The one-prompt run took 86 minutes, Sael says, and drew on files from his earlier projects rather than a blank workspace.
By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
· Published
Primary source: [X](https://x.com/RyanSael/status/2102591147927654847)
Why it matters #
Sael's demo shows how a coding model can turn a specific teaching brief into a usable interactive page. His prior project files and maximum-effort setting are essential context for the one-prompt, $25.66 claim.
Ryan Sael (@RyanSael) published an interactive camera-lens lab on September 23rd after asking Anthropic's newly released Opus 5.5 to build an explanation of focus. In his post on X, Sael said the model completed the project in one run lasting one hour and 26 minutes, at an API cost of $25.66. Those time and cost figures are Sael's account of this run, not a measured cost for reproducing the project.
https://x.com/RyanSael/status/2102591147927654847 Sael brought relevant experience to the prompt. His portfolio lists earlier work building VR and AR experiences at FXMedia Singapore, leading interface design at Bridestory, and designing charity software at N3O. He has also made interactive web projects under his SAEL Lab name. That background matters here because Sael said he asked Opus 5.5 to read files from his previous projects and set it to maximum effort. His description of a "one-shot" build refers to the prompt-and-run workflow; it does not mean the model started without examples or direction.
A camera lesson you can manipulate
The published page, titled "The Plane of Focus," lets visitors turn a focus ring and see the represented lens elements and sharp plane move through a scene. Visitors can select foreground, middle or background subjects, change the aperture among f/2, f/5.6 and f/16, and switch between assembled and exploded views of the lens. Explanatory text connects those controls to depth of field and the circle of confusion, the blur produced when light from an out-of-focus point reaches the image plane as a disc. The page specifies that its numbers use a 50 mm lens on a full-frame camera.
That combination is what makes the demo useful to examine. It joins an explanation, a visual model and controls that let someone test the explanation by changing a setting. A static answer could describe why a wider aperture produces a shallower sharp zone; Sael's page gives the visitor an aperture control and depicts the resulting change. The published artifact establishes what the page presents. It does not, on its own, independently validate every optical calculation behind the visualization.
Sael's account also narrows what the price tag proves. He told another X user that his prompt included instructions to scan files from previous, sometimes unpublished projects, and told a commenter that he used maximum effort. Reusing that context may have helped the model match the kind of interaction and finish he wanted. The $25.66 figure is therefore a reported bill for Sael's particular run, with his project materials and settings, rather than a quote for anyone asking a model to make the same lesson.
A timely test of a new model
Anthropic introduced Opus 5.5 on September 22nd and emphasized coding performance and lower costs for work billed by token. Anthropic lists API prices of $4 per million input tokens and $20 per million output tokens. Those rates provide context for Sael's reported spend, but they cannot verify his bill without the run's token usage. Anthropic's own coding evaluations address a different question from this demo: how a model performs across specified tasks, rather than whether one interactive lesson can be built in a single guided run.
Sael said more than 1,000 people visited the lab within four hours of his post. The traffic claim, like the build time and API cost, comes from Sael. The page itself is available to inspect, and its appeal is concrete: a reader can change focus and aperture instead of taking either Sael's description or the model's prose on faith. For builders weighing AI-assisted development, the demonstration is strongest as an example of a skilled designer supplying prior work and a specific learning goal, then publishing the result for others to use.