You watched OpenAI's DevDay keynote and moved on. Here are the 4 announcements (
By Jeremygrandillon · October 2, 2026 · Curated by George's Blog
You watched OpenAI's DevDay keynote and moved on.
Here are the 4 announcements (actually) worth acting on for GTM:
→ Dots
ChatGPT's out-of-the-box agents.
The real GTM value is in adoption: onboarding flows a lead/client can complete.
Later, you share Dots that do specific GTM job, like qualifying a lead or generating a list, right inside the conversation where your client already lives.
Build a Dot that finishes one GTM task for the customer, and you're in the workflow before they know your name.
→ GPT-6.1 Sol
Astra is best in class today, and Sol 6.1 just landed: excellent and cheaper.
The point isn't the benchmark, it's availability.
You get frontier-level intelligence you can actually afford to run in every GTM workflow. Enrichment, scoring, drafting, and research all run at volume, without the bill deciding what you're allowed to try.
→ Agents API (with computer use)
This is the big one.
Agents that can use CLI tools, so you can programmatically execute real workflows through the API.
Those are real multi-step jobs that end in an action, not a draft.
→ Decisions API
The judgment layer.
Finite questions, predefined answers, fast and cheap.
It's OpenAI’s version of Jev (you may have heard about it recently): a model that decides instead of writes.
On its own it's nice.
Combined with something like Clay, it's huge.
Most of your workflows aren't writing. They’re the same small calls repeated 40,000 times.
Is this title a buyer? Is this company in our ICP? Is this reply a "not now" or a "never"? Each needs judgment but produces a label, a score, or a yes/no, and the Decision API is built exactly for that routing at scale.
The LLM writes; the decision layer decides, and you only pay for writing on rows that cleared the bar.
Save this.
Come back when you're mapping your first GTM build.
Repost ♻️ to help a GTM operator in your network.
P.S. Which of the four moves first for you: Dots, the model, agents, or decisions?