Brandon Thomas, the developer behind PhotoCraft, an AI-built open-source set of replacements for seven Adobe applications, has walked back the project’s initial hype and said it is “not anywhere close to ready.” The candid update came after he posted the work to Hacker News, where he acknowledged that he was not ready to show it publicly and said he would share it again once it clears beta.
PhotoCraft’s bold pitch runs into a reality check
PhotoCraft aims to mirror Photoshop, Illustrator, Premiere, Lightroom, After Effects, InDesign and Acrobat Pro, with all of it written in Rust and built for the browser through WebAssembly. Thomas used Claude Opus 5.5 to help build the applications, and the code is released under MIT and Apache licenses.
The project had already been promoted more aggressively elsewhere. Announcement posts appeared across five subreddits days earlier, presenting the software as in beta and suggesting it would reach roughly 100% parity within a month. The repository, however, now aligns with Thomas’s more cautious Hacker News message.
What the roadmap says about PhotoCraft
According to PhotoCraft’s ROADMAP.md, updated on October 11, the project is broader than it is finished. The repository says feature breadth is around 76%, based on counts that include 628 of 628 menus, 56 of 68 tools, 30 of 35 panels and 13 plus 4 of 31 formats. But that broader coverage does not translate into production readiness.
The same roadmap gives a much lower score for real-world use: readiness for real work is about 45%. It breaks that down further, saying a mainstream practitioner would get 43%, while someone using only core features would get 61%. Beta is estimated to be 1,100 to 1,800 hours away, with full parity 2,300 to 4,100 hours out, using Opus 5.5 agent wall-clock hours for one agent working sequentially.
The target is also narrowly defined. The roadmap says the benchmark is Photoshop 2026, version 27.11.0, measured from the installed application on October 10.
Why the project still calls itself alpha
The repository’s own description of the work is blunt. It lists 579 open issues for users of versions 0.5 and 0.6, says 22 of 25 performance budgets are missed, and notes that “PSDs we write are not yet reliably accepted by Photoshop.” For a project that wants to replace Photoshop, interoperability is not a minor flaw.
That gap is especially important because PhotoCraft is trying to do more than imitate a UI. It is attempting to produce files and behavior that are compatible with existing professional workflows, which means the bar is not just visual similarity but reliable round-tripping with Adobe software.
The scale of the codebase grew quickly
Even with those limits, the pace of development is substantial. The first commit landed on September 30, and the latest on October 13. In those 14 days, the repository accumulated 2,152 commits, 1,004 merged pull requests and 539,017 lines of Rust across 1,306 files.
Git records 317 distinct contributor identities, which means the “solo developer” framing stopped fitting almost immediately. Claude appears as a named commit author 277 times, reinforcing how heavily the project depends on AI-assisted coding.
Users who tried it found the rough edges quickly
Feedback from early users maps closely to the roadmap’s sober assessment. One Reddit commenter, Painted-Arcana, said the Photoshop clone was “bad,” arguing that even basic shortcuts and layout flexibility were missing and concluding after a few minutes that it was not a Photoshop replacement.
Photographers raised other concerns. Commenter Sea-Temporary-6995 reported noise and banding when lifting shadows on a Sony a7c RAW file. Another user, OspreyChaser, said several Nikon bodies were not supported for RAW work. On the other hand, the simplest app in the suite appears to be more stable: on Hacker News, commenter Krasnol said PDFCraft worked without bugs.
Clean-room development, tests and automation
Little of the discussion around PhotoCraft focused on the process documented in the repository. Its AGENTS.md says the team studied Photoshop and other proprietary editors for behavior and appearance only, while avoiding copying code, shaders, profiles or assets. Instead, the project says it is implemented from public specifications, including Adobe’s PSD spec, ICC, ISO 32000 blend modes and related papers, plus observation.
The repository also describes a test-heavy workflow. Every change is gated by tests, a panic_hunt test fuzzes each command with adversarial parameters, and format work is checked against a corpus of Photoshop-authored PSDs. The apps expose a JSON control channel and an MCP server, allowing the coding agent to drive the running app and screenshot the result to evaluate its own interface work.
Why the clean-room claim is controversial
That approach has drawn skepticism. One commenter, kmeisthax, dismissed it with the line: “There is no such thing as clean-room vibe coding.” The usual clean-room idea assumes separate teams, one documenting behavior and another implementing it without seeing the original source, which is not how this project is framed.
Still, not everyone is writing the project off. A commenter named az_liberal_geek, who is building an open-source SketchUp replacement called Hew with Claude, said an early version came together quickly but was buggy and needed many hours of testing and refinement after each major feature. That developer put parity at about a year, which is closer to PhotoCraft’s own estimate than to the original “within a month” messaging.
Thomas has already softened the timeline
Thomas has since moderated his public expectations as well. In a comment on Hacker News, he said “99% parity will take a while, but I’m sure it’ll be measured in months and not years.” That is still ambitious, but it is a far cry from the earlier beta-and-parity claims.
For now, PhotoCraft looks less like a finished Adobe alternative than an unusually fast-moving experiment in AI-assisted software development. Its codebase is already large, its roadmap is unusually transparent, and its limitations are equally explicit. The project may yet improve quickly, but by Thomas’s own admission, it is not ready for prime time.
Source: Original report
Was this helpful?
Explore more: Software Development More Software Development Tech News
Last Modified: October 22, 2026 at 10:33 pm
0 views
