You can now personalize the Codex app with themes that match your taste. Font, colours, contrasts, sidebars and more can now all be tampered to make your Codex feel like yours exclusively.

There is also the ability to use ready made templates such as Absolutely, which surprisingly resembles Codex’s fierce rival. Claude.

Codex also released a feature that allows it to autonomously act when certain requirements are met. Such as a timer going off, a Slack message sent or a PR pushed. They called this feature “Automations”.

A larger context window means that models are able to view & incorporate more of your codebase and grasp a more coherent understanding than before.

The graph that Anthropic posted (below) shows that Claude gets only 13.6% and 25.5% worse when using the entire context window.

To put these results into perspective, GPT 5.4 gets 42.7% worse and Gemini degrades 61%, which makes Claude’s results particularly impressive.

This matters because real-world coding often requires models to reasons across large repositories, multiple folders and long chains of dependencies. If performance collapses as context window fills up, the generated code can worsen. Claude maintaining a relatively stable performance indicates that it shall excel with larger projects, comparitively.

It can be overwhelming can’t it?

Anthropic releasing features faster than you can try them out. Fellow AI coders making tens of apps a month. Free credits being given out everywhere. New models that supposedly crush benchmarks being released every single week.

There is so much noise out there, that it is easy getting caught up in it all, especially when the noise sounds exciting.

How often have you abandoned a project just a week after you couldn’t sleep as it excited you so much? Only to jump to the next idea…which also led nowhere.

Shiny object syndrom is real. Shinnier than ever before in this golden age of software development, where apps have become a few prompts away. Yet if we want to build something valuable, something big, then we have to resist it.

Now, I am not saying that we aren’t allowed to experiment, fiddle and play. In fact, I ENCOURAGE you to. There is a difference though in learning and experimenting versus being endlessly distracted from your goals.

Whether your goals are building a unicorn, a $20k/month saas or a side project that only you will ever use it doesn’t matter. Yet all these goals require focus, consistency and persistence.

You got this.

I have been chronically on Reddit and X where I couldn’t help but stumble upon a TREASURE TROVE of projects that awesome AI coders have cooked up. There are so many that I am going to have to cover more in future newsletters, enjoy!

Now, onto the timeline.

We have a VERY exciting edition for our next newsletter, until then!