We had two big rival model releases this week - Claude Opus 4.6 from Anthropic and GPT-5.3 Codex from OpenAI. Both focused on coding. Both big deals for agentic AI. Anthropic also released their Super Bowl ad, taking a pop at OpenAI testing ads in ChatGPT, and Boston Dynamics released a fun Atlas video.
There’s a small number of Web 4.0 and Ethics News articles - Gemini has now surpassed 750m weekly users, very quickly closing the gap with ChatGPT. There’s also a report on the $100bn NVIDIA/OpenAI deal and the security challenges inherent in OpenClaw.
There are also a good couple of Long Reads to check out - Enjoy!
Anthropic releases Opus 4.6 with new ‘agent teams’
On Thursday Anthropic released the latest version of its flagship model, Opus, bumping it to 4.6, up from the 4.5 version which was only released in November last year.
The big focus of this update is on the model’s coding skills, especially as these skills are what are starting to unlock a lot of amazing agentic capabilities, including helping people with ‘everyday’ tasks in their Cowork application.
On the same day, OpenAI also introduced GPT-5.3 Codex. Again, the focus here is on coding capabilities, but OpenAI haven’t started applying this to helping people with ‘everyday’ tasks. I suspect we’ll see something big from OpenAI in this space in the next few months.
Mark Andreessen said in 2011 that "Software is eating the world". Now AI coding agents like Claude Code and Codex are eating software, and beginning to control the world by proxy.
Anthropic | TechCrunch | OpenAI
Anthropic says ‘Claude will remain ad-free,’ unlike ChatGPT
In Anthropic’s Super Bowl spot they have taken a pop at OpenAI for testing ads in ChatGPT. Sam Altman got exceptionally testy over this and the whole thing has highlighted that there is still a huge rivalry between the two companies.
For the last couple of years it has seemed like OpenAI were very focused on Google Deepmind as their main rival and Anthropic were happily focusing more on the enterprise market and catering specifically to developers with the coding capabilities of their models and Claude Code.
But as coding skills have become more central to agentic AI over the last 3 months the rivalry between OpenAI and Anthropic has really heated up, culminating in the joint releases of Opus 4.6 and GPT-5.3 Codex this week. It’s great to see the competition driving capabilities forward, and I hope they don’t get too distracted by taking pops at each other - there’s plenty of room for lots of different AI platforms in the future.
Atlas Airborne
The research version on Atlas has reached the end of the line, now that they’re heading into full production, so the researchers thought they’d really try and push the limits of what it could do. Pretty impressive!
Web 4.0
AI Ethics News
Long Reads
Stratechery - Microsoft and Software Survival
Artificial Ignorance - The Codex App Has Upended My Daily Workflow
Latent Space - Experts Have World Models. LLMs Have Word Models
“The future is already here, it’s just not evenly distributed.“
William Gibson