OpenAI launches Codex: An 'Software Engineer' agent that runs parallel tasks

Ewan MacLeod

16 May 2025 — 1 min read

Just the other day I wrote a long post about Anthropic's Claude Code (triggered by a Jeff Dean Q&A) and I noted that OpenAI's Codex CLI equivalent was pretty limited in comparison (so far, anyway!)

With the noise building about what Google might bring to market soon, OpenAI just announced their latest release in the form of Codex.

I was hoping they might bring a rather exciting update to the Codex CLI system to bring even more competition for the Anthropic team. But no. The OpenAI team has taken a slightly different approach, launching a, "cloud-based software engineering agent that can work on many tasks in parallel, powered by codex-1. Available to ChatGPT Pro, Team, and Enterprise users today, and Plus users soon."

I really like the lateral thinking.

Instead of a system sitting on your desktop or server, they're approaching things another way. The Codex 'agent' will access your code from (for example) your github repository and then act accordingly in terms of discussing, explaining, modifying and adding code to it.

The idea that it could do multiple tasks in parallel is extremely exciting – and that, I would imagine, is the base level thinking behind having Codex work in a cloud environment. Once it knows what to do, it can spin up many instances to rapidly complete tasks at-the-same-time.

Right now, me and my Claude Code buddies (I have 5 open at the moment) are all working sequentially – on one thing at a time. Parallel would be awesome. Indeed, one of the reasons I have multiple Claude Code instances operating is because it's all sequential.

Count me in as highly intrigued.

I will endeavour to check it out as soon as I can.

Meantime, here is the live stream replay - it's about 25 minutes if you'd like to watch the intro from the team:

Fiverr's CEO Micha Kaufman's "AI Letter" to employees

I wanted to make sure I documented this here, especially following the awesome Q&A from Fiverr's Sapir Hadad (their Executive Director of AI) that I published earlier this week. About a month ago, Micha Kaufman, the CEO of Fiverr, sent an open letter to Fiverr employees

Eric Schmidt: "The AI revolution is underhyped"

I always enjoy listening to what Eric Schmidt has to say about the AI world – and he's back again, this time in a 25-minute TED conversation published today. I'd recommend sitting and watching what he has to say, especially in the context of Conversational AI and

Google's Jeff Dean: We're 'about a year-ish' from having 24/7 AI engineers

Sequoia Capital held their AI Ascent event earlier this month and it included a session with Google's top man in AI, Jeff Dean. The whole event is captured in quite a few YouTube videos but this one stood out for me because of one of the questions Jeff

Stripe's ML team used a transformer approach on payment transactions to double fraud detection overnight

I came across this seriously fascinating post from Gautam over at Stripe. He is the Applied Machine Learning leader at the company. Too often, organisations hire ML/AI people such as Gautam and they essentially get nothing done. They're often stuck in proof-of-concept hell, with little support from

Read more

Fiverr's CEO Micha Kaufman's "AI Letter" to employees

Eric Schmidt: "The AI revolution is underhyped"

Google's Jeff Dean: We're 'about a year-ish' from having 24/7 AI engineers

Stripe's ML team used a transformer approach on payment transactions to double fraud detection overnight