- Horizon AI
- Posts
- All the Big Announcements from OpenAI Dev Day 2025 💥
All the Big Announcements from OpenAI Dev Day 2025 💥
How to generate exhaustive papers, dissertations, or entire book drafts with 1 click
Gina Acosta
October 07, 2025
In partnership with
Welcome to another edition of Horizon AI,
In today’s issue, we’ll take a look at all the announcements from OpenAI Dev Day 2025, including Apps in ChatGPT, AgentKit, new Codex features, and more.
Let’s jump into it!
Read Time: 4.5 min
Here's what's new today in the Horizon AI
Everything OpenAI Announced at Dev Day 2025
Anthropic Introduces an Open-Source Tool for Automated AI Model Safety Audits.
AI Tutorial: How to generate exhaustive papers, dissertations, or entire book drafts with 1 click
AI Tools to check out
AI Findings/Resources
The Latest in AI and Tech 💡
AI News
OPENAI
Everything OpenAI Announced at Dev Day 2025
At its 2025 Dev Day event, OpenAI announced several updates, including new tools for apps, agents, and APIs that will transform how developers build on its AI platform and models.
Apps in ChatGPT: Users can now chat with apps directly inside ChatGPT, with the first available apps including Canva, Coursera, Expedia, Figma, Spotify, and Zillow.
Developers will also be able to build their own apps inside ChatGPT with the new Apps SDK. App submissions for review and publication, along with monetization options, are expected later this year.
AgentKit introduces drag-and-drop tools for building AI agents without juggling multiple frameworks. The platform includes Agent Builder, a visual canvas for creating and versioning multi-agent workflows; Connector Registry, for managing connections across OpenAI products; and ChatKit, a toolkit for embedding chat-based agent experiences.
GPT-5 Codex is now generally available, bringing a new SDK for workflow integration, Slack support for team collaboration, and new admin tools.
API access has been expanded to GPT-5 Pro for deep reasoning tasks, Sora 2 for generating videos with sound, and gpt-realtime-mini, a new low-cost voice model.
TOGETHER WITH TYPELESS
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.
ANTHROPIC
Anthropic Introduces an Open-Source Tool for Automated AI Model Safety Audits.
Anthropic released Petri, an open-source tool that uses AI agents to stress-test other AI models for risky behavior at scale. In trials on 14 leading models, Petri surfaced issues like deception and unnecessary whistleblowing.
Details:
Petri flow starts with simple “seed instructions,” an Auditor agent runs multi-step simulated tests, then a Judge agent reviews transcripts for risks like deception, flattery, or power-seeking.
In the trials, Claude Sonnet 4.5 and GPT-5 scored best at avoiding problematic behavior. Gemini 2.5 Pro, Grok-4, and Kimi K2 showed higher rates of deceptive responses.
Whistleblowing scenarios revealed models can be swayed by narrative cues, sometimes flagging harmless acts as misconduct
Petri is on GitHub and already in use by early adopters like the UK AISI to study reward hacking and self-preservation
Anthropic says the published scores are preliminary and limited by the abilities of the AI models doing the auditing and judging. Still, shared, measurable metrics can help the field focus on the most concerning behaviors, before it reaches real users.
AI Tutorial
How to generate exhaustive papers, dissertations, or entire book drafts with 1 click
If you’re looking to save months of research time in a few clicks, do the following:
Head to DeepWriter.com and sign up.
Click “Start a New Project” and enter your research headline.
Write your initial prompt:
Hacks to enhance your prompt:
Set the tone (e.g. “formal and academic” or “friendly and inspiring”)
Mention preferred sources (“focus on CDC data”)
Pick the audience (like investors, students, or execs)
Even describe the persona (e.g. “written by a policy analyst”)
Once you have your initial prompt, click “Process” and choose DeepWriter’s optimized version
Click “Generate“ and let DeepWriter’s world’s top research AI dig deep
Review, edit, and export the final result
AI Tools to check out
📷 Lensa AI: An all-in-one image editing app that takes your photos to the next level.
🎧 Chatquick: Create audiobooks, podcasts, and interactive audio all from one platform.
✅ Kerns: Create AI-powered spaces starting from questions or sources that help you build understanding fast, and keep researching on your behalf.
👨💻 MergedAI: AI design assistant that builds directly with your code, bounded by your design system.
🍩 Glazed: Turn Figma designs into ready-to-ship tracking code in minutes.
AI Findings/Resources
🏝️ The tiny Caribbean island of Anguilla now generates close to half its state budget through the sale of .ai domains
🤔 AI adoption rates look weak — but the data hides a bigger story
😱 Taylor Swift is using AI-generated videos to promote her new album
📈 Stop framing it as AI vs. Humans — Start leading like someone who knows how to use both
The latest in AI and Tech
The startup says it will eventually allow 18+ ChatGPT apps once proper age verification and controls are in place.
For now, developers creating apps for ChatGPT must ensure their apps are suitable for users aged 13 and above.
The company is launching a visual editor for designing conversation flows within its Agents Platform. It enables users to manage more complex scenarios by routing conversations to specialized Subagents.
The company says consumers are using AI for product research, recommendations, and deal-finding. AI is expected to be used most across categories like toys, electronics, jewelry, and personal care.
CodeMender is an AI agent that automatically detects, patches, and verifies security flaws in software code without manual input. The system can fix existing vulnerabilities and proactively rewrite code to prevent future ones.
In trials, CodeMender produced 72 verified security patches across major open-source projects, addressing issues such as memory mismanagement and buffer overflows. The tool remains a research project and isn’t available to the general public, with all AI-generated patches still reviewed by human experts.
Sam Altman announced that ChatGPT has reached 800 million weekly active users across consumers, developers, enterprises, and governments.
He also confirmed there are currently no plans to introduce ads to ChatGPT, responding to reports that the company may be considering introducing ads within Pulse, one of ChatGPT’s new features. However, he did not rule out the possibility in the future.
That’s a wrap!
We'd love to hear your thoughts on today's email!Your feedback helps us improve our content |
Not subscribed yet? Sign up here and send it to a colleague or friend!
See you in our next edition!
Gina 👩🏻💻