In partnership with

Greetings from above,

OpenAI just dropped their biggest update of 2025, and if you're building anything with AI, this changes your entire playbook.

Yesterday at DevDay in San Francisco, they announced six major features that went from "coming soon" to "available right now."

We're talking apps inside ChatGPT, visual AI agent builders, and APIs that give you capabilities that didn't exist last week.

I watched the whole thing so you don't have to. Here's what actually matters.

Today, we'll talk about:

📱 Apps running inside ChatGPT conversations (7 already live)

🛠️ AgentKit - build AI agents without writing code

💻 Codex officially launched with Slack integration

🎥 Sora 2 video generation API for developers

🎯 GPT-5 Pro and cheaper voice models in the API

Let's dive in!

Typing is a thing of the past

Typeless turns your raw, unfiltered voice into beautifully polished writing - in real time.

It works like magic, feels like cheating, and allows your thoughts to flow more freely than ever before.

With Typeless, you become more creative. More inspired. And more in-tune with your own ideas.

Your voice is your strength. Typeless turns it into a superpower.

📱 ANNOUNCEMENT #1: APPS IN CHATGPT

ChatGPT isn't just a chatbot anymore. It's now a platform running full applications inside conversations.

What this means:

You say "Spotify, make me a focus playlist" and the actual Spotify app appears in your chat. You're not redirected. You're not opening another tab. You're using Spotify right there.

7 apps launched today:

  • Booking.com - search and book hotels

  • Canva - create slide decks from outlines

  • Coursera - watch courses with AI explanations

  • Expedia - plan and book entire trips

  • Figma - convert sketches to diagrams

  • Spotify - build custom playlists

  • Zillow - browse homes on interactive maps

For developers:

The Apps SDK is in preview. You can build apps that reach 800 million weekly ChatGPT users without dealing with app stores, ads, or complex distribution.

It's built on Model Context Protocol (MCP) - an open standard. Your app can tell ChatGPT what the user is looking at, making conversations feel natural instead of repetitive.

Submissions open later this year with monetization options including instant commerce inside ChatGPT.

🛠️ ANNOUNCEMENT #2: AGENTKIT

The complete platform for building AI agents. The breakthrough? You don't need to code.

Four components:

1. Agent Builder - Visual drag-and-drop interface

  • Drag boxes (agents, tools, logic gates, guardrails) onto a canvas

  • Connect them to create workflows

  • Someone built a working agent on stage in 8 minutes

  • Companies like Ramp cut development time by 70%

2. Connector Registry - One admin panel for all data sources

  • Connect Google Drive, Dropbox, SharePoint, Teams, OneDrive once

  • Available everywhere across your organization

  • No rebuilding the same connections for every team

3. ChatKit - Pre-built chat interface

  • Drop into your app, customize colors and branding

  • Canva built a support agent in under an hour

  • Handles streaming, history, rich content automatically

4. Evaluation Tools - Measure and improve performance

  • Trace every decision step-by-step

  • Create test datasets

  • Automated prompt optimization

  • Test with models from different providers

Plus: Open-source Guardrails

  • PII masking

  • Jailbreak detection

  • Content moderation

  • Custom rules for your use case

Real results:

  • Albertsons: 2,000+ stores using agents for instant sales analysis

  • HubSpot: Built Breeze support agent

  • LY Corporation: Built work assistant in 2 hours

Available now for Plus, Pro, Business, Education, and Enterprise users.

💻 ANNOUNCEMENT #3: CODEX GOES OFFICIAL

Codex, the AI coding assistant, moved from research preview to general availability. Plus three major updates.

What Codex does:

Works in your code editor, terminal, GitHub, and cloud. OpenAI's engineers now merge 70% more pull requests per week because of it.

Three new features:

1. Slack Integration

  • Tag @Codex in any thread

  • Ask it to fix code or add features

  • Reviews the full context

  • Posts a link to review changes

  • Coding help where your team already works

2. Codex SDK

  • Use Codex in automation and workflows

  • Set up GitHub Actions for automatic code review

  • Build custom development tools

  • Available in TypeScript (more languages coming)

3. Admin Tools for Companies

  • Edit or delete cloud environments

  • Set configuration rules

  • Monitor actions and usage

  • View analytics on code review quality

The model: GPT-5-Codex trained specifically for coding. Better at refactoring and reviews. Thinks longer on harder problems.

Note: Starting October 20, Codex Cloud counts against usage limits. Check your plan pricing.

🎥 ANNOUNCEMENT #4: SORA 2 VIDEO API

Generate videos programmatically. Already in the Sora app, now available via API.

Two versions:

  • sora-2: Faster, cheaper, perfect for testing

  • sora-2-pro: Slower, pricier, higher quality for production

Key features:

  • Synchronized audio included (dialogue, sound effects, background)

  • Upload images as starting frames

  • Remix existing videos ("make the sky purple")

  • Videos up to 10 seconds

  • Control resolution and aspect ratio

Five endpoints: create, get status, download, list, delete

Restrictions:

  • No real people or celebrities

  • No copyrighted characters or music

  • No inappropriate content

  • No faces (for now)

Real use: Mattel animates toy prototypes before building physical versions.

Currently in preview.

🎯 ANNOUNCEMENT #5: GPT-5 PRO IN THE API

OpenAI's smartest model now available to all developers.

When to use it:

  • Financial analysis requiring perfect accuracy

  • Legal document review

  • Healthcare applications

  • Complex reasoning where mistakes aren't acceptable

When NOT to use it:

  • Simple tasks

  • Speed matters more than perfection

  • High-volume processing where cost adds up

Thinks longer, shows reasoning, catches its own mistakes.

Costs more, takes longer, but delivers better answers.

You must be really rich to use it in API though, it’s EXTREMELY expensive.

Available in the API now.

🔊 ANNOUNCEMENT #6: GPT-REALTIME-MINI

New voice model. Same quality as the large realtime model, 70% cheaper.

Features:

  • Speech-to-speech (no text conversion)

  • Natural conversation with emotion

  • Handles interruptions

  • Multiple languages

  • Low latency

Perfect for:

  • Voice assistants

  • Customer service bots

  • Accessibility tools

  • Educational apps

Available in the API right now.

📦 WRAP UP

What you learned today:

ChatGPT became a platform - apps run inside conversations reaching 800M weekly users

AgentKit lets you build AI agents visually without coding

Codex is official with Slack integration and SDK for automation

Sora 2 API generates videos with synchronized audio programmatically

GPT-5 Pro and cheaper voice models expand what's possible via API

The bottom line:

These aren't future features. They're available right now. Your competitors might already be using them.

The gap between "AI user" and "AI builder" just got way smaller. AgentKit's visual builder means you don't need a dev team to automate your work. The Apps SDK means you can reach 800 million users without traditional distribution headaches.

What changed:

  • Platform grew from 100M to 800M weekly users in 2 years

  • Token processing increased 20x

  • Everything announced is built for production scale

📚 WANT THE FULL BREAKDOWN?

I wrote a complete guide covering every detail, use case, and technical specification from DevDay 2025.

You'll get:

  • Detailed technical specs for each announcement

  • Real company case studies and results

  • Step-by-step guides for using each feature

  • Code examples and implementation tips

  • Everything you need to start building today

And as always, thanks for being a part of my lovely community,

Login or Subscribe to participate

Keep learning,

🔑 Alex from God of Prompt

Reply

or to participate

Keep Reading

No posts found