How the new site is being built

Before Andy sat down, he read the site first

The meeting was booked, and Isaac spent the hour before it going through davidlocksmith.ca page by page — around twenty-two of them. The reviews. The 20,000-Calgarians page. Since 1982, over and over. The discount for anyone named Larry or Peggy. His own verdict, before anyone else was in the room: it looks “not good… really bare bones and low quality and dated.”

That reading is why everything after it is specific. Nothing in this project was ever designed for a locksmith in general.

What the meeting was for

Isaac asked Andy to sit down and talk about what his ideal website would look like — not to pitch him anything, but to get the dream out of his head and build a plan around it. Isaac put ideas in too. Almost everything that happened afterwards traces back to this conversation.

What the two of them covered, in the order it came up:

• “No flashy, hokey, gimmicky.”Said before the meeting even properly started, and held all the way through. Every rule about taste that came later is a refinement of this one.
• “Extremely unique, every aspect.”Nothing generic. The test that came out of it: if a thing on the page could belong to any locksmith, it does not belong on this one.
• The site should be a helping page, not an advertising page.“I want that sales pitch element on our website, and automated.” Solve the customer’s problem on arrival rather than describe the company.
• The intake ticket, in detail.Year, make and model for a car. Photos, with a template showing the customer how to shoot the keyway, because “we do rely on pictures.” Specialties surfaced per audience, so a property manager sees evictions, rekeys and repairs. Proof of recent work placed at the confirmation, not cluttering the front. This became the single most-built thing in the whole project.
• Why a filled-in ticket matters.Andy pointed out it works the way a submitted rental application works: once somebody has done the work of filling it in, they are committed. That one observation is why the intake is the centrepiece and not a contact form.
• Texting is the sales channel.Seven years of it. “That’s how you do the sale.” And: “I would love it if the site could actually message us directly, and when they get that text back, it’d be like, Isaac is writing back — and Isaac is not a chatbot. We can even say that.”
• The uniqueness comes from real objects, and they listed them.David’s hand-built key-cutting machine. The 1992 Yellow Pages ad. The red badge. Real worn keys, dial faces, and their working hands. That is a photograph list, written three weeks before anyone asked for one — and four designers who never saw this meeting arrived back at the machine on their own in August.
• The best thing on the site is the thing nobody can find.The service pages were indexed; the safe content was not. That is the business argument for leaving Google Sites, and it is Andy’s own material being buried. Re-checked August 11: ten pages of davidlocksmith.ca come back in a search and not one of them is a safe page.
• Selling it on, one day.Branching the intake to locksmiths in other cities — the two of them built that idea out together across the conversation, including what it would be called and how vetting would work. Not being built now. It is on the page because it is the reason the intake is worth building properly rather than quickly.
• Guardrails, in the same breath.Usability wins for the panicked customer. The phone number visible on every screen and nobody ever trapped. And price-range honesty as a trust move.

Five steps, in this order

1. The ticket and intake interfaceThe centrepiece and the value.
2. A clean homepage with two pathsFix my problem, or let me look around.
3. An easy way for Andy to change things, earlyNot bolted on at the end.
4. Migrate all the existing contentNothing lost, organised into about five sections instead of seventeen flat ones, and finally findable by search engines.
5. Then the content that earns search trafficReal job photos, proof of work, video, the blogs, and an About section with Isaac, Andy and David, real faces.

Everything built since has been step one and step two.

Three things, in this order

A mockup, built before Andy arrived, out of that morning’s reading. A brief, which is what the two of them made together over coffee. And at the very end, a written plan — the site assessed, the search problem, Andy’s editing question, the intake tool, the agreed build order, and the decisions still to make.

The mockup is the oldest artifact in this story, and it is here because it failed usefully.

It came out polished and premium and generic — and in doing that it dropped the forty years of heritage completely. That is exactly the trap Andy named an hour later — walked into before he had said a word.

Why nothing happened on the website for three weeks

The meeting was July 15. The first line of work on the website was the evening of August 6. That is 22 days, and it was not idle time.

