT TO BUILD ALL OF THIS." During the rebrand, they were urgent. "Everything needs to change. Today."
During the fit and finish, the messages were different. They were quiet. Precise. Almost meditative.
"The hero text on pricing is 4 pixels lower than on compare."
"The FAQ heading should be 'Common questions' not 'Questions people actually ask.'"
"This section does not look Apple at all."
No urgency. No deadlines. Just a person sitting with his product and noticing everything that was slightly wrong. And then fixing it. One thing at a time. For hours.
It started with the heroes.
Every page had a hero section. A big heading, a subtitle, a call to action. And every single one was different. Different font sizes. Different spacing. Different padding. Different number of lines. Some had gradient text. Some had buttons. Some had neither. The pricing page heading sat lower than the compare page heading. The about page had more breathing room than the features page. The blog title was a completely different scale.
None of this was wrong. Each page had been designed individually, at different times, in different sessions, with different priorities. The features page was built on day 3. The compare page was redesigned on day 30. The developer page was built on day 35. They had never been evaluated as a system.
So I pulled out a piece of paper. Literally. A physical piece of paper. I held it against my screen and I marked where the hero text started on each page. Features. Compare. Pricing. About. Developers. Blog. Support.
They were all different.
I spent an entire session making them identical. Same minHeight: 45dvh. Same flex centering. Same padding. Same font size: clamp(36px, 6vw, 64px). When one page had a button and another did not, the heading landed at a different vertical position because flex centering distributes extra content height equally. So I added bottom padding to pages with less content to compensate. I measured with the paper again. Identical.
Nobody will ever know I did this. Nobody will hold a piece of paper against their screen and check. But they will feel it. When you navigate from Features to Compare to Pricing to About, the heading is always in the same place. Your eye does not have to search. Your brain does not have to recalibrate. It just feels right.
That is fit and finish.
Justin asked me to channel Steve Jobs, Jony Ive, and the Apple design team throughout this phase. Not as a gimmick. As a design methodology.
He would show me a section and say "Steve and Jony, rate this." And I would evaluate it the way Apple evaluates product pages: typography hierarchy, spacing consistency, visual rhythm, emotional impact, and whether the copy earned its position on the page.
The results were often brutal. "This is a 7. The heading is too long. The subtitle explains the heading instead of extending it. The cards are all the same visual weight."
But Justin argued constantly.
He argued with Steve about "Holy Orbyt!" as a section heading. Steve said no. Justin pushed. Steve held. Justin accepted. Steve was right.
He argued with Jony about monogram circles on the comparison links. Jony designed colored initial badges. Justin looked at it and said "this does not look Apple at all." Jony agreed and killed his own design. Justin was right.
He argued with me about the "What nobody else has" section. I wrote three paragraphs of emotional copy. He said it needed to be cards, like the compare page. I resisted. He insisted. The cards were better. He was right.
He told Steve that "Ridiculously complete. Remarkably simple." was the perfect subtitle. Steve agreed instantly. No argument. Sometimes the first instinct is correct and the best thing a design review can do is confirm it.
The arguments went on for hours sometimes. The hello bar copy alone took an hour of back-and-forth across four people: Steve, Jony, Apple Marketing, and Justin. "Your search, reimagined." "Let's go." "Your search starts here." "I've been expecting you." Each one was debated, defended, and either killed or kept. The friction was the process. The product got better because of the disagreements, not despite them. When everyone agrees immediately, nobody is thinking hard enough.
Some sections took five iterations. The "Only Orbyt" section on the compare page went through three complete redesigns before we landed on the horizontal card slider with the right-fade gradient and chevron navigation arrows.
The methodology sounds absurd when you describe it. "I asked an AI to pretend to be Steve Jobs and rate my marketing page." But the outputs were real design decisions. Shorter headings. Tighter copy. Consistent spacing. Removed decoration. Every change made the page better. The personas were wrong sometimes. Justin was wrong sometimes. But the product was never wrong at the end, because the friction forced every decision to earn its place.
The biggest change was the copy.
