RT
@BrannonSD: the discourse around epic’s AI use is getting ridiculous. the way some of you talk about this you’d think epic just handed every engineer a $20 claude subscription, blindfolded them, and told them to start shipping whatever it spits out.
that’s not how any of this works lol
first off, AI writing code and AI-generated code being blindly accepted into production are two VERY different things. code generation is one part of a much larger software engineering process. architecture, testing, code review, security analysis, integration, CI/CD pipelines, deployment controls, etc. don’t just magically cease to exist because an LLM was involved.
epic’s own VP and GM of the Epic Games Store, literally stated that their developers managed and supervised the AI-assisted launcher rewrite. AI agents were used to write and review code, but the development team was still directing the process.
does that automatically guarantee the code is secure? obviously not. but neither does having a human manually type every line. that’s why we have testing, reviews and security checks in the first place.
and i genuinely think a huge portion of the people arguing against this have absolutely no idea how capable modern agentic AI has gotten, especially in just the past couple of months alone.
it’s not 2024 anymore. we’re not talking about developers asking chatgpt for random code snippets and blindly copy-pasting them into their projects. modern coding agents can navigate entire repositories, understand relationships between components, make coordinated changes across dozens of files, execute builds and tests, identify failures, debug their own output and iterate on implementations. all while operating within whatever tooling, permissions and constraints the developers give them.
the improvements in reasoning, tool use and long-running autonomous development over even the last few months have been ridiculous. yet so much of the criticism still seems to be coming from people whose entire understanding of AI-assisted programming is based on what chatgpt could do two years ago.
none of that means modern models are infallible or that security reviews are suddenly unnecessary. but acting like AI-generated code is inherently some unmaintainable security nightmare while human-written code is automatically trustworthy is equally ridiculous. the quality and security of the resulting software depend on the actual engineering and validation processes, not simply whether a human physically typed every character.
and then there’s the claude screenshots..
you basically asked an AI leading hypothetical questions about what happens when AI writes code without proper oversight, and somehow took its answers as confirmation that epic’s products are suddenly going to be full of security vulnerabilities.
claude hasn’t audited epic’s source code. it hasn’t examined their development pipeline, reviewed their authentication architecture, or found a single exploitable vulnerability. it’s answering the hypothetical YOU gave it.
how on earth does that translate to stolen credit cards and more fortnite cheaters?
we’re talking about an AI-assisted rewrite of the epic games launcher that’s being released as a preview alongside the existing launcher. not some announcement that fortnite’s servers, anti-cheat and payment infrastructure have all been handed over to autonomous AI agents with unrestricted access.
sure, shared systems and integrations can introduce security risks, but you would need to actually identify some affected component or vulnerability before making claims like that.
but so many people seem to have already decided that AI involvement automatically means the absence of engineering oversight, despite having absolutely no idea what epic’s actual development process looked like.
there’s plenty of room for legitimate criticism of AI-assisted development without inventing an entire security disaster based on two claude responses to hypothetical questions.