What I brought home from OpenAI DevDay

On the drive home from the airport after OpenAI DevDay, I spent about 45 minutes on a phone call with AVA, my dot. Before I got home, I had planned the next day and moved work forward across several projects.

That was when it really clicked for me. It rings like an actual phone call. You answer and talk through the work. It feels familiar, which makes it much easier to use.

Thank you, OpenAI, for flying me to San Francisco and having me be part of DevDay 2026. It was an extremely rewarding experience. Thank you to everyone I met at DevDay, at the surrounding events, and over coffee and drinks.

Walking into DevDay at Fort Mason

More than two weeks with AVA

I used to keep all my AI work organized in Projects. Everything had its place before I started.

Now I can call AVA while I’m in the car or walking my dog and talk through what I’m working on. That drive home from the airport showed me how much I could get done in one call.

For me, removing friction is the single most important thing for AI adoption. Being able to pick up the phone and work with AVA has made that clear.

Dots are OpenAI’s always-on agents, powered by GPT-6 Astra. You give your dot a goal and connect the apps it needs. I can return to AVA and talk through work across my projects without figuring out where to start the conversation.

We also got to design our dots at DevDay, and OpenAI printed custom stickers of them. That’s AVA wearing headphones in the top left of my MacBook Air in the banner photo.

The most valuable conversation of the trip

My go-to-market discussion with Dr. Ulrich Keller was the most valuable time I spent there. He was amazing, generous with his time, and is helping connect me with OpenAI’s partner program and people on their go-to-market team.

We’ve made a bet on OpenAI at Digital Meld, so I appreciated being able to talk with him about our work and where we go from here. Thank you, Dr. Keller.

The sessions I went to

I recorded my sessions and photos in my trip timeline. Here’s what I went to and how I see it applying to our work at Digital Meld.

Brad at the DevDay keynote, alongside a view of the audience

Opening Keynote

There was a lot to take in at the keynote: dots, GPT-6.1 Sol, collaboration, developer tools, and integrations. I’m interested in how we can put them together for customers, from discussing a task with an agent to sharing the result with a team or building an app around it.

Brad’s check-in photo for What Comes After Loops

What Comes After Loops?: Deep Dive Into Agentic Architecture

I went to Raj Pathak and Eric Provencher’s session on agent architecture. For our customer work, I see this applying to tasks that pass between people and agents. Researching a request, preparing a response, and reviewing it all need the same brief and a clear handoff.

Brad’s check-in photo for Harness-as-a-Service

Harness-as-a-Service: Building with the Agents API

Steve Coffey and Charlie Guo’s session covered the Agents API. Its computer-use capabilities give us another way to connect work across the applications a customer already has. That could save us from having to build a separate interface for each step.

Brad’s check-in photo for Launchpad

Launchpad: Go-to-market

This was the small-table discussion with Dr. Ulrich Keller I mentioned above. His help connecting me with the partner program and go-to-market team made it the most valuable time of the trip.

Brad’s check-in photo for the Higgsfield session

Higgsfield: Building Agents That Provision Their Own Compute

Craig Dennis and Alex Mashrabov’s session covered agents choosing models and provisioning compute, including throughput and failure modes. For a customer running media-processing jobs, I would apply that to choosing capacity, controlling spending, and recovering from failed runs.

The session slide showing Teams alongside Slack

Codex Has Left The Laptop

I posted a photo of a slide showing Microsoft Teams alongside Slack. My reaction at the time was that OpenAI should be shouting about Teams support from the rooftops. We work with Microsoft customers, and having AI available in a tool their teams already use removes a step. The session covered Codex across the CLI, app, and web.

Brad’s check-in photo for From Single Player to Multiplayer with Codex

From Single Player to Multiplayer with Codex

Jason Liu’s session followed an overnight outage from the first Slack message to a reviewed fix and a team update. That is relevant to the support and maintenance work we do for customers. An investigation needs to end with both a fix and an explanation for the people affected.

