Wordware

Wordware is presented on its current site as an AI context lab in San Francisco rather than as the prompt-authoring IDE this entry was written around: the homepage states that the company started in 2023 on the conviction that words are the next programming language, and that its main product is now Sauna, an assistant built for compounding context that is meant to learn a user's taste, detect patterns and work proactively. The navigation keeps a Wordware V1 entry pointing at an app subdomain alongside Sauna, and Sauna itself is sold as a proactive teammate with its own computer that builds routines and apps, reachable over iMessage, Slack and the web, with a free start and tools listed in the thousands.

It suits people who want an assistant that accumulates context about their work and acts without being asked each time — Sauna's pitch is running parts of a company from messaging apps, with free credits to try it. It suits less well anyone who came to Wordware for the natural-language agent builder this entry used to describe: the homepage now markets Sauna as the main product, so whether the prompt-as-program workflow is still developed, still supported or still sold under the V1 name is not something the pages read here establish, and a team that needs that tool should confirm it before committing.

Not verified: only the homepage and the announcement page were read on 12 October 2026, plus the Sauna landing page; the Wordware V1 application and its documentation were not opened, so what remains of the original product, its pricing and its support status are unknown. The homepage's credentials — a $30M seed round described as the largest in Y Combinator history, led by Spark Capital with Felicis, Y Combinator and others, and a first-place Product Hunt launch — are company statements carried on the site's own announcement page and were not confirmed elsewhere. Sauna's free credit figure and its model, integration and tool counts are the vendor's own, and nothing was run here, so the context-compounding behaviour and the proactive claims were not tested.