Skip to content
How I AIHow I AI

OpenAI DevDay 2026: Dots, Spaces, and ULTRAFAST

I spent the day at OpenAI’s DevDay in San Francisco, and I have good news and bad news: OpenAI released a lot of stuff. In this episode, I break down the announcements worth paying attention to - and show you what happened when I tested some of them early. We’ll meet my Dot, explore why Spaces and Sites could matter for how teams work, and get into the model and API updates I’m most excited about as a developer. I use the Decisions API to find podcast thumbnails where nobody looks awkward, build a collaborative sketchpad with Astra ultrafast, and let my kids redesign a 3D world in real time. That last experiment cost about $97. My wallet has thoughts. These are my early impressions: what’s promising, what still feels rough, and what I think you should try first. What you’ll learn: 1. What OpenAI’s Dots can do, how I’ve been using mine, and why I’m waiting to give a full verdict 2. Why Spaces might be one of the most underhyped announcements for collaboration between humans and agents 3. How Sites with connectors and plugins could help teams share internal tools with the right data permissions 4. Where GPT-6.1 Sol fits in my model stack—and why speed and cost matter 5. What vision adds to the Decisions API, including my thumbnail-selection and hot dog demos 6. What Astra ultrafast makes possible for interactive AI apps, from collaborative drawing to a changing 3D game 7. Where the speed feels magical, where the experience still needs work, and what it costs In this episode, we cover: (00:00) OpenAI DevDay recap—and pressing the Codex reset button (00:58) Dots: early impressions and rough edges (06:57) Spaces: working with humans and agents (10:37) Sites, connectors, and sharing internal tools (13:06) Models and platform: GPT-6.1 Sol (14:36) Decisions API: fast decisions with vision (15:27) Finding better podcast thumbnails with AI (16:29) Hot dog or not hot dog? (17:17) Astra ultrafast: speed, pricing, and possibilities (18:50) The Other Pencil: drawing alongside AI (19:45) Little Starship: a 3D world you can change with a prompt (21:24) The $97 AI game—and what it makes possible (22:25) Agents API, computer use, plugins, and plan updates (23:03) What I’d try first — Tools referenced: • ChatGPT — Dots, Spaces, and Sites: https://chatgpt.com/ • Codex: https://openai.com/codex/ • OpenAI API — GPT-6.1 Sol, Decisions API, and Astra ultrafast: https://platform.openai.com/ • Jev: https://typesafe.ai/ Other references: • OpenAI DevDay 2026: https://devday.openai.com/ Where to find Claire Vo: ChatPRD: https://www.chatprd.ai/ Website: https://clairevo.com/ LinkedIn: https://www.linkedin.com/in/clairevo/ X: https://x.com/clairevo Production and marketing by https://penname.co/. For inquiries about sponsoring the podcast, email jordan@penname.co.

Claire Vohost
Sep 30, 202624mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

DevDay 2026: Always-on agents, collaborative Spaces, ultrafast models, vision decisions

  1. Claire recaps OpenAI DevDay 2026 as an unusually dense release slate spanning agents, collaboration surfaces, site/app sharing, and new model/runtime options.
  2. Dots are introduced as a persistent, multi-channel agent (ChatGPT, Codex, phone calls, Slack) with its own computer and optional local access, though Claire finds its thread/surface structure confusing and prefers multiple specialized agents today.
  3. Spaces is highlighted as an under-hyped enterprise feature: AI-native documents and slides designed for human-and-agent collaboration with robust permissions and sharing.
  4. Sites gains the ability to bundle connectors/plugins so shared internal dashboards and tools respect each viewer’s data access—directly addressing governance and distribution for “vibe-coded” enterprise prototypes.
  5. On the model/platform side, GPT‑6.1 Sol emphasizes cheaper speed, the Decisions API adds fast constrained outputs with vision, and Astra ultrafast demonstrates compelling near-real-time apps despite steep costs.

IDEAS WORTH REMEMBERING

5 ideas

Dots aim to be a single, always-on “master agent,” but the UX and mental model are still unsettled.

Dots are positioned as a persistent, cross-surface delegate that can act proactively, use a dedicated cloud computer, and optionally access your local machine—making it closer to an “always-on coworker” than a chat session.

Spaces could become the enterprise collaboration wedge if adoption follows the governance and sharing features.

Claire sees Spaces as the long-missing AI-native place to co-edit documents and slides with both people and agents, with enterprise-grade sharing/permissions—potentially challenging Google’s collaboration moat.

Sites + connectors is OpenAI’s answer to “vibe-coded internal tools” distribution, governance, and data access.

By allowing Sites to bundle connectors/plugins (e.g., Snowflake) and preserve per-user access controls when shared, OpenAI is tackling a key enterprise pain: what to do with internally built, AI-generated dashboards and prototypes.

GPT‑6.1 Sol targets the “fast + cheap” sweet spot many developers actually need.

GPT‑6.1 Sol is framed as a strong, cheaper workhorse option—near frontier usefulness at substantially lower token pricing than Astra—ideal for everyday engineering tasks when cost matters.

Fast decision models become far more valuable when they can see—vision turns “routing” into “real automation.”

The Decisions API is described as a Jev/Luna-style constrained decision model, but with an important differentiator: vision support, enabling fast classification over images/frames (e.g., auto-picking non-awkward podcast frames).

WORDS WORTH SAVING

5 quotes

The good news is OpenAI released a lot of stuff. The bad news is OpenAI released a lot of stuff.

— Claire Vo

Dots are an always-on long-running agent that can work across ChatGPT, across Codex, in its own computer, on your computer, on your phone, whether through text or through call, even in your Slack, to get work done for you.

— Claire Vo

We have been waiting for an AI native document, slide, HTML collaboration platform with all the right sharing built in, with all the right data controls built in, with all the, like, agent native kind of aspects of this built in.

— Claire Vo

It has vision. It has vision. So we get the speed, cost of, like, a Decisions API, a Jev style model, with eyeballs, with computer eyeballs.

— Claire Vo

Now, real talk, it cost me, like, $97 or something to run this for 30 30 minutes.

— Claire Vo

Dots long-running agent across surfacesDot computer + local computer accessSpaces for human-and-agent documents/slidesSites app builder with bundled connectors/pluginsEnterprise sharing, permissions, and governanceGPT‑6.1 Sol pricing and positioningDecisions API: constrained outputs + vision + speed

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.