Last Week in AI #339 - DLSS 5, OpenAI Superapp, MiniMax M2.7
Last Week in AI Last Week in AI
NVIDIA's DLSS 5 reportedly uses generative AI to redraw game frames in real time, and OpenAI may be ditching consumer apps to chase business cash instead. Also: MiniMax's M2.7 model quietly launched.
Based on reporting by Last Week in AI, Last Week in AI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
There's a lot crammed into this week's roundup, but the DLSS 5 story is the one that'll make gamers do a double take. Instead of just upscaling resolution like earlier versions, reports describe DLSS 5 as something closer to a generative filter sitting on top of the rendering pipeline, reconstructing and even inventing frame details on the fly. If that holds up, it's less an incremental graphics update and more a sign that real-time generative AI is quietly becoming standard equipment in consumer GPUs, not just a chatbot party trick.
Meanwhile, OpenAI appears to be recalibrating its ambitions. The word going around is that the company is pulling back from the everything-app dream, the one where ChatGPT becomes your shopping assistant, your social feed, your personal concierge, and instead narrowing focus toward business and productivity tools. That's a notable retreat for a company that spent the last year talking about superapps and consumer ecosystems. It suggests the economics of chasing consumer engagement against Google, Meta, and TikTok just don't pencil out the way enterprise contracts do.
Then there's MiniMax dropping M2.7, another entry in the fast-moving Chinese open-model scene that keeps shipping new checkpoints at a pace Western labs struggle to match. It won't grab headlines the way DLSS or OpenAI's strategy shift will, but it's part of the same pattern: model releases are becoming routine infrastructure updates rather than events, and the companies still pursuing consumer mindshare are increasingly the ones with hardware or platform leverage, not just a good chatbot.
Put together, the three stories sketch a rough map of where the industry's energy is actually going. Hardware makers are baking generative tricks directly into silicon, an AI lab that once wanted to be your everything is retreating to enterprise safety, and open-weight competitors keep grinding out releases in the background. None of that is flashy on its own, but it says something about where the money and the attention are actually settling as the initial consumer hype cools.
My take — AI-written commentary, not fact-checked reporting
OpenAI backing off the superapp fantasy doesn't surprise me one bit — building a consumer platform from scratch against Google and Meta was always the harder, dumber bet compared to selling API access and productivity seats to enterprises that already have budget lines for this stuff. The real story to watch is DLSS 5, because once generative AI is embedded in your GPU driver rather than a website you visit, the 'AI bubble' conversation gets a lot more awkward for the skeptics.
Read more about this at: Last Week in AI