Portal and Half-Life writer Erik Wolpaw recently appeared on a podcast and said that he and a small team at Valve have been "poking around" at using AI to generate text, audio, and more. But don't expect Portal 3 or Half-Life 3 to be built by machines. Wolpaw was explicit about what this actually is: a small group of people at Valve looking at some AI stuff, not a strategic mandate from the company.
This distinction matters. As 2026 arrives, AI has moved from experimental novelty to production reality, with half of all studios actively deploying AI tools and nearly all developers leveraging AI-assisted workflows. When Valve talks about experimentation, people listen. When Wolpaw clarifies it's just a small group, people still worry. The real story isn't what Valve is doing right now—it's what Wolpaw reveals about where AI might actually help, and where it unambiguously fails.
Wolpaw was clear on AI's limitations: it's not good at being especially creative, and it's not good at being funny. He's tested it. The technology hasn't mysteriously improved since November or changed its nature. What matters is where he sees genuine potential. Wolpaw explained that in game writing, developers have always had to simulate characters reacting to whatever players do, and it's the one place where he feels AI is worth investigating.
The distinction is important. AI struggles to invent compelling narratives or comic timing. It doesn't get why a joke lands. But generating contextual responses to random player input—making an NPC acknowledge the chaos you're causing without needing a thousand hand-written scenarios—that's different. Wolpaw imagined a Grand Theft Auto scenario where you're creating physical and social chaos, and the AI plays the straight man, reacting to whatever insanity happens. It's reactive composition, not creative invention.
Wolpaw said he's very interested in AI in this specific case because it would allow you to do something that is impossible, no matter how many humans you throw at it. This is honest work. It acknowledges a practical constraint: you cannot hire enough writers to script every possible player interaction in a sufficiently complex game. You can hire AI to handle it.
Yet Wolpaw also made clear what this isn't. He said the idea of using AI to replace people or make games cheaper didn't interest him at all, and he just thinks the tech could be used to make a better experience for people playing the game. This is where institutional incentives collide with creative intentions. Wolpaw wants better games. Executives want lower costs. Both cannot always be the same thing.
The gaming industry's relationship with AI has soured. Over one-third of game industry professionals are using generative AI tools as part of their job. Usage is rising rapidly. Sentiment is plummeting. Only 7% of respondents think generative AI is having a positive impact on the game industry, down from 13% in the previous year. Workers in visual art, game design, and programming hold the most unfavourable views. They're not opposed to tools that make their work better. They're concerned about tools that make them redundant.
This is the real conversation Valve's quiet experiments hint at. AI isn't good or bad in the abstract. It's a tool whose value depends entirely on how it's deployed. Used to replace junior artists and writers to pad executive margins, it erodes an industry built on talent. Used to solve specific, unsolvable problems—like generating thousands of contextual NPC responses—it might create experiences that weren't previously feasible. The difference is execution.
Wolpaw's comments are valuable precisely because he isn't evangelising AI. He's describing its actual limits and actual uses with the precision of someone who has spent weeks testing it. That honesty is rare in the current landscape, where venture capitalists and some studio leadership speak as though generative AI will make game development cheaper and faster while creating better art simultaneously. It won't. The real question is whether companies will use it as Wolpaw describes—to augment human creativity in narrow, specific ways—or as a cost-cutting hammer that erodes the craft.