AI-first game engine analysis: test 93x faster claims

The phrase “vibe coded” hits game development discourse like someone threw a flashbang into a Jira board. Everybody sees white, nobody reads the ticket. PC Gamer’s Jacob Fox, syndicated by Yahoo Tech, reports that an ex-Blizzard employee claims to have built a “100% AI-first” engine for Em-8er, with backing from AMD. That is the kind of claim that deserves neither instant worship nor instant dunking. It deserves the sacred gamer ritual: open the settings menu, check the numbers, and see what actually changes.

The 93x claim is the boss fight

According to PC Gamer’s Jacob Fox, the game is Em-8er and the custom engine is the Kun-Zhi Engine. The reported claim is not just “we used AI,” which is basically every pitch deck wearing a wizard hat now. The sharper claim is the quoted line, “I spent 6 months using AI to replace Unreal Engine from under our game Em-8ER,” followed by the reported productivity claim that the team is developing “93x faster” than before. That number is the whole raid boss. “100% AI-first” is branding, and branding has the nutritional value of a loot box animation. “93x faster” is at least a testable workflow claim, even if the public evidence so far is still a claim rather than a benchmark suite. If you are a developer, do not argue with the sticker on the box. Ask what task got faster, who measured it, and whether that speed survives asset import, debugging, builds, tooling, deployment, and the thousand tiny gremlins that live between prototype and shipped game.

AMD backing makes this more than garage sorcery

PC Gamer’s report says the engine has backing from AMD, which matters because engine strategy is not just code. It is hardware relationships, driver behavior, toolchain support, and the brutal reality that “works on my machine” is how production pipelines go to die. AMD’s own Adrenalin Edition page says its software includes AI-powered features such as AMD Chat, AMD FSR, and an optional AI Bundle tied to the AMD ROCm platform. That does not prove Kun-Zhi is production ready, but it does show AMD is already packaging AI as part of its broader gamer and developer stack. Here is the constructive read: AMD backing gives the claim a bigger stage, not a free win. I am absolutely not handing out a Game of the Year trophy to an engine because a chip company is in the credits. That would be like rating a restaurant five stars because the oven manufacturer followed it on LinkedIn. But if an AI-native engine can connect faster iteration with real hardware support, then developers should pay attention, because iteration speed is where small teams either live, die, or spend six weeks fighting a shader bug named Kevin.

Score the workflow, not the buzzword

PC Gamer frames the project as “vibe” coded, and yes, that label is going to vacuum up the discourse like Kirby with a media studies degree. But builders should ignore the meme fog and score the system like a tool. My provisional rating is 7 out of 10 profiler windows, interesting enough to watch, nowhere near enough to trust without receipts. The right questions are boring, which is why they are good. Does “93x faster” mean faster feature creation, faster engine migration, faster content iteration, or faster code generation that still leaves humans holding the mop? Does the Kun-Zhi Engine reduce dependencies that made Unreal Engine too heavy for this team, or does it create a new dependency on AI-generated architecture that future developers will inherit like cursed armor? If the answer is “AI did it,” congratulations, you have described magic, not engineering.

The take for developers watching from the queue The lesson

from PC Gamer’s report is not that Unreal Engine is suddenly cooked, or that every studio should throw its pipeline into an AI blender and hit puree. The lesson is that AI-first tooling should be judged by where it changes production math. If a tool really makes a team dramatically faster, it should be able to show before and after evidence across tasks developers actually repeat. That is where this gets useful instead of just loud. Track build times, bug resolution, asset turnaround, migration cost, tool maintenance, and how much human review is needed before code stops behaving like a raccoon in a server room. If Kun-Zhi and Em-8er can back the “93x faster” claim with reproducible workflow data, it becomes a serious case study. If not, it is another hype goblin in the cave, glowing nicely, dropping vendor trash. For readers building games, the next move is simple: do not buy the label, test the pipeline. AI-native engines may become real leverage, especially for teams without AAA headcount, but only if the speed claims survive contact with production. Watch for demos, workflow breakdowns, and hard comparisons against existing tools. Until then, keep one hand on the hype meter and the other on the profiler.