ThursdAI - The top AI news from the past week · From Weights & Biases, Join AI Evangelist Alex Volkov and a panel of experts to cover everything important that happened in the world of AI from the past week

📆 Oct 9, 2025 — Dev Day’s Agent Era, Samsung’s 7M TRM Shock, Ling‑1T at 1T, Grok Video goes NSFW, and Serverless RL arrives

·1 hr 41 min·5 clips

Samsung’s 7-million-parameter TRM hits 45% on Rkgi-1 and 8% on Rkgi-2.

The episode starts in a joking but uneasy place, with the hosts asking for proof that everyone is real. They frame the week around a world where Sora, the upcoming release from the Big G, possible Meta video updates, and Grok video make the boundary between digital and physical feel thin. That setup leads straight into the platform question. Builders are asking how to keep their apps specific when chatbots blur context and speak for the product. The conversation turns to OpenAI’s position in that stack. The panel treats it as a serious move to sit between users and software. They describe a future where people may not go into apps themselves at all. Instead, they will ask an assistant to do the task and let the assistant drive the app. That lands as both smart and unsettling. The hosts point out that OpenAI’s scale makes the strategy possible. They also note the tradeoff for builders, who lose some control over how their content and interfaces are presented. The episode keeps returning to the same pressure point: who owns the user relationship when the UI becomes an AI conversation. The discussion makes room for the current state of the tooling too. For now, a lot of the integrations still feel like controlled interfaces rather than fully agentic action. The panel says these MCP connections are still being handled through an interface, like an iframe with buttons. That keeps the episode grounded in what is actually shipping, not just the marketing promise. The tone stays fast and opinionated throughout. The hosts move between seriousness and jokes without slowing the pace for long. They treat the release cycle as something to metabolize live, not summarize from a distance. That gives the episode a specific texture: excited, messy, and very current. The segment on post-reality pushes the same theme further. The hosts talk about a world where it is getting harder to tell whether a video was shot in actual meat space or generated. They connect that feeling to the broader AI news cycle and to the way these tools change public trust. The concern is not abstract. It is tied to product design, moderation, and audience perception. The episode also keeps one eye on the audience itself. The hosts mention the show’s growth, the newsletter version, and the fact that the podcast this week will carry more of the questions from the recording as well. That makes the episode feel like a live operation, not a finished studio artifact. It is a working room, and the listeners are being brought into that room on purpose. By the end, the show has covered a lot of ground without pretending the week was neat. It leaves listeners with the sense that the AI platform race is already reshaping what an app is, where the interface lives, and how much control builders still get to keep. The episode is less about one headline than about the collision of all of them at once. The closing stretch keeps the same brisk rhythm. The hosts thank listeners, point them to the newsletter for links, and sign off with the kind of energy that suggests the next batch of releases is already on deck. That gives the episode a newsroom feel, but with more personality and less polish. It sounds like people trying to stay ahead of the wave while standing inside it. The result is an episode that works as both a news sweep and a strategic warning. It is not trying to calm anything down. It is trying to name the shape of the thing while it is still moving. That is what gives it force. The final impression is of a show that understands AI as product, infrastructure, and culture all at once. It keeps the jargon close, but never lets the jargon hide the stakes. The listener comes away with more than a list of releases. They come away with a sense of where the center of gravity is shifting. The episode also leans into the social side of the AI moment. The hosts talk as if the audience already lives inside the ecosystem. That assumption lets them move quickly through references to agents, app layers, and platform control without stopping to tutor the room. It is efficient, but it is also a statement of identity. There is a steady undercurrent of practical anxiety beneath the jokes. If an AI assistant becomes the front door to every app, then the app maker is no longer just designing a product. They are negotiating with a platform. The episode keeps that tension alive without flattening it into theory. That is why the discussion feels more useful than a standard roundup. It is not merely collecting headlines. It is trying to predict the shape of the next workflow. The panel keeps asking what builders lose, what users gain, and who gets to mediate the experience. Even the light banter serves that larger theme. The show’s fast exchanges mirror the pace of the field it is covering. Releases land, reactions follow, and the meaning gets negotiated in real time. That makes the episode feel immediate in a way a scripted recap would not. Listeners who follow AI closely will recognize the value in that. The episode is messy in the way the week itself is messy. It keeps the uncertainty visible. That is what makes it informative. The sign-off reinforces the show’s role as a weekly dispatch. The hosts push the newsletter, tease the next format, and invite people back into the loop. It feels communal, but still pointed. The episode ends like a live room that knows the news will not wait.

As heard by us

A loose but sharp look at AI's reality problem and the product questions it raises.

The episode stays with the uneasy shift that follows AI into everyday use: if people cannot trust what they are seeing, what does that do to the software built around it?

Read the full review in PlayNext →

Why you'd press play

You want the AI week, but with the volume turned up and the stakes explained.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from ThursdAI - The top AI news from the past week