Brad’s check-in photo for the closing Q&A

Closing Q&A

I ended the day’s sessions with Sam Altman and Thibault Sottiaux’s Q&A, then headed to the reception at Fort Mason.

What the announcements mean for businesses

GPT-6.1 Sol improves coding, computer use, and professional work at lower standard token prices than Astra. Sol Ultrafast followed after DevDay. I want to see where the combination of capability, price, and speed fits the tasks we run for customers.

ChatGPT Space and Pages let teammates work on shared documents with AI. Sites are available on eligible Business and Enterprise plans, so a team can describe an app it needs, try it, make changes, and share it.

Plugin extensions add interactive panels and other interfaces to ChatGPT. Being able to create plugins and tools as a need comes up is really freaking cool. I can see this helping with the small business apps that people want but rarely have time to build.

Here are the 25 announcements from OpenAI’s DevDay recap:

DevDay announcement guide

Dots: Takes on ongoing work in the apps you connect.

GPT-6.1 Sol: Improves coding, computer use, and professional work at lower standard token prices than Astra.

Ultrafast: Speeds up responses from Astra and, following the October rollout, GPT-6.1 Sol.

Private Intelligence: Adds privacy options for businesses, including a preview of Private Inference.

Codex in the cloud: Runs coding tasks in the cloud so work can continue across devices.

Refreshed Codex CLI: Adds voice input and a view for managing agents in the terminal.

Code Review: Lets developers inspect changes and discuss findings with Codex.

Codex Security Cloud: Scans repositories, investigates vulnerabilities, and prepares proposed fixes.

Decisions API: Classifies inputs and selects from defined answers; now in public beta.

Agents API with computer use: Lets agents interact with existing software to complete tasks.

Bedrock Managed Agents powered by OpenAI: Runs OpenAI agents within AWS.

Plugin extensions: Adds interactive panels, sidebar views, and file viewers to ChatGPT.

Improved plugin creation, submission, and discovery: Helps developers build plugins and people find them.

Sites hosting plugins: Lets teammates use plugins in a shared Site with their own connected accounts.

MCP events for automations: Starts an automation when something changes in a connected app.

ChatGPT Space: Brings a team’s documents and related work into a shared space.

Pages: Lets teammates and AI write, research, and edit a document together.

Collaborative slides: Announced shared slide editing for teammates and agents.

Teams and shared tasks: Lets teams share instructions and delegate recurring tasks.

@ChatGPT in Slack and Microsoft Teams: Lets people ask ChatGPT for help within a team conversation.

Meetings plugin: Saves meeting notes and action items in Space.

Shareable profiles: Collects Sites and plugins people can share and reuse.

Sign in with ChatGPT: Lets people use their ChatGPT account with participating tools.

New Pro tier: Adds Pro 500 with higher usage allowances and Ultrafast access.

OpenAI Marketplace: Lets eligible enterprise customers apply part of their OpenAI commitment to approved partner software.

They have kept shipping since I got home

GPT-6 with Intelligent UI adds visual and interactive answers. The October 9 dots update lets people create a dot from their phone and delegate work to Codex. Since I got home, OpenAI has also rolled out Sol Ultrafast and moved the Decisions API into public beta.

Other updates include audio uploads, document scanning, composer predictions, and GPT-Live-1.1 with new voices. These are smaller changes, but being able to upload a recording or scan several pages of notes saves people steps before they even start asking for help.

After meeting so many people on the team, the pace makes more sense to me. They were passionate about what they were building and dedicated to making it useful. I appreciated getting to spend time with them.

Why we are investing in this

Digital Meld is an OpenAI Partner Network partner. I hold the ChatGPT Expert, Codex Expert, and OpenAI API Expert certifications, and I have submitted my application to the Codex Ambassador Program.

I already share my work through my AI Dev Days repository and talks. As I said on Twitter, I hadn’t realized I was already kind of doing it. I’d like to keep doing that through the program.

Thank you again to OpenAI for flying me out and having me there, to everyone who made time for coffee or drinks, and to Dr. Ulrich Keller for his help.