How No Context Bot works
This page explains, step by step, what happens between someone posting a line in a connected Slack channel and a painting showing up in the thread. It describes the product as it runs today.
1. Cheap checks first
The bot only reads channels it has been added to. Before any AI model sees a message, simple rules skip the obvious non-quotes: messages under three words, messages over 280 characters and messages made only of emoji. A message written as a Slack blockquote (starting with >) skips these checks, because quoting is an explicit signal.
2. Is it a quote?
A language model from Anthropic or OpenAI, depending on your plan, reads the message and answers in a fixed format: whether it is an out-of-context quote, how confident it is, the quote text and who it is attributed to. Messages it is not confident about are left alone, so normal conversation in the channel is never painted.
3. Picking a style
Each channel has a set of enabled art styles. The same model picks the style from that set that best fits the tone of the quote. You can browse the styles on the examples page.
4. Writing a workplace-safe prompt
The quote and style become an image prompt. The prompt instructions require the scene to be suitable for a workplace: no violence, racism, sexism, stereotyping, explicit content, nudity or gore. The image providers apply their own safety filters on top of this.
5. Rendering and posting
An image model from OpenAI or Google renders the painting. The image is stored privately and posted back to the original Slack thread, where only members of your workspace can see it.
What we do not do
- We do not use your quotes or images to train AI models.
- We do not publish your quotes or images.
- We do not read channels the bot has not been added to.
The data side of this is covered in the security overview, the privacy policy and the list of sub-processors. How the guides on this site are written is in the editorial policy.