There is always a line somewhere, running across any project like a fault in bedrock, where you cross from ambitious architecture into things nobody can actually use. You do not notice the line until you stand on the other side and look back. Capacitor got installed, iOS scaffolded, Android nested inside the same repo which felt clever enough to convince me that multi-platform meant something real. 35 commits stacked before noon, progress log reading confident enough to lie to itself. The simulator environment was acknowledged as a blocker and dismissed with the sort of optimism only someone who has not yet met their own assumptions can manage. It would be sorted soon. It was not.
Then came the cookie thing. Which is always the moment your architecture meets reality and loses. Deployed to Vercel, logged in on my phone, discovered Supabase session cookies had grown too large for browser storage to hold. Split them into chunks. Added console logging just to watch the failure mode one more time, as if observation might change the outcome. 27 tests still passing. Build green. Everything functional, everything solid, sitting on top of an architecture the people I was building for could never reach. And that was the truth I kept pushing against. Almost every person accessing this internet lives on mobile Chrome, probably on a family phone somebody else bought, probably in Bandung or Surabaya or somewhere entirely outside my head. All of them. Almost all of them.
The web-first decision landed around 7pm. 41 commits. Ten hours of convincing myself that engineering greatness lives in the cleverness of the architecture rather than the simplicity of what people can actually use. Admission filed under feat(capacitor) in the morning and pivot by evening. Branch kept. Nothing deleted. PWA manifest added. Because giving up on an idea and keeping a door open are not quite the same thing, though the distinction helps at the time.
And then, quietly, without any of the drama the rest of the day demanded, the first lesson prototype shipped. Inflation 101. Animated CSS and SVG visual showing mie ayam price rising from Rp 12k to Rp 16k over time. Bite-sized concept. Indonesian example. Inline quiz. Source trust card citing Bank Indonesia. This was the thing you could point at and say "this is what we are building," which shifts the entire energy of a project, even when you are 3 hours into the next day and genuinely questioning whether everything is just a very elaborate procrastination technique masquerading as ambition.
Some days you convince yourself you are an engineer because the architecture is clever. Other days you realise the only architecture that matters is the one people can actually reach. Building something real is harder than building something ambitious.
Which is almost the whole point.