Benjamin Code and Autotrim: building solo with AI

Building an application with AI comes with a frustration for me: spending time with Codex and Claude Code, accumulating generated code, and still waiting for a stable MVP or a first sale. Benjamin Code’s account of building Autotrim, in the video linked to this article, felt both intimidating and illuminating.

He describes launching it in 90 days, on his own. What interests me here is how he organised development and found his first users.

Architecture and review of generated code

In this account, Benjamin Code says he did not know Rust and Tauri well at the outset. He describes a vibe coding approach in which AI acts as a co-developer and generates 90% of the code.

That figure describes his experience. For me, the interest lies in the work around the code: defining the architecture, writing instructions and reviewing results. Fast generation still leaves errors to resolve and a product to iterate on.

The landing page and first testers

In the video, he explains that AI generated his multilingual landing page and its SEO without him writing frontend code. That choice lets him focus on the message and value proposition.

For testing, he turns to his network, including people who previously applied to work as video editors. This gives a concrete example of finding people likely to use the software.

Moving beyond iteration

What I take from this account is that the work also includes presenting the product and talking with users. An application can stay in iteration for a long time even when code generation moves quickly.

A first unsolicited sale remains a goal I care about: it shows that someone found a reason to pay for the product. The 90-day timeline described here does not guarantee the same outcome for another project.

What is the next step in your AI-assisted project?