Adobe Built the Walled Garden. AI Might Be the Way Out.

When the software you depend on is controlled by someone else, how much control do you really have? That question usually stays abstract until a dialogue box blocks the path to yesterday’s files. In June 2024, Adobe put that question on every Creative Cloud screen that demanded a fresh acceptance of its Terms of Use.

Adobe's June 2024 terms fight made rented software feel suddenly visible

The controversy did not begin with a new generative model. It began with language. Adobe pushed a routine re-acceptance of updated Terms of Use to Creative Cloud and Document Cloud customers. Creatives seized on wording about Adobe accessing content through automated and manual methods, and many read that language as a broad licence to inspect, reuse or train on their work. Forums filled with cancellation threats. Trust, already strained by subscription pricing and AI anxiety, snapped into public view.

Adobe answered first with clarification, then with a rewrite. In a 6 June 2024 blog post, the company said the update was meant to explain moderation processes, not to claim ownership of customer work. On 10 June, Adobe president of digital media David Wadhwani’s follow-up promised clearer terms by 18 June. Adobe stated that customers own their content; that Adobe does not train generative AI on customer content; that Firefly training draws on licensed and public-domain sources; that local files on a user’s machine are not scanned; and that automated scanning of uploaded cloud content is aimed at illegal material such as CSAM, with human review in limited cases. Users, Adobe said, could opt out of product-improvement analysis that uses machine learning short of generative AI.

Those clarifications mattered. They also revealed the deeper wound. People were not only arguing about a paragraph. They were discovering how little leverage they felt once years of projects lived inside one vendor’s apps, formats and billing system.

Subscriptions, formats and cancellation friction keep the garden closed

The problem is rarely the sticker price alone. It is the accumulated cost of leaving. Creative Cloud sells access by the month. Cancellation can involve friction that users experience as a trap, a complaint Brandon Thomas later cited as a spur for ArtCraft. Project files accumulate in proprietary formats. Muscle memory settles into keyboard shortcuts and panel layouts. Studios standardise on Photoshop documents, Premiere projects and shared fonts. Switching is never “download the other app.” It is retraining people, converting archives, renegotiating client delivery, and accepting that some effects will not travel cleanly.

That is the walled garden Adobe built over decades: not only features, but an ecosystem where leaving hurts more than staying. Alternatives have always existed on paper. Getting a whole team through the gate is another matter.

In early October 2026, that dependency story collided with a different kind of software ambition. Brandon Thomas, a developer with a long Rust background and ties to ArtCraft (and earlier Storyteller.ai / FakeYou), announced seven free open-source applications aimed at familiar Creative Cloud workflows. PhotoCraft stands in for Photoshop, VectorCraft for Illustrator, FilmCraft for Premiere Pro, LightCraft for Lightroom, EffectCraft for After Effects, DesignCraft for InDesign and PrintCraft for Acrobat Pro. The repositories describe pure Rust implementations, native builds for major desktops, WebAssembly builds for the browser, and dual MIT or Apache-2.0 licensing. PhotoCraft’s README calls the work a clean-room reimplementation from public specifications and observed behaviour, denies Adobe code or assets, and labels the status early alpha. Thomas has said Anthropic’s Claude Opus 5.5 assisted the build, and that token sales around ArtCraft’s existing AI studio are meant to fund ongoing support.

The turning point he is selling is simple. What once required a large engineering organisation might now be attempted by a much smaller team with AI as a force multiplier. Whether that attempt becomes a profession-ready suite is a separate question.

Open source already had tools; AI is speeding familiar-workflow clones

ArtCraft is not walking onto an empty field. GIMP has long covered serious raster work. Krita is a first-choice painting environment for many illustrators. Inkscape handles vectors. Kdenlive and other editors cover non-linear video. Affinity’s one-time purchase model, DaVinci Resolve’s free tier and other commercial rivals already chip at Adobe’s price story. Open source itself is old news in creative software.

What feels new is the promise of proximity. Thomas’s pitch is less “here is another capable editor with its own conventions” and more “here is something that behaves like the tool your hands already know.” PhotoCraft’s own marketing leans into that familiarity: menus, shortcuts and panels where Photoshop users expect them, plus real PSD round trips claimed in the README. FilmCraft screenshots in press coverage deliberately echo Premiere’s layout. That is the wager. If the switching cost is mostly muscle memory and file fidelity, then a clone that preserves those may matter more than a technically strong tool that feels foreign.

