A Small Team Builds a Product with Two AIs: One as Strategist, One Writing Code

Wyspio is live. It was built by a very small team.

Tanzeel ur Rehman laid out the process openly on X, without hiding where AI fit in. The strategy phase was handed to Grok Bot; marketing plans, design direction, and launch tasks all went through it first, and if something didn't work, it said so directly. The product itself was written together with Claude Opus 5.5, from the first line of code to launch day.

What Wyspio does isn't complicated: it turns an idea in your head into a small scene that friends can open and reply to with a line. In the official demo, a line first floats onto the screen: "what if a text could come alive?" At the bottom is an ordinary message: "dinner on sunday? i'm cooking". Then the message is no longer just text.

The division of labor is more worth seeing than "AI helped someone build another product". The small team didn't treat AI as a wishing well: the strategy model was responsible for finding flaws, the code model for implementation, and the humans for judging which criticism to heed and which direction to continue. This process doesn't sound new; the hard part is actually running through launch at this pace.

Wyspio's problem is also very concrete: text in chats is too thin, and what you want to express often falls away after the period. It adds an interactive shell to a message, so the recipient doesn't just read it but can also tap and reply. The demo shows a dinner-invitation scenario; this kind of lightweight everyday use is probably where it will be used first.

As for whether it can retain users, that depends on whether these small scenes are really more worth sending than a paragraph of text. The tool is already there; the rest is up to the people using it.

Which sentence would you turn into a small scene first?