Blog

B Engineering 15 Oct 2026 · 1 min read Big Canvases, Smaller Bills Building a game or a long document on a canvas used to get slower and more expensive with every turn. We rebuilt how the assistant works with large canvases. What we changed - Piece by piece. The assistant reads and patc… T Engineering 24 Sep 2026 · 3 min read Three Pages a Second On the morning of 22 September, torun.ai slowed to a crawl. A farm of automated browsers, more than two thousand addresses in three quarters of an hour, was requesting our public pages one at a time, and at the worst min… T Engineering 14 Sep 2026 · 6 min read The Spend Page Read High, and No Money Moved Start with the part that is easiest to lose while reading the rest: no money was taken. No wallet was debited twice, no card was charged, no subscription allowance was spent on work that did not happen. What broke was a… E Engineering 12 Sep 2026 · 4 min read Every Reference the Model Takes, Over the Wire A video model will take more than a prompt. Some take a first frame and a last frame. Some take a handful of reference images. Some take a reference clip, and then behave differently because of it — muting their own soun… W Engineering 7 Sep 2026 · 3 min read When a Picture Has Words In It Someone attached a picture — a card with Russian text on it — and asked, in Russian, for the same card in Turkish. What came back was two images, both billed, on a different background, with lettering that was not words… L Engineering 5 Sep 2026 · 3 min read Links That Show What They Are Paste a ToRun link into a group chat and, until this week, you got the platform's default card: same title, same image, whether the link was a poem, a spreadsheet or a video of your dog. The cause is mundane. The app is… W Engineering 16 Aug 2026 · 2 min read What Auto Is Actually Choosing A fresh "hello" came back badged with a model that has no function calling at all. Nothing looked broken — it answers well enough — but that turn, and every turn behind it, ran with no web search, no picture or video gen… W Engineering 14 Aug 2026 · 2 min read When a Model Has a Bad Minute Every AI provider has bad minutes. A model gets overloaded, a region hiccups, an endpoint answers with an error for ninety seconds. There is nothing unusual about that. What matters is what the app does during those nine… K Engineering 8 Jul 2026 · 2 min read Keeping the Model Catalog Honest A multi-provider platform lives or dies by its catalog. If the price shown is stale or a retired model is still offered, every promise about transparency collapses. Three rules keep ours honest. 1. Prices are never edite…