I think the biggest problem with AI in game development is that the focus remains on technically impressive visuals, and not on making games fun. That would make games with AI much more popular and acceptable to mainstream audiences, ultimately.
Using Codex or Claude in Unreal MCP makes me think that Unreal Editor is actually the bottleneck. It's slow, it stutters, it crashes, and now it's hardly necessary if Codex can do most of the actual development. It should just be a lightweight editor for precision work.
A strange thing about the rise of assistant apps like Dots or Grok Bot or Duet is finding that it's more enjoyable to just do most software development through that. Now I understand the people who would put Claude Code on a VPS and just text it to build a web app through a phone
PSA for those of u vibecoding naked on your laptop with no containers: tell your agents to DELETE Playwright NOW. Even the latest (1.64.1) bundles early Chromium.156 with
CVE-2026-118306
CVE-2026-118331
CVE-2026-118348
CVE-2026-118344
And the Playwright SANDBOX IS OFF BY DEFAULT
I've had good experiences with Longcat, but I also think it's far more akin to Sonnet than Opus. It's good for getting work done, but it's not as good at design or figuring out problems as Opus or Sol might be.
Meituan’s new LongCat-2.5 just dropped, and it’s already outperforming Opus 5 and GPT-5.6-sol on AutomationBench and BrowserComp. 🚨
A massive 1.6T MoE native multimodal model (~48B active) built specifically for long-horizon agents and real-world software engineering.
Massive step up from v2.0 with deeper visual reasoning and autonomous agent execution.
Pricing (per 1M tokens):
• Uncached In: ¥2
• Cached In: ¥0.04
• Out: ¥8