• First, research on AI itself — not on websites.Isaac had been working with Claude through a chat window. He went and learned how to use it properly, which is a different tool with a different way of working. None of that research was about web design.
• ⭐ Then Andy bought him a computer and brought it to his place.Andy set the machine up; Isaac installed the tooling on it. It is the machine everything in this presentation was built on, and it is the machine typing this sentence. Not the fastest computer in the world — and it got him started.
• Then about a week building the workshop, not the product.A second brain that holds everything about the business and remembers it between sessions. A button on his phone that talks to it from anywhere. A link into Wheelhouse, his scheduling app. None of that is the website. All of it is why the website could then be built in five days instead of five months.

Then, three weeks after the coffee, the actual work started.

This time, to inventory what has to survive

Three weeks after the first read-through, the site got read again — not to prepare questions this time, but to write down every page, every asset and every line of copy that has to come across to the new one.

The work started with reading rather than drawing. Every page of davidlocksmith.ca was pulled down and read end to end, then looked at again for how it actually appears on a screen. Not to find faults. To find what was already there and working.

Plenty was working. The best writing on the site is Andy’s own, and most of it sits on inside pages where almost nobody lands. This line about Isaac, on the safe openings page, does more than any slogan a designer could invent:

And in the third paragraph of the About page, the machine again, in the shop’s own words:

David Leung was a journeyman machinist before he was a locksmith. He built a key machine out of raw metal and it still cuts keys. Andy has it. David Reid has used it. It was the first key machine Isaac ever used, because Andy lent it to him when he was starting out. Four locksmiths, one machine, forty-four years.

What was actually asked, and what came back

Before a single line of the new site was written, twelve research agents were run across eleven passes — each one given a title and a single question, each working on its own and reporting back. These were the questions. This is the part that does not show up in a finished website, so it is worth seeing.

• What can a website even be in 2026?Asked what the most awarded sites on the web are actually doing. Answer: the frontier is browser-native games and films — and it is reachable, but it is not where a locksmith site should start.
• What should this be built with?The answer was: keep doing what Isaac already does. Hand-built single-file HTML with Claude beat every design tool examined. Nothing needed buying.
• What does a panicking, locked-out person actually need?Call first, not form first. The two best-funded emergency-trade sites in the world both put a tap-to-call above everything else. A form is for the person who can wait.
• Should the safe dial be the navigation?No — it should be the signature, not the steering wheel. Every successful instrument-as-interface site does exactly that, and the ones that make you drive with it fail.
• What does the anti-generic look actually look like?The AI-generic look of 2026 is now a documented, named cluster: purple gradients, the same headline font, three rounded cards with thin-line icons. Naming it is what lets you avoid it on purpose.
• What moves the page?GSAP for the two things that matter, plain CSS for everything ambient. No heavy library where a light one does.
• Can it use real 3D on a phone?Not in version one. Measured against real mobile budgets, SVG plus one pre-rendered image sequence gets the same effect without the cost.
• Can pages transition like machinery?Yes, and free. Modern browsers do it natively, and older ones simply do not see it.
• How does Andy change things himself?A free editing layer where Andy never needs a developer account. One fetched fact settled it.
• What happens to a submitted ticket?The whole pipeline is real, cheap and buildable on the free tier — form, spam check, stored ticket, email out. Under three dollars a month.
• Funnel or brochure, for search?Both, split by what the person typed. Emergency searches want the funnel; the safe work wants the brochure. And the Google Business Profile, with its 4.9 stars, is the real source of leads.

Eleven passes, twelve questions — the first was run twice. The twelfth covered sound and accessibility, and settled it in four words: sound is opt-in or nothing.

The same brief, four times over, as an experiment

Rather than commission one design, four were run at once. Four separate design agents, given the identical brief and evidence, none of them able to see the others. The question was whether independent designers would land in the same place.

Three of the four did, separately: the shop has two rooms — a counter where customers are served and a bench where the work happens — and the site should have both. The fourth broke away entirely and turned the site into the shop’s carbon-copy work order.

The same design, built three different ways

One design was picked and then built three times over, by three different builders, to see how much the execution changes the thing.

Isaac’s feedback on all three was that they read generic — the drawings looked like drawings, and the shop was not in them. That single note is what sent everything afterwards toward real objects and real photographs.

One pin, made of light

Not a website. A single pin stack from inside a lock, in real 3D: a brass key pin, the steel driver above it, the spring, the chamber wall, and the shear line drawn as a horizon that flares at the moment the pin sets. It runs on a nine-second loop and you can drag it to look around it.

