Developer Behind the AI-Built Adobe Clones Says They Are “Not Anywhere Close to Ready

Brandon Thomas put his AI-built open-source replacements for seven Adobe applications in front of Hacker News and immediately deflated them. "I wasn’t ready to post this to HN," he wrote. "It’s not anywhere close to read" He said he would post again once the project clears beta.

Announcement posts had gone up across five subreddits days earlier carrying a different message, projecting beta and roughly 100% parity within a month. The repository sides with the Hacker News version.

Thomas used Claude Opus 5.5 to build the applications in Rust, with WebAssembly builds for the browser, mirroring Photoshop, Illustrator, Premiere, Lightroom, After Effects, InDesign and Acrobat Pro. All of it ships under MIT and Apache licenses.

PhotoCraft’s ROADMAP.md, updated October 11, puts feature breadth at roughly 76%, from counts: 628 of 628 menus, 56 of 68 tools, 30 of 35 panels, 13 plus 4 of 31 formats. Then comes a second number. Readiness for real work is about 45%. A mainstream practitioner gets 43%; someone using core features only gets 61%. Beta is 1,100 to 1,800 hours out and full parity 2,300 to 4,100, counted as Opus 5.5 agent wall-clock hours for one agent working sequentially. The target is named exactly: Photoshop 2026, version 27.11.0, measured from the installed application on October 10.

Screenshot of PhotoCraft, an open-source Photoshop clone, editing a Hokusai print with adjustment, type and shape layers visible in the layers panel. (Source: GitHub)

The roadmap’s own account of why the project is still alpha is harder on it than its critics are. Users of 0.5 and 0.6 have filed 579 open issues. Twenty-two of twenty-five performance budgets are missed. And "PSDs we write are not yet reliably accepted by Photoshop." For software meant to replace Photoshop, file interchange is not a rough edge.

The scale is real regardless. First commit September 30, latest October 13. In those fourteen days the repository took 2,152 commits, 1,004 merged pull requests and 539,017 lines of Rust across 1,306 files. Git records 317 distinct contributor identities, so the "solo developer" framing stopped being true almost at once. Claude appears as a named commit author 277 times.

People who installed the apps split along a line that matches the roadmap. Reddit commenter Painted-Arcana went at the Photoshop clone directly:

It’s bad. You can’t even eye dropper by pressing alt. You can’t change the layouts much beyond the static presets. I could go on and on but after 2 minutes with the app it is very clear this isn’t a photoshop replacement at all.

Photographers hit the RAW pipeline. Commenter Sea-Temporary-6995 found that raising shadows on a Sony a7c file produced noise and banding. Commenter OspreyChaser found no RAW support for several Nikon bodies. The simplest app fares best: on Hacker News, commenter Krasnol reported PDFCraft working without bugs.

Little of the coverage mentioned how much process sits in the repository. AGENTS.md carries a clean-room rule: the team "studied Photoshop and other proprietary editors for behaviour and look only," never copying code, shaders, profiles or assets, implementing "from public specs (Adobe PSD spec, ICC, ISO 32000 blend modes, papers) and observation." Tests gate every change, a panic_hunt test fuzzes every command with adversarial parameters, and format work is checked against a corpus of Photoshop-authored PSDs. The apps also expose a JSON control channel and an MCP server, so the coding agent can drive the running app and screenshot it to check its own interface work.

That language is where opinion divides. Commenter kmeisthax was flat about it: "There is no such thing as clean-room vibe coding." The doctrine normally assumes two separated teams, one writing specs and another implementing without seeing the original.

The most measured voice came from someone running the same experiment. Commenter az_liberal_geek is building Hew, an open-source SketchUp replacement, also with Claude:

The very first version of Hew came out fast and looked and worked fantastic. But it was also super buggy and failed in unexpected ways. Adding a major feature takes a prompt + several hours of Claude develop, followed by many more hours of my own testing and refining.

He put parity at a year, not a month. The repository’s own estimate lands in similar territory.

Thomas has since adjusted the public figure too, telling Hacker News that "99% parity will take a while, but I’m sure it’ll be measured in months and not years."