INDEPENDENT AI NEWS / AI-assisted. Human-directed.EDITION 006 / 5 November 2026

Software & ownership

The competition would like a screenshot.

Craft Anything studies existing software and uses a coding agent to build another version. We are delighted by the possibilities, particularly until someone points it at ours.

Here’s what happened.

Craft Anything is an open-source desktop app that puts inspection, rebuilding and comparison into one workflow. Choose a website, application or game, describe what you want, and its builder creates a separate project before checking it against the reference. Its v0.3.0 release is dated 2 November. The default Clean room mode uses visible evidence, such as screenshots and observed behaviour; a separate Deep mode can inspect code and files. Building requires a coding agent installed and signed in on your computer, with Claude Code the primary documented route. This is an early alpha: the project distinguishes working website comparisons from partial app and game behaviour checks, and lists integrations that have not been tested live. A convincing recreation of a screen is therefore a useful result to examine, rather than evidence that the original product has been reproduced in full.

Source: Craft Anything project and current status ↗

Our silver robot photographs a founder’s application while a second monitor displays a similar layout.
AI-generated editorial illustration: a metaphor about easier software reproduction, not a Craft Anything screenshot or a demonstrated rebuild.

The We Are So Done take

We have spent years explaining that our product is more than its interface. There are integrations, awkward exceptions, decisions made after customers called at terrible hours. We are absolutely right. It is still uncomfortable to watch the interface become somebody else’s starting material.

Naturally, we want the tool. We have an old application that deserves a better life, a prototype we could finally rebuild properly, and several very sensible reasons to ask the machine to study something that already works. Every one of these reasons is excellent. We assume our competitors have terrible reasons.

The unsettling part is the changing price of a head start. We used to count the months it would take someone to arrive at our first version. Now we may need to explain why customers should stay after somebody else has arrived at something similar. Relationships, reliability and all those boring details suddenly look extremely innovative.

The alpha has plenty left to prove. We can breathe. We can also use the time to make something harder to replace than a screenshot. We would prefer to do that with the new tool, obviously.

Outside this newsroom

The internet has thoughts.

The public issue tracker supplies a useful alpha check. A relayed Windows report says the app did not detect Claude Code; a reply says the connection works on another setup. A separate Linux report describes being blocked by browser detection. These are individual setup reports, not a measured failure rate or tests reproduced by us.

Selected public reactions, checked 2026-11-05. Individual experiences, not a representative survey.

For those of us already clicking

Try a small rebuild before declaring the moat gone.

Access
Public alpha downloads or a local source build, plus a configured coding agent. Agent usage has its own account and costs.
Start here
Use a small site you own or are allowed to examine. Ask for one page in a separate project. Compare layout, navigation, keyboard use and one real task at several window sizes. Keep a list of what the comparison cannot establish.
The small print
No rebuild has been tested by this newsroom. The project marks several integrations and behaviour checks as partial. Its Clean room setting controls the evidence it uses; it does not establish rights to reuse someone else’s material or prove functional equivalence.

Follow the evidence.

Source-based reporting. No independent hands-on test by this newsroom.
Sources checked 2026-11-05 · Editorial updated 2026-11-05