Four concepts, and none of them the answer alone

Four more agents, four different ideas of what the whole site could be — written up in detail rather than built, so they could be compared cheaply before anything expensive got made.

• The scroll journeyYou travel from the street into the lock as you scroll.
• The two doorsLife only at the thresholds — structurally unable to become gimmicky.
• The funnel heroThe intake itself is the front page. This one won.
• The bench as galleryA craft museum with rooms, the counter kept simple.

The verdict was to take the funnel out of the third one and build that first, because it is the part that earns money — which is also step one of the order agreed in July.

Then nothing for three days. That Friday afternoon was the last of it before Shabbat, and the weekend went to other work. The website picks up again on Monday evening.

Two things shipped in seven minutes

Isaac stops it

He looked at both and said they were something put together out of panic — built fast because the day was running out, not because they were right. He stopped the build there.

That stop is why the next thing that happened was research and not another demo.

What Isaac actually wants it to feel like

Isaac sent sixteen videos of websites he admires. Every one was watched, written up, and judged against one question: what would this be if it were built for a locksmith?

They are linked below so you can watch them yourself. This is the imagery that is driving the whole look of the site — and if something in here sparks an idea, that is worth more than anything written about them.

• Watch ›An AI insurance site where one morphing shape carries you the whole way through.
• Watch ›The $10,000 website elements — what expensive actually buys.
• Watch ›“Sorry bro, Claude is not making this” — the ceiling, honestly stated.
• Watch ›A burger restaurant. The closest thing to home on the whole list — a real small business, made to look like it matters.
• Watch ›A personal build, start to finish.
• Watch ›A studio site that alternates light and dark rooms, and treats type as a material.
• Watch ›The tools list — what the people making these actually use.
• Watch ›A renaissance-themed commerce site, where one art direction carries everything.
• Watch ›Gravity as the single metaphor, used once and used properly.
• Watch ›A wireframe-industrial look.
• Watch ›Scene choreography — the camera moving through a built space.
• Watch ›Texture and material study.
• Watch ›Type and motion together.
• Watch ›Hand-crafted web feel, one of the two Isaac later made required viewing.
• Watch ›The second of his two required ones.
• Watch ›The sources list — the studios worth studying.

The rule that came out of all sixteen: the ones that would embarrass this shop are the ones where a technique keeps its original subject and the shop just gets dressed in it. Different words, the same rule Andy set in July.

Four more passes, on what any of it costs

Four further agents went and priced the techniques in those videos. The most impressive sites are not doing the expensive thing. Two of the three showpieces run no heavy 3D in the browser at all — they are pre-made and played back. Which means the look is reachable on a free plan, and that is the finding the whole design rests on.

Five designs, and something to measure them against

A control design was written and sealed before any of the designers started, so there would be something honest to compare the field to. Then five designers ran one at a time, each blind to the others.

The blindness got tightened partway through: after one designer admitted that catching sight of a rival’s file name had changed its design, the rest were forbidden from even listing the folder.

Isaac’s verdict on all five: every one of them refused spectacle on the first screen — which was right — but the real problem was density. Too much on screen at once, and no single obvious thing to do.

The brief rewritten almost entirely out of Isaac’s own words

• No invented conceits.The deepest correction of the night. A metaphor worn like a costume gets thrown out; the idea has to come from the trade rather than be dressed onto it. This is Andy’s July rule arrived at again, from the other end.
• The logo is untouchable.The badge is Andy’s father’s. Nobody redraws it, recolours it or modernises it.
• The site is about the company and about Andy’s father’s legacy.Isaac said this himself, unprompted, and it changed several designs. He is part of the story and never the subject of it.
• One door, not two.Every version offering a choice of two ways in was cut. A person locked out does not want a menu.
• Mechanical accuracy is not optional.If a lock is drawn, it has to work the way that lock actually works. Every locksmith Andy knows would spot a fake instantly.

• Heavy graphics live deeper in.Isaac settled the argument between cinema and speed himself, and it is now the architecture of the whole site: the front door stays fast, the craft lives further in.
• A locksmith’s site, not a hobbyist’s site.Lock imagery is welcome, and ordinary mechanisms fascinate people more than exotic ones do. The proof is in his own numbers: the most-watched thing he has ever posted is a cabinet lever lock, 53,781 views, and it is not a locksport video at all. The comments are almost entirely people locked out of real things — a great-grandmother’s cabinet nobody could open, an antique desk, a display case with the key broken off inside. Serve the locked-out person, and the craft audience comes anyway.