AI enters as an accelerant for that wager. Reverse engineering and clean-room reimplementation are old techniques. The claim now is that large language models can compress the tedious mapping of behaviour into working code, so a small team can attempt suite-scale scope. That claim should be handled carefully. Speed of scaffolding is not the same as years of edge-case hardening. Still, the competitive landscape changes if the cost of attempting a familiar alternative falls sharply, even when the cost of finishing it remains high.

An impressive alpha is still not a production studio stack

Generating a large volume of interface and feature scaffolding is not the same as engineering a dependable professional application. Production-grade creative software is judged on crash rates under real files, colour-managed pipelines, plugin ecosystems, GPU drivers across machines, accessibility, security updates, and the boring work of keeping yesterday’s documents opening next year. PhotoCraft advertises substantial automated tests and PSD round-trip claims; those are encouraging signals, not a substitute for months of studio abuse testing.

Independent coverage has already applied a cold cloth. Creative Bloq’s Joe Foley, after trying the AI-coded suite in mid-October 2026, concluded it was impressive for its weight and origin story yet not stable or complete enough to replace existing software. Ars Technica similarly framed the apps as ambitious, free and nowhere near finished, while noting Thomas’s parity timeline slipped from a blunt “within a month” on Reddit toward “months and not years” on Hacker News. Missing features, rough performance and incomplete tools are normal in early alpha. They are also exactly why professional teams do not migrate on announcement day.

Maintainability is the quiet risk. If the suite is largely AI-authored, the question becomes who understands it well enough to fix a colour-management bug at 2 a.m. before a client delivery. Thomas says ArtCraft has team support funded by its AI product line. That is a plan, not a track record of decade-long Creative Cloud parity.

Clean-room claims still meet copyright, trade dress and contract lines

Law draws lines that slogans blur. Copyright protects expression in code and assets, not bare ideas or functional behaviour in the abstract. Independently implementing a feature described in a public file-format specification sits in a different posture from copying protected source, icons or presets. Classic clean-room practice separates the people who study a product from the people who write the replacement, precisely to reduce the risk of unconscious copying. ArtCraft’s repositories instruct contributors and agents never to read or disassemble Adobe application bundles and to observe behaviour by using the apps. That is the project’s stated method. Independent auditors have not certified that process.

Trade dress is another pressure point. Even when code is new, an interface that too closely mimics a competitor’s distinctive look and feel can invite a different kind of claim. Ars Technica’s Kyle Orland noted that risk when FilmCraft’s Premiere-like appearance circulated. Contract and licence terms can also restrict reverse engineering in some jurisdictions and products, separate from copyright’s outer bounds. None of this is a courtroom verdict on ArtCraft. It is the map of fights that become more likely as AI makes ambitious clones cheaper to start.

The policy tension underneath is older than Claude. How do we protect creators and companies that invest in tools without making meaningful competition impossible? If every familiar workflow is treated as untouchable territory, users remain tenants. If every clone is waved through because the bytes differ, investment in hard software may look irrational. The workable middle has always been contested. AI raises the temperature by changing who can afford to contest it.

Cheaper features do not erase Adobe's deeper advantages

Suppose AI keeps making software features cheaper to reproduce. Adobe’s durable advantages may shift further toward workflow integration, reliability, distribution, customer relationships, Sensei and Firefly services, stock and font ecosystems, and the simple fact that clients still ask for a PSD. Building software could become less exclusive. Building a trusted software business would remain difficult. A free clone that opens many files and then corrupts one critical archive is not a bargain.

What is still missing from the public record is decisive. There is no independent forensic confirmation of ArtCraft’s clean-room process. There is no long-run evidence that a community-plus-token-revenue model can match Adobe’s update cadence across seven professional surfaces. There is no settlement of the trade-dress question. There is also no proof that most working creatives will leave tools that already pay their rent. The open question is not whether Adobe should feel competition. It is whether competition arrives as durable infrastructure or as a viral alpha that fades when the next model demo lands.

Return to June 2024. The Terms of Use fight hurt because users felt cornered. Credible alternatives matter for that reason alone. They give people somewhere to point their feet. They force clearer contracts and better behaviour from incumbents. ArtCraft does not need to dethrone Photoshop to count. Even a partial, trustworthy escape hatch changes negotiations inside studios that today treat Creative Cloud as weather.

If AI lets more people build the tools they need, software users may remain dependent on a handful of dominant platforms, or they may gain real room to choose. The next few years will show which story we are actually living in. The garden walls are no longer imaginary. Neither is the possibility of a gate.