Is the quiz meant for the reviewer or the author? The quiz is the part I find genuinely new here. My one hesitation: since the questions come from the same model that wrote the summary, passing might just prove I absorbed the AI's reading of the diff, not that the reading was correct or that I understood the actual code. Curious whether the questions are grounded in the diff itself. Only tried the example on the landing page.
What do you do about portion size, and the calories a photo can't see like oil or dressing? Those are the parts that have always made me distrust the number these apps spit out. Curious how you're handling it. Only looked at the landing page so far.
Does anything control the order the paths get drawn in? That feels like the thing that would sell it as a hand sketch, versus strokes appearing all over the canvas at once. Might already be handled, I only played with the demo briefly.
Former bookseller angle is a nice touch. My one worry as a reader: at ~500 words a day a normal novel takes half a year. Can I dial the pace up, or send myself the next drip early when I'm hooked and don't want to wait until tomorrow?
Visual SQL builders live or die on the import direction — pasting an existing messy query and seeing it laid out is way more valuable than building from scratch. How robust is the SQL → canvas parsing? That's the feature I'd actually come back for.
Nice, local is the right call. What's the local AI model — a small NER model bundled in, or calling out to something? Curious about the size/footprint for a desktop app.