He also designed a section outright, out loud: scrolling toward a building, finding an old safe inside it, getting close enough to watch the handle turn, and then going through the dial itself to see the mechanism from within. That is the safes section, and it is his.

Not on guesswork this time

Four more designers, blind again — but this round they were handed everything: the sixteen references, the three hours of rules, and the density verdict. What follows are architect’s drawings rather than finished pages. They are deliberately cheap to load, and the grey areas are labelled in words with what the real thing puts there.

Isaac’s verdict on the second design-off

So the second design-off is not the answer either. The drawn keys and drawn pins in those four boards are exactly what he does not want on the real site. The next round has to be built from photographs of real locks rather than drawings of them — which is the third time this project has arrived at the same conclusion: the July meeting, a design agent in August, and Isaac on the morning of August 11, all separately.

The story becomes part of the story

Isaac asked for this page so you could see everything at once — not just the versions, but the order, the research, the arguments, and the two times he sent it back to the drawing board.

And then this page went through the same cycle as everything on it. It opened in August and left the July meeting out entirely. It argued that the website must be visual while being 4,767 words with no pictures on it. Every one of those was found by checking it against the records rather than from memory, and fixed before the next read. Both earlier drafts are kept below. Nothing gets thrown away here either.

Then Isaac read it, and it went round again

He opened it properly the next morning and the cycle repeated. Every link on it was dead, and he found that in about a minute. The first mockup was in the wrong place — going back to the record showed it was built before Andy arrived, not during the meeting. And he remembered a second document from that day that had never been shown: the written plan, with the build order in it. It is on this page now.

Eleven other things were fixed the same morning, from the typefaces to the dates to the pauses in the narration. The three drafts kept below are the ones where something visibly changed; the rest were corrections, and they live in the record rather than on this page.

And the voice reading this to you

The narration you can play from any section of this page is Isaac’s own voice, trained on the night of August 11 on recordings of him reading. It is not finished — it has heard about fifteen minutes of him so far, and it still cannot say his own name properly, which is why the narration says Yitzhak. It runs entirely on that computer you brought him, with no internet connection at all.

Four things, in order

• Isaac reviews the four newest versions on camera.He walks through each one out loud and says what is wrong and what is worth keeping, and the recording becomes the brief. Nearly every good rule in this project came out of one of those.
• Another research pass on more reference sites.Same as the sixteen: one at a time, written up, always asking what it would be if it were built for a locksmith.
• Photographs. This is the thing that unblocks everything.A full shot list exists: the key machine, the three of them together, hands working, and a reference library of locks, cylinders, pins and safe dials to build the 3D from.
• Then the next round of designs,built on real photographs instead of drawings — which is the whole difference between this round and the next one.

Where these stand

Five things were left open in July. Isaac has answered all five of them. They are here so you can see what was decided and push back on any of it.

• Who answers a ticket when everyone is busy? — answeredDavid is on the phones; if he is not free it goes to Andy, and then to Isaac. When someone is in the middle of a job they usually free their hands and take the call anyway, and when they genuinely cannot, they call back within about ten minutes — or text the number to whoever is free. So the site does not have to invent a promise. A real person answers, and when they cannot, a real person calls back quickly.
• Does the badge stay? — answered: yes, and it is not a design questionRedesigning the branding would, in Isaac’s words, make it feel like this is not Andy’s thing and he is not part of it. The point is to enhance the business and honour David’s legacy, not to cover it up and replace it.
• Do prices go on the site? — answered: not yet, and not by one personIsaac would consider a general list with a plain note that difficulty varies — but he has said outright it is not his call alone. It needs Andy and David. And the question surfaced something bigger: there is no price list at all, not even a private one. Everyone prices from experience and checks with each other. That is a shop conversation, not a web page.
• Is the move off Google Sites settled? — answered: yes, about 99%Google Sites is extremely limited. Where the new one goes is deliberately still open — that decision is not worth making yet.
• Will Andy be able to edit it himself? — answered: yes He will have access to the code and be able to change things on his own. That does not change, whatever the site is built with.