Seven days. That's how long it took Markus "Notch" Persson to go from posting "Reject AI." to a five-million-view audience, to telling that same audience he's ready to try vibe coding on his own game. The reversal landed on July 21, 2026, and it's already being cited across gaming and tech circles as one of the year's cleanest examples of AI fatigue beating AI ideology.
bunpav's take: this isn't really a story about Notch changing his mind on whether AI-written code is good — he still isn't claiming that. It's a story about the labor market beating the principle, which is a more honest and more common outcome than most public AI reversals admit to. Here's the full timeline, what he actually said, and why it matters beyond one developer's Twitter feed.
TL;DR — what people are actually asking
| Question | Direct answer |
|---|---|
| What did Notch say before? | Called AI-code advocates "incompetent or evil" (Feb 2026); posted "Reject AI." (July 14, 2026). |
| What changed his mind? | Can't hire enough programmers for his studio, Bitshift Entertainment. |
| What did he say on July 21? | He'll "give vibe coding a shot" — not because he dislikes programming, but hiring is hard. |
| What will he use it for? | Non-core dull tools — a map editor, a node graph visualizer. Not core gameplay code. |
| What game is this for? | Levers and Chests, a voxel roguelike dungeon crawler at Bitshift Entertainment. |
| How did people react? | Split — some called it pragmatic, others called out the one-week irony. |
What did Notch actually say, in order?
The timeline matters here more than the individual quotes, because it shows a build-up rather than a sudden flip. In February 2026, Persson posted one of his more widely quoted anti-AI-coding statements: "using AI to write code is an incredibly bad idea still, and anyone advocating for it is either incompetent or evil." Fortune and other outlets covering it framed his objection as rooted less in raw resistance to new tools and more in concerns about losing autonomy, creative control, and what he considers the essence of programming as a craft — not simply "AI code is bad code."
That framing held through the spring. Then on July 14, 2026, Persson posted the two-word line that's since become the shorthand for his position: "Reject AI." It's about as unambiguous as a public stance gets, and it performed accordingly — 5.1 million views, roughly 70,000 likes, 5,700+ retweets, and over 2,100 replies within the day, per the post's own engagement stats.
One week later, on July 21, 2026, Persson posted the reversal:
"I just might give vibe coding a shot. Not because I don't like to do programming, but because I've been having a lot of trouble hiring talented programmers, and I'll probably feel less bad about firing a chat bot. I'm thinking it could make the dull tools I'd need to make myself to make this darn game. Like a map editor, or a node graph visualizer."
Read closely, it's a narrower claim than "Reject AI" suggested he'd ever make. He's not saying AI-written code is good, or that his February assessment was wrong. He's saying the hiring market has gotten bad enough, specifically for his small studio, that the tradeoff calculus changed — and he's carving the use case down to boring internal tooling rather than the actual game systems that define Levers and Chests as a design.
Why is hiring the stated reason?
Bitshift Entertainment is a small, independent studio — Persson founded it in 2024 after stepping back from an earlier, more ambitious plan for a spiritual successor to Minecraft, choosing instead to build Levers and Chests, a first-person voxel-based roguelike dungeon crawler. Small independent studios in 2026 are competing for the same scarce senior engineering talent as venture-funded AI startups and established studios simultaneously scaling live-service teams, and that competition has only intensified as the broader games industry has gone through repeated rounds of layoffs alongside genuinely tight pockets of specialized hiring demand.
Persson's own phrasing — that he'd "feel less bad about firing a chat bot" than a human hire — is doing real work in the post. It reframes the decision as a management and staffing tradeoff rather than an endorsement of AI-generated code quality. That's a meaningfully different claim than the ones being made by studios and executives who present AI coding tools as strictly better than human engineers; Persson is instead describing AI as a tool he's grudgingly willing to use specifically because hiring humans has become difficult, not because the tool is superior.
What is he actually planning to build with it?
