After a month of IPOs, model launches, and a government ban, today the action moved to where AI actually touches you. Snap launched real consumer AR glasses built to replace your phone. ChatGPT turned itself into something that runs tasks on a schedule instead of waiting for you to ask. And ByteDance's Seedance, already the best AI video model in the world, is lining up its next leap. Three companies, three fronts: AI on your face, AI on a schedule, and AI video closing in on indistinguishable-from-real. Here are the three drops.

At Augmented World Expo yesterday, Snap unveiled Specs, its first consumer AR glasses, and CEO Evan Spiegel made the pitch explicit: the smartphone era is ending. His words, almost 20 years after the iPhone launched, people are ready to think about computing differently. Specs cost $2,195, are available to preorder now with a $200 refundable deposit, and ship this fall in the US, UK, and France.

The hook is the AI, not the hardware. Specs are fully standalone, no tether or puck, running two Snapdragon chips, a 51-degree field of view, and 7-millisecond latency. But what Snap is selling is contextual AI you wear. In the company's framing, AI is no longer limited to a text box. It can see what you see, understand what you are trying to accomplish, and help in the moment. Look at an object and ask about it. Get directions painted onto the street. See where to pour the coolant under your car's hood.

For your audience specifically, here is the detail that matters: developers build for Specs using Claude Code, Codex, and Cursor, the same agentic coding tools you already use. The competition is crowded, Meta's Ray-Bans, Google's upcoming glasses with Samsung, Apple's rumored N50, and at $2,195 this is a bet on enthusiasts first. But it is the clearest sign yet that every major company now believes the screen you stare at is the thing AI is going to replace.

OpenAI rolled out a real scheduled-tasks system in ChatGPT today, with a dedicated Scheduled page in the sidebar where you can see, pause, edit, and delete anything you have asked ChatGPT to do in the future. You can set one-off or recurring tasks, at a specific time or across a window like morning or evening, and set up monitoring tasks where ChatGPT proactively checks the web or your connected apps and pings you. It is rolling out to Go, Plus, Pro, Business, and Enterprise users on web and mobile.

The bigger move is what it replaces. OpenAI is sunsetting Pulse, its proactive daily-summary feature, within 14 days, and folding that into scheduled tasks. Instead of a separate Pulse product, you just tell ChatGPT to give you a daily briefing based on your interests and connected apps, and it runs as a scheduled task. There are real guardrails worth knowing: tasks run at most once per hour, you can have up to 10 active at a time, and unattended ones may auto-pause.

Why it matters: this is ChatGPT becoming a background worker instead of a thing you talk to. In OpenAI's own framing, most people never even realized they could schedule a prompt. Making it a front-and-center feature nudges hundreds of millions of users toward treating ChatGPT as something that does things on its own time, not just answers when spoken to. That is the same shift the whole industry is making, from assistant-on-demand to agent-on-a-schedule.

ByteDance's Seedance is, right now, the best AI video model in the world, and it is reportedly weeks from getting better. First the confirmed part: Seedance 2.0, released in February, sits at the top of the Artificial Analysis video arenas for both text-to-video and image-to-video, ahead of Google's Veo, Kuaishou's Kling, and OpenAI's Sora. It generates cinematic clips up to 15 seconds with native audio from text, image, audio, or video inputs. It is the model behind the viral clips earlier this year, the fake Brad Pitt versus Tom Cruise fight, the Friends-as-otters videos, that were realistic enough to trigger cease-and-desists from Disney and Paramount and a letter from US senators.

Now the leaked part, framed honestly as leaks: according to Dev Mode sightings circulating today, Seedance 2.5 is expected in early July with support for generations longer than 15 seconds, and Seedance 2.0 is expected to get 4K resolution support imminently, possibly this weekend. ByteDance has not officially confirmed either date, so treat the timing as rumor, but the direction is consistent with the public roadmap, which has long pointed to 4K and real-time generation.

Why it matters: AI video is the frontier where "is this real?" is genuinely getting hard to answer, and the current leader is a Chinese model most of the US AI conversation underweights. A jump to 4K and longer clips pushes it further toward replacing actual production work, which is exactly why Hollywood is alarmed. Whatever your view, the best video model is getting better fast, and it is not made by the US labs everyone watches.

Three drops, three fronts, one direction: AI is pushing deeper into the real world.

Snap is putting it on your face, betting the assistant belongs in your field of view, not your pocket. OpenAI is putting it on a schedule, turning ChatGPT from something you query into something that works in the background. And ByteDance is pushing AI video toward the point where generated footage is indistinguishable from filmed reality. Worn AI, scheduled AI, and AI that manufactures convincing video. For two years, using AI meant opening an app and typing. This week the frontier moved to your face, your calendar, and your screen's relationship with the truth. The chat box was the first form factor. What comes after it is being built right now, on three different fronts at once.

Three fronts moved today. Snap launched Specs, its $2,195 consumer AR glasses for what Evan Spiegel calls the post-smartphone era, with standalone hardware, contextual AI that sees what you see, and developer support through Claude Code, Codex, and Cursor, shipping this fall. OpenAI rolled out a proper scheduled-tasks hub in ChatGPT, with a dedicated page to manage recurring and monitoring tasks, while sunsetting Pulse within 14 days, turning ChatGPT into a background agent that works on its own schedule. And ByteDance's Seedance, already the top-ranked AI video model ahead of Veo, Kling, and Sora, is reportedly weeks from Seedance 2.5 with longer clips, plus imminent 4K support for 2.0, though ByteDance hasn't officially confirmed the dates. The thread: after a month of IPOs and bans, AI moved off the screen and into glasses, schedules, and video that's getting hard to tell from real. The chat box was the first form factor. This week three companies bet on what comes next.

Stay building. 🤖