← All notes · 5 min read
The last day before the lock
A magazine closing, a final shooting day, a launch week. Now a game. It is always the same day.
Tonight I'm going to type one word in capital letters: KİLİTLE. Lock. After that word the store build of Game Show gets made, and whatever is inside it is version one.
I haven't typed it yet. This note is about the thirty hours before it.
The lock was supposed to happen on November 5. Then I added a whole new mode where two friends play against each other, and ads. Five days gone. I don't regret either. But I could feel the date getting soft, and a soft date never arrives.
Last night
Around 1 a.m. we had 110 new picture questions for party mode. Four pictures, pick the right one. The map questions looked bad, so I cut them. We redrew ten as city silhouettes in the game's own style. Still boring. Cut. At 1:24 I parked the whole pack. It was pulling the quality of the questions down, and an evening of work is not a reason to ship something.
Same night, the store screenshots. I tried a style without phone frames. Didn't like it. Went back to the first idea, changed my mind again, and ended on plain clay phones with real screens from the game. Ten posters, done before 2.
Somewhere in there I saw how the game looks on a tablet. Not great. I said leave it, write a note, later.
Today
At 9:17 an audit report was waiting for me. Everything that was wrong before release, a couple dozen items. Holes on the server. A test payment still open on the web. Six "coming soon" packs sitting on the shelf with no questions in them. By 10:34 the server ones were closed. The empty packs I just removed. A "coming soon" label is a promise, and I have enough of those.
At 5:43 in the evening I asked to see our app icon. It was still a placeholder. On the last day. Two AIs agreed the icon should be letter tiles, so people understand it's a word game. I said no. The logo, 3D is fine, make the studio behind it darker. A few minutes later it was in both packages.
What the AI said we couldn't
The AI is careful. On a last day, careful sounds like "later."
I asked whether the game really works on the new folding iPhone. The answer was honest: it runs, it isn't ready for the badge, don't hold the first release for this, put it in the first update. I said we do it now, that badge matters to me. By evening the game was being built with the new tools and the wide screens had their own layout. The tablet problem I had parked at midnight got fixed on the way.
I asked what Apple likes to see in a game. I got a good list and a plan: Game Center, accessibility, a widget, a card on the lock screen. 22 to 26 working days. I asked, why so long, based on what? Then I picked the first and the last and said today. The plan gave those two about ten days. They were in the main code the same afternoon.
I don't think the AI was wrong. It estimates like a careful team with a calendar. It doesn't know how much I'm willing to drop to get one thing in, or that I'll be at the desk until midnight. Only I know that, so I have to say it.
And the caution was not useless. Those two features went in fast, and nobody has seen them on a real phone yet. That comes back at the end of this note.
At some point the AI explained that the release branch was about seventy changes behind the main one, and that things should be carried over piece by piece. I didn't follow all of it and I didn't need to. I'm building one thing. Everything I approved is version one. One package, nothing left behind.
Close to 8 p.m. one session gave me a full status report and I wrote back: you threw too much at me at once, I can't take this in. Don't show me what's solved. Show me what's left. And show me, separately, what I chose to leave out.
That second list is the real work of a last day.
Two store events needed new things inside the game. Good ideas, both. Version two.
By tonight there were more than 200 commits since yesterday noon and about twenty AI sessions open, each on one job. One of them does nothing but watch the others.
At 9:13 I wrote, "I think we're ready to say lock." The answer came back: the build on your phone is from yesterday. Nothing from today has run on a real iPhone yet. And the crash reports arrive but can't be read. So one more test build first. Then the word.
I know this day
I have lived it before, in other jobs.
A magazine on closing day. The pages are what they are. Someone finds a typo, someone wants to swap a photo, and at some hour the files go out. The argument doesn't get solved. It gets printed.
The last shooting day. I spent years on sets keeping the story straight. On the final day there is a list of shots and less light than the list needs. You don't get everything. You get what the story can't live without, and you make sure it cuts together.
The week before a launch. Every hour someone brings one more thing that would be nice. Most of them are right. Almost none of them go in.
Different work, same day. Nothing is ever finished. Someone says stop, and what's on the table becomes the thing. What the day asks for is small, fast decisions. Cut. Park. Later. This one, now. And enough calm to tell a real problem from a good idea.
I used to think of these days as the stressful exception between long stretches of normal work. Now I think they are the parts I actually remember. A life looks a lot like the sum of its closing days.
The test build is on its way. I'm going to go check my phone.
— Kemal