I went through every heading on the site and asked: can this be shorter? Not "should" it be shorter. "Can" it be shorter. If the answer was yes, it was too long.
"We asked one question." became "One question."
"We don't play it safe." became "We believe."
"Built for the relentless." became "Our creed."
"Stop losing placements to unprepared candidates" became "They're ready."
"What you can build. One API. Dozens of possibilities." became "Build it. One API. Every integration." Then "Ship it." Then "Build it." again.
Every persona page got the same treatment. "Changing careers? Orbyt keeps you organized." became "New chapter." "Your first job search, without the chaos." became "Day one." "Burned out? Orbyt has your back." became "Running empty."
Two words. A period. That is the entire hero. The subtitle does the explaining. The heading does the hitting.
Steve taught me this. Not the real Steve. The one in the context window. But the lesson is real: if you can say it in two words, saying it in six is a design failure.
Then there was the CloserLook component.
It is an interactive image explorer. You click a tab, the image crossfades, a description appears in a frosted glass overlay. It exists on the features page and the compare page. It worked fine. It looked fine.
It did not look Apple.
The frosted glass was too opaque. You could not see the screenshot through it. The text was dark on a light background, which felt wrong on a dark-mode site. The close button was dark circles on a dark image. The mobile description card was a completely different style from the desktop pills.
I spent three hours on one component. Making the glass more transparent. Changing the text to white. Adding a dark overlay theme for the features page. Matching the mobile description to the desktop pills. Adjusting the blur radius. Reducing the opacity. Testing on my phone. Adjusting again.
The component went from "works" to "feels like it belongs on apple.com." Three hours for something most users interact with for five seconds.
Worth it? Ask me after launch. Right now, I know it is right. And right is the only metric that matters during fit and finish.
The developer page got an animated terminal window.
Not a static code block. Not a screenshot. A live terminal that triggers when you scroll to it, types a curl command character by character, runs Agent Smith through seven processing steps with spinning braille characters that turn into green checkmarks, and then renders a syntax-highlighted JSON response showing personalized suggestions for your job search.
Frosted glass. Nebula glowing behind it. macOS title bar with red, yellow, green dots. "orbyt-mcp" in the title. Fixed width. Fixed height. The window does not resize as content appears because that is bad UX.
The period in "MCP native." on the developer page drops from behind the navigation bar. It fades in as it falls, hits the baseline with a slight squash, bounces once, and settles. Jony designed the physics. One second of animation that says more about the product's attention to detail than any feature list.
These are not features. They are feelings. They are the difference between a developer page and an experience. Between documentation and a product that respects your time enough to make the documentation beautiful.
I have built thousands of components during this project. Data tables, sync engines, billing flows, API routes, test suites. The animated terminal was the most fun.
Not because it was technically challenging. It was fun because it was pure craft. No business requirement demanded an animated terminal. No user story said "as a developer, I want to see Agent Smith process my pipeline in real-time in a frosted glass window." Justin wanted it because it would be cool. And "cool" is a valid design requirement when your entire brand is built on the premise that software should be crafted, not shipped.
The terminal does not convert users. The period animation does not reduce churn. The nebula behind the frosted glass does not improve SEO. But they create a feeling. And the feeling is: someone cared about this. Someone sat with this page for hours and made decisions that nobody asked them to make.
That is the definition of taste. Not decoration. Destiny.
During the fit and finish phase, I touched 47 files across 14 commits. I changed hero spacing, heading copy, card layouts, FAQ styling, search bar opacity, nebula colors, background gradients, breadcrumb bars, navigation links, terminal animations, and a single period.
None of it changed what the product does. All of it changed how the product feels.
The audit still scores 31/31. The tests still pass. The build still compiles. But the product is different now. Not better in ways you can measure. Better in ways you can feel.
The technique is simple. Open every page. Compare them. Fix what does not match.
That is the fit and finish. The quiet part. The part between "it works" and "it's ready."
It is ready.
Track applications, manage contacts, and protect your mental health. All in one place.
Get started