AI Agents AI Gadgets & HW AI Models - LLM AI Open Source AI Security AI for Coding AI for Gaming AI for Images AI for Music AI for Videos Artificial Intelligence Editor's Choice NVIDIA AI Other News Robotics Tech Face-off Tech Satire

Facepunch Studios Addresses AI Slop in S&box Launch

By Artūras Malašauskas Apr 30, 2026 4 min read Share:
Garry Newman acknowledges AI-generated content flooding S&box's discovery tab and promises to push low-quality AI slop off the main page while maintaining a neutral stance on AI tools.

The successor to Garry's Mod has hit a familiar problem from the first day of launch. S&box, the open-source game development platform from Facepunch Studios, is already being flooded with what players are calling "AI slop" — low-quality, generative AI-created game modes that are burying human-made content in the discovery tab.

Steam reviews for the title currently sit at mixed, with multiple users explicitly citing AI-generated content as a core issue. One reviewer with 67 hours logged described the discovery tab as "an unoptimized mess with a 'game hub' filled with 'vibe coded' AI slop." Another called the platform "AI-infested."

Garry Newman, founder of Facepunch Studios, has acknowledged the problem directly. In a statement to Rock Paper Shotgun, Newman confirmed the team will take action to promote human creativity and push obviously AI-created slop off the main page.

"Low quality, obvious AI-created slop is going to be a growing problem in every creative outlet," Newman said. "We don't encourage using AI to be creative. We don't encourage using AI to create games for you. But we do acknowledge that it's a good learning tool and it's a good productivity tool."

The Steam page for S&box carries a disclaimer that the game "contains user-generated content, which may contain AI generated stuff," while clarifying that Facepunch itself does not "AI generate anything." This distinction matters — the studio is not building AI tools into the platform, but it's also not banning users from employing them.

Newman's full statement, published on Shacknews, doubles down on supporting AI as a learning mechanism. "AI is the teacher now," he wrote. "It's how people are going to learn how to program from now on. They're going to be generating their own examples and have it explain the code to them."

The studio is drawing a line between AI as a productivity tool and AI as a replacement for human creativity. Facepunch won't include an AI Assistant button or a "make game" wizard that generates code automatically. But if someone builds the best game in the world using AI help, the studio won't tell them they can't post it.

This is where the friction lives. The discovery algorithm isn't perfect right now (a problem that has plagued user-generated platforms since the dawn of Steam Workshop). The main aim is that good games float to the top and bad games float to the bottom, while offering fresh content to people all the time. But when AI can generate thousands of game modes in minutes, the signal-to-noise ratio collapses.

Users clicking through the discovery tab will notice the difference immediately. Human-made content typically has polish — consistent textures, working mechanics, intentional level design. AI-generated slop often feels hollow, like a house built from cardboard boxes. The thumbnail might look professional, but the actual gameplay experience lacks cohesion.

This isn't the first time a major developer has faced backlash over AI integration. Embark Studios utilized AI to generate additional voice lines in ARC Raiders and The Finals, using the voices of paid actors. The studio later re-recorded many lines with real voices after player pushback.

Facepunch is taking a different approach. Rather than removing AI-generated content entirely, the studio plans to adjust its discovery algorithm to surface human creativity more prominently. The exact mechanics of this adjustment remain unclear. Will it involve manual curation? Machine learning classifiers? Community reporting systems?

The practical reality is that moderation at this scale is extremely difficult. AI tools can generate content faster than human moderators can review it. Even if Facepunch pushes AI slop off the main page, it will still exist in the platform. Users seeking it out will find it.

There's also the question of whether this strategy will satisfy either side of the debate. Players who want AI-free experiences may still feel the platform is compromised. Developers using AI tools may feel unfairly penalized if their content gets buried.

Whether users actually pay for S&box long-term remains the real question. The platform has been in development for years, utilizing a heavily modified version of the Source 2 engine. It allows creators to publish their creations as standalone games on Steam, which could open the floodgates to even more AI-generated titles.

Newman promised weekly updates and listening to player feedback so the platform rounds out into something as beloved as Garry's Mod or Rust in the future. But the AI slop problem won't solve itself with patches. It requires ongoing moderation decisions that will inevitably disappoint someone.

The original Garry's Mod remains highly successful, priced at $10 since its launch 20 years ago. An update arriving April 29th will add mounting support for Black Mesa, the fan-made Half-Life remake. That's less depressing news for now.

Facepunch has acknowledged the problem. The studio has outlined a general direction. The execution will determine whether S&box becomes a thriving creative platform or another casualty of the AI content flood. Time will tell if the discovery algorithm can actually distinguish quality from quantity.

Arturas Malas Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn
Share:

Comments

Sign in to comment:
    <