Ben Newell

I've spent years coaching, studying exercise science, and building software. ImprvHealth 2.0 is my attempt to bring all of that together into one thing: An agentic sports scientist in your pocket. This is the third major Codex goal I've run for the project. All of them have gone 10+ hours. This one is currently ~70% complete. Excited to show people what's coming. Thanks @thsottiaux and @OpenAIDevs

Everyone is focused on the Agentic Engineer. I’m starting to think the role that becomes even more important is the Agentic Technical Program Manager. As agents get better at writing code, generating tests, reviewing PRs, and completing bounded tasks, the bottleneck shifts. Someone still has to coordinate work across teams, manage context, validate outputs, own release readiness, track dependencies and risk, create proof gates, and decide what ships and what doesn’t. In many ways, AI increases the need for strong execution and orchestration. Curious what others think. Which role becomes most important in an AI-native engineering organization? - Agentic Engineer - Agentic Technical Program Manager - Product Manager - Engineering Manager Drop a comment below 👇

@rezoundous technical program management is needed more than ever now

We're no longer vibe coding. We're agentic engineering now.

Opening my gardening project in ChatGPT to see the new image capture improvements - this is a really cool upgrade.

Hey @OpenAI where on the merch store can I get this table.

@thekitze @madsmadsdk 😂

@madsmadsdk just sell the car if u have that sticker

Enjoying a nice weekend trip. Just me, my boy and tranquility at my moms place 🫶

@jxnlco @MichaelArnaldi domain expertise

@MichaelArnaldi Llms could learn a lot from me

Amazing quality of response!

@MTSlive the skateboarding on is iconic

SITUATION DETECTED: OpenAI has made the first ever beautiful tech billboards.

This looks super interesting

@thdxr common courtesy?

who ever needs a receipt why is the default to ask if you need a receipt

Is walking underrated as a real workout?

@clarefinds go door to door

You have $0 for marketing. Your product just launched. How will you get users?

@gabriel1 what if it wasn’t a dream?

woke up sweating dreaming i started a mildly successful company 2018, being a stuck saas founder ceo watching everyone around me participating in what feels like sci-fi

What combination of AI, exercise, recovery, and engineering practices produces the highest-performing human?

@petergyang It would be cool if codex could tell you the probability of a successful reconnect. Sometimes i'm not sure if i should wait another 5-10 minutes or just start a new thread and move on.

This is probably the most annoying part of using Codex

@mvanhorn @petergyang you made it

@petergyang omg

He doesn't have a CS degree, doesn't read code, and doesn't even look at AI's plans most of the time. Yet @mvanhorn has built one of the most impressive open source track records I've seen. My next episode is for every non-technical person who has doubts that they can build with AI agents. 📌 Subscribe to see how Matt does it tomorrow: @PeterYangYT?sub_confirmation=1" target="_blank" rel="nofollow noopener">youtube.com/@PeterYangYT?s…

@PolymarketMoney @egorkabantsov More motivation to create a better model

JUST IN: Andrej Karpathy, a top AI scientist at Anthropic, is reportedly barred from accessing the company’s most advanced AI model because he is not a U.S. citizen.

@thekitze bro I have been saying this for weeks. I see why you have the 🐐in your handle

actually can we like ban all the models so we can all chill a bit

@morganlinton I’ve been wanting to get into local models - can you recommend any to start with?

@ben_ai_eng Yes, and I think it’s very likely. And for a lot of routine tasks, you don’t need frontier-level chain of thought, there’s a lot you can, and likely will do, with local models.

Soooo, I can't stop thinking about the fact that I was able to get the same results with Fable on low effort, as I was with Opus 4.8 and GPT 5.5 on high and xhigh, with so many different test that I ran last week. Now I can't get it out of my head, woke up, wrote this.

@petergyang @mvanhorn impressive