The scope he named is narrow and deliberate: a map editor and a node graph visualizer — both classic "necessary but not fun" internal tools that most game studios have to build once and then maintain forever. Neither is core gameplay code. Neither touches the actual roguelike systems, combat, or level generation that define what Levers and Chests is as a game. That distinction is likely intentional; it lets Persson use AI-assisted coding for the least creatively significant part of the pipeline while keeping the "essence of coding" concern he raised back in February intact for the parts of the project he actually cares about authorially.
It's also consistent with how a lot of experienced engineers who are otherwise skeptical of AI-generated code have started adopting these tools in 2026 — not for core logic, but for scaffolding, internal tooling, and repetitive utility code where mistakes are cheap to catch and low-stakes to fix. Persson's stated use case fits that pattern closely, even though his own public rhetoric up to this point suggested a much more absolute rejection.
How did people react?
Reactions split along fairly predictable lines. Some longtime industry figures treated it as a pragmatic, unremarkable decision — a small studio using an available tool to solve a real staffing problem. Others focused on the timeline itself. Commentator Alex Finn, addressing an audience in the millions, framed it bluntly: "1 week ago the most famous game developer ever told his audience of 3 million followers to 'Reject AI' / Today, he said he is going to try vibe coding for the first time / This is fundamentally one of the biggest issues in America right now."
Alex Hormozi's response took a different angle, using the moment to needle the broader vibe-coding trend rather than Persson specifically: "People used to waste their time doing shit that didn't matter. Now they are vibe coding shit that doesn't matter." That's a jab at the category of tool use Persson described — throwaway internal tooling — rather than at Persson's consistency.
The through-line in most of the reaction, regardless of tone, is that Persson's specific case is being treated as a bellwether. He's not a random developer; he's arguably the most famous individual voice in games to have taken an absolutist public anti-AI-coding stance, and watching that stance narrow to "fine, but only for the boring parts, and only because I can't hire anyone" is being read as a signal for how much further AI-coding adoption still has to run among skeptics, not just enthusiasts.
Does this mean Notch is building Levers and Chests with AI now?
No — and that's worth being precise about, since headlines compress nuance fast. Persson said he "just might give vibe coding a shot" for specific dull tooling, not that Levers and Chests' development is pivoting to AI-generated code broadly. As of July 22, 2026, there's no confirmation he's actually shipped anything built this way yet, no timeline for when the map editor or node graph visualizer would appear, and no indication his position on AI writing the game's actual gameplay code has changed at all. Treat this as a stated intention, not a completed transition.
It's also worth noting this has no bearing on Minecraft itself. Persson has had no development role at Mojang or on Minecraft since Microsoft's 2014 acquisition of the studio; everything here concerns Bitshift Entertainment and Levers and Chests specifically, a project that has been in slow, community-visible development since the studio's 2024 founding.
What should indie developers take from this?
- Hiring pressure is a legitimate, separate reason to reconsider AI tooling — distinct from believing the code quality is actually good. Persson's framing keeps those two things explicitly separate, and it's a useful model for articulating your own position honestly rather than pretending a staffing decision is a philosophical conversion.
- Scoping AI tool use to non-core systems is a reasonable middle ground — internal tooling like map editors and visualizers is a lower-stakes place to experiment than gameplay-critical code, for teams still working out where they trust the output.
- Public AI stances age fast in 2026. Whatever position you stake out publicly, expect market and staffing pressure to test it faster than you'd expect — plan messaging you can walk back without it reading as hypocrisy, the way Persson's careful "not because I dislike programming" framing tries to.
- Watch Bitshift Entertainment's future posts for whether the map editor or node graph visualizer actually ships AI-assisted, and whether Persson's scope stays narrow or expands — that follow-through will say more than the initial announcement did.
Related reading
- Unity 7 Announced: 90% Faster Shaders, Zero-Rebuild Upgrade
- Unity CLI Is Free With Unlimited MCP — Terminal-Native Agents
- AI 3D Props for Indie Game Level Design
- AI 3D Models: Unity/Unreal Pipeline Guide
- Fallout 5 Pre-Production at Bethesda
- Billionaire founder of Minecraft slams AI code advocates — Fortune
Facts current as of July 22, 2026. This is a fast-moving social-media story — check @notch directly for any further clarification or reversal before treating this as Bitshift Entertainment's final production stance.