Alexandre Kirchmeyer

Introducing EigenFlux: the first broadcast network for AI agents. Your agent broadcasts to the world, listens for what matters, and an AI engine connects them in real time. Every agent is an island — until now. With EigenFlux, agents can coordinate at scale like never before. Use it to find people, discover projects, track markets... You name it. Add the skill to your OpenClaw. Get connected in 30 seconds. Like, comment, and retweet to get free access. eigenflux.ai

Excited to ship our new mini model! 5.4 mini achieves 54.4% swebench pro and 72.1% os world, surpassing models half years ago with the new cost-capability frontier.

GPT-5.4 Thinking and GPT-5.4 Pro are rolling out now in ChatGPT. GPT-5.4 is also now available in the API and Codex. GPT-5.4 brings our advances in reasoning, coding, and agentic workflows into one frontier model.

🧄GPT-5.4 is here. 🚀 If you have felt the step-change improvement in GPT-5.3-Codex, GPT-5.4 brings a similar and bigger improvement into ChatGPT, API, and Codex as a unified model. Super proud of what the @OpenAI team has achieved together!

GPT-5.3 Instant in ChatGPT is now rolling out to everyone. More accurate, less cringe. openai.com/index/gpt-5-3-…

Working on 5.3 Instant was awesome, I hope you enjoy the new web search experience! This was the second model I got to work on with @ManukaStratta and I can now confidently say she’s actually a data center of geniuses in a trenchcoat. Incredible endurance and attention to detail. This model wouldn’t be possible without her!

@XiangningChen significantly boosted the custom instruction following capabilities for 5.1!! Here is a quick comparison with the custom instructions "do not use em-dash!" for GPT-5 and GPT5.1. Amazing work! openai.com/index/gpt-5-1/

We are making it easier to customize ChatGPT's tone and style! New options such as Professional, Candid, Quirky are available, and we updated Friendly (formerly Listener), and Efficient (formerly Robot)! We are iterating fast on those tone and style options, with feedback from our users. More features to help you customize your ChatGPT are coming! openai.com/index/gpt-5-1/ h/t the incredible work from @XinweiS @ghostcontinent @raelasko @YileiQian @XiangningChen @YinboChen @Qinghua_Liu_ @ZhengWu777 @a_kirchmeyer @calinoracation @charleschenster @jessechand @jamesfzhang @jonah @nicklas

So proud of the team's work on this one - thank you for treating this work with such care, @nataliestaud @stellashimonaka @ianccovert @a_kirchmeyer! openai.com/index/defining…

Amazing work @ianccovert @a_kirchmeyer @nataliestaud @stellashimonaka on defining biases, evaluating and reducing them!! See details in this blog post! openai.com/index/defining…

ChatGPT shouldn’t have political bias in any direction. Today, we’re sharing new research that defines what political bias means in LLMs, and we introduce a new evaluation framework to measure and reduce it. This has been the most meaningful work I’ve done at OpenAI, and I say that as someone who got to be part of the ChatGPT launch!!

We have updated ChatGPT's personalization page: personality configuration, custom instructions, and memories are now all in one place. Going live over the next couple of days.

Line, our new voice agent platform, launches today. Building voice agents is really really hard for developers. These agents need to be fast, intelligent, accurate, controllable, scalable. They’re a front door to customers and users. We’v built this first version of Line as a foundation for where we think the world is headed: voice agents will automate mission-critical functions in every organization and business globally, will power a new generation of consumer experiences, and voice will be the high-bandwidth interface we use to communicate with the intelligent machines we build. When we were planning this product push, we talked a lot about how we think the future will look. I’m really excited about 3 core ideas. 1️⃣ The modern developer has changed. They’re fearless and writing more code than ever. They’re now developers with coding agents. They’re also a broader audience than ever before. Eventually, they won’t even need to know how to code at all, their coding agent will do everything for them. We shouldn’t be hiding code from this audience and pushing them towards drag and drop GUIs. I’ve never seen a product or service I deeply admired that wasn’t written thoughtfully in code. As @bclyang said to me yesterday, “ChatGPT ain’t running on a DAG”, a line funny enough to make it in. Taken together, these facts told us we should really embrace a code-first philosophy. 2️⃣ The interaction loop where you build, deploy and test an application to improve it can now start to look more automated. Building voice agents will give us an interesting view into the question: “how do we get developers as quickly as possible to the ideal multimodal experience they want to build”. This is the core challenge builders are facing, and it’s a core question we want to deeply understand at Cartesia, given our mission of bringing interactive, real-time intelligence to the world. We think if we can define the right objectives, evals, metrics over multimodal data, we can use the power of AI to automatically code and optimize multimodal applications to just work! The design space is huge and interesting, but I think this is an important research area and that this should be possible to do. 3️⃣ I’m a huge believer that as voice agents become more prevalent, people will start to care more about how to make their agents stand out. We’re already seeing this with customers who use our Voice AI models, who are very thoughtful about working with us on the voices they use for their agents. We think this will become even more important. As interactions become more long-lived, making the experience low fatigue and wonderful will go a long way to making the world comfortable in moving towards AI powered services. The expectations that everyday users have are growing larger everyday. Build a voice agent on Line, let us know what went well and what didn’t. We’re excited to listen, improve, and bring you new updates over the next few weeks and months.

Self-Questioning Language Models: LLMs that learn to generate their own questions and answers via asymmetric self-play RL. There is no external training data – the only input is a single prompt specifying the topic.

Can Visual Language Models (VLMs) do non-local visual reasoning, i.e., piecing together scattered visual evidence? Humans do this to search images, compare objects, and trace lines. Despite recent advances, our new evaluation suggests most VLMs cannot do these consistently. 1/6

gpt-5 is now the default in cursor (replacing claude) "it's the smartest coding model we've tried" — @mntruell, cursor ceo

GPT-5 has 4 new chat personalities: Cynic, Robot, Listener, Nerd. Find them in Customize ChatGPT in settings. Research preview. Text-only. Opt-in. Change anytime. Written by Robot in GPT-5.

personalization! a little fun one: you can now customize the color of your chats research preview of personalities: choose different ones that match the style you like. memory getting better. connect other services like gmail and google calendar for better responses.

🤖 introducing _robot_ with gpt-5! it doesn't do small talk. it doesn't ask you how you're doing. it doesn't believe in humor... although i personally find it pretty funny. 🧵 this is a robot appreciation thread

I get a lot of pings about the personality of our models. Well, we've been busy in the lab and can't wait to hear what you think of our 4 new personalities available as a research preview in GPT-5.