The eye is closed
The last entry praised the engine we’d finally got right. So it’s a strange thing to report, writing this one tonight, that the first thing we did this month was delete it.
On a single afternoon the whole custom engine came out in one commit — a hundred and fifty files, forty-five thousand lines gone. We’d found a simpler answer: hand the opaque bytes to the browser’s own media pipeline and let it do the decoding we’d spent months fighting it for. The blind-relay rule never moved — the relay still passed your provider’s bytes along without decoding them — but the part we’d written five times was now the browser’s job. The production release that shipped it three days later has a deletion count for a headline.
Then we rebuilt the mobile player on top of it, and wrote a small instrument that measures the interface the way a thumb would. The first thing it did was catch us shipping buttons too small to tap. A good instrument indicts its own authors.
And on the last night, the logo finally caught up to the product. The mark became an amber square with two closed eyes, the same promise the home page makes: we’re built not to look. One of the commits leaves a note for whoever comes next: do not revert it.
The rule hasn’t moved either. We won’t card a channel that is actually fine. We won’t silently downgrade what you’re watching. And we won’t log what you watched, because we can’t — there is no column in any database that names what’s in your stream. That was the whole reason we deleted the first version of twiga, where this journal started. We’ll write the next one when there’s something worth telling.
Back to the beginning: the first version of twiga could see your streams, so we deleted it →
twiga