Casey Neistat on ModRetro's Chromatic at OpenAI DevDay, vibe-coded games, and why hardware constraints make better experiences

Casey Neistat on ModRetro, vibe-coded games, and the case for constraints

ModRetro used OpenAI DevDay to put its Chromatic handheld into the hands of every attendee — 8,000 custom transparent units, announced on stage by Sam Altman. Within hours, one had appeared on eBay for $4,000.

The giveaway wasn't just a stunt. ModRetro built a custom Codex plugin that lets users vibe-code games and flash them directly to a Chromatic cartridge, skipping the laptop entirely. The demo that landed internally was ModRetro CEO Torin Herndon and his five-year-old vibe-coding a game together, then handing the cartridge to the kid to play immediately. The integration is deliberate: constrained hardware makes AI-generated games more completable and less janky than open-ended prompts aiming for AAA output.

“We made 8,000 of them to make sure every attendee could leave with one... We built a custom plugin for Codex and it enables you to sort of vibe code games. What's special about it is that you're not constricted to play that on your laptop. You can flash it to this cartridge immediately... Torin, our CEO, and his five year old vibe coded a game and then he was able to flash it and hand the chromatic to his kiddo.”

Constraints as the product

Neistat's argument is that constraint is a feature, not a compromise. A Game Boy-scale screen and four buttons set a ceiling that current AI can actually clear — something he says the breathless "one prompt, one Call of Duty" demos ignore entirely. ModRetro's recently launched M64, a Nintendo 64 console released about six weeks before this conversation, is drawing the same response: parents in their mid-40s who were 15 when the original shipped, introducing it to their kids and physically sitting beside them to play Mario Kart. Neistat contrasts that directly with finding his daughter hiding in her closet, iPad six inches from her face, playing Roblox alone.

Palmer Luckey's response to the nostalgia-bait critique — is this just old-man sentiment? — is that it isn't nostalgia. It was just better. Neistat uses the example of buying his daughter a point-and-shoot camera this past summer. She has a camera on her iPad. The single-purpose device still meant more to her. He sees the same dynamic playing out with wired headphones, film cameras on the Brooklyn Bridge, and Chromatics.

The LG controversy

Neistat flags what he describes as a real privacy issue with LG's premium OLED televisions. According to his account of a YouTube investigation by a pair of hackers (he couldn't recall the channel name), the microphone included for voice commands on LG's latest sets has been found to record ambient audio continuously — storing audio files locally when the TV is offline and uploading them when it reconnects to the network. Neistat says he now unplugs his bedroom LG from the wall. The episode is what pushed him to take Luckey's desire to build a "dumb TV" seriously.

The Samsung complaint is simpler: downloading Netflix on a guest-room Samsung requires creating a Samsung account he doesn't want and has no reason to have.

Creator economy and vibe-coded games

Neistat's optimism about AI-generated content is grounded in his own trajectory. After seven years of a successful YouTube presence, he had 280,000 subscribers. He started daily vlogging, maintained that cadence for 18 months, and reached 10 million. The inflection required cheaper cameras and faster editing tools — technology lowering the barrier, not talent alone. He sees AI doing the same thing at greater scale: the volume of output goes up, the volume of genuinely good work goes up with it, and eventually something AI-generated will be undeniably excellent in a way that forces even skeptics to concede.

The friction he identifies is curation, not generation. He watches both Mark Felton — a British historian he describes as an objectively poor filmmaker who makes the best videos on YouTube — and fully AI-generated World War II documentaries with clear human editorial oversight. He tolerates the AI versions because the images are transparently illustrative rather than deceptive, and because the factual grounding feels intact. Where he loses trust is in high-volume accounts that appear to pipe forum content into video without anyone actually reviewing it, producing advice that's already expired by the time it's published.

His underlying bet is that authentic expertise still draws attention regardless of production value. The HI Sutton model — a naval warfare specialist who presses record, narrates slides, and explicitly doesn't use AI — will always have an audience. So will three-hour podcasts watched by an editor who debates individual frames in his own work. The range of what people give their attention to is wider than the AI-content anxiety implies.