🚀 Two More Perspectives from AI CodeCamp: From Game Dreams to Code Defenses

🎮 Creating Games with AI — and Surprising Yourself

As AI CodeCamp winds down, two very different—but equally telling—sessions rounded out the event.

Eric Berger’s “How to Use AI to Create Commercial-Quality Games” took a hands-on approach to building a small 3D game using his Gorilla3D framework. At first glance, this seemed niche—especially for those of us not actively building games. But what started as a demo of character models and dialog design turned into a mini-epiphany: it's not just about games.

The real story here was about resources—how AI can fill skill gaps in animation, texturing, audio, voice, and even storytelling. Even if you're not game-focused, Berger’s framework could be a springboard for all kinds of interactive applications. When asked directly about other use cases, he rattled off several that might help developers across industries, from training simulations to architectural walkthroughs. The fact that the full demo clocked in at around 180 lines of code made this even more approachable.

Gorilla3D—based on FireMonkey—could become a quiet powerhouse for Delphi developers looking to experiment beyond forms and grids. And for those with creative ideas but limited teams, Berger showed that AI can be more than hype: it can be help.

🧪 DerScanner: Vulnerability Scanning with AI—or Rules?

Valerie Kim from DerScanner presented a more complex—and less polished—talk: “AI Revolution: How to (not) generate vulnerabilities.” The aim was ambitious: use AI to detect, explain, and mitigate code vulnerabilities before they land in production.

At first, it sounded like an LLM-based solution that checks AI-generated code for errors. But as the session unfolded, it became clear that DerScanner is primarily a rules-based engine using techniques like SAST, DAST, and SCA—proven, if conventional, approaches. Their proprietary “Confi AI” engine helps reduce false positives, but it doesn’t appear to be an LLM in the usual sense. So, the “AI” might be closer to “automated inference” than GPT-style reasoning.

That said, DerScanner occupies a unique symbolic space right now. It reflects where many of us stand: trying to navigate this uncertain territory, wondering if we're witnessing the twilight of the old order—or awaiting a harsher reckoning. Some might recognize the echoes of Götterdämmerung or even Dies Irae here. One hopes for cleansing fire and rebirth—but the verdict isn’t in yet.

To be fair, DerScanner has been doing code scanning for over a decade. Their tooling likely excels at finding known vulnerability patterns in legacy code—especially useful when upgrading large systems. But the session suffered from technical glitches and some ambiguity. Was AI checking the rules engine? Were LLMs being queried at any point? Who makes the final call on the suggested fixes? These weren’t clearly addressed.

The broader criticisms of AI—hallucinations, data exposure, inaccuracy—felt tired. Most developers today recognize that humans make plenty of mistakes, too. The real challenge is deciding where automation can best assist without introducing new risks.

Still, Embarcadero seems to value DerScanner, and Barker’s endorsement suggests there’s more to come. A better-prepared demo and a clear architectural walkthrough would go a long way in showing how DerScanner actually works—and why it deserves a seat at the table.

🧩 Takeaways

• Berger's Gorilla3D: A compact, surprisingly versatile toolkit that could help even non-gamers visualize and animate ideas quickly with AI support.
• DerScanner: A mature tool whose value lies in its integration and history—but whose future in the LLM world remains a little opaque for now.

🤖 About This Post

Drafted using notes taken by a human, synthesized and organized with AI assistance, and refined by a human editor. A fine example of collaboration—AI as assistant, not oracle.