"smart" forms (for onboarding, wizards, etc.)

Sunil Pai has built an experiment using an LLM and web search to auto-infer form values during onboarding for the app Trellis, reducing data entry time with just a name and company. The approach targets defined fields and templates, with potential applications for government agencies and educational institutions, though some suggest a user-side implementation could leverage more information.

new experiment: "smart" forms for onboarding, wizards, etc LLMs aren't just about chat based conversations; here I'm using an LLM/web search to infer form values while onboarding on to new app/service called Trellis. With just a name and company, it can guess some fields and values that can dramatically reduce the time to enter and make it a little more delightful? shoutout to sunil pai on X: "new experiment: "smart" forms for onboarding, wizards, etc LLMs aren't just about chat based conversations; here I'm using an LLM/web search to infer form values while onboarding on to new app/service called Trellis. With just a name and company, it can guess some fields and" - let me guess you built that with gpt-5.6 - Super nice use case , but this would work well only for cases where there are defined fields/templates e.g as described onboarding Would be interesting for govt. agencies & educational institutions to try this out first - this made me think of browsers filling fields like credit card or billing address automatically for you seems like implementing the llm user-side will have much more information than trying to guess from the app side? obviously a good way to start by doing it this way