My Thoughts on AI: Week of June 12

Developer workspace comparing stable code execution with polished design review panels

I am not trying to crown one AI tool as the winner forever. These tools change too fast for that. What matters to me is more practical: when I sit down to do real work, which one helps me keep going, make fewer messy mistakes, and understand what happened when the session is over?

Codex and Claude Code both do serious coding work. They can read files, make changes, run commands, and help move a project forward. But they feel different. Codex feels more like the tool I use when I need to finish implementation. Claude feels stronger when the work needs taste, presentation, and design sense.

This is also just where I am right now. My opinion is based on GPT-5.5 in Codex and Opus 4.8 in Claude Code. Fable 5 also came out this week, on June 9, 2026, so even while I am writing this, the ground is already moving.

Codex Feels Better When the Job Has a Clear Finish Line

Codex works best for me when the task is concrete. Read the repo. Follow the project rules. Make the change. Run the check. Tell me what passed and what failed.

For that kind of work, Codex feels steady. But that same directness can feel too plain when the task needs design judgment. It can solve the ticket, but not always understand the vibe.

Claude Code Feels Better With Presentation and Design

I am seeing this right now while working on my wife's portfolio website. I am not really a designer type of person. She is a fashion designer, not a techie, so she looks at the result differently than I do.

When she looked at Codex's design work, she felt it was lazy and hodgepodge. It did not follow her comments closely enough, and it made up some design choices that did not match what she asked for. Claude's results felt more polished. It followed her design cues much closer. It did not feel stubborn or lazy in the same way.

Plans, Limits, and Model Releases May Change My Stack

The pricing and limits also matter. I am thinking about downgrading Codex Pro and adding Claude Pro, because I may not need one tool to do everything. I might switch back and forth depending on what OpenAI and Anthropic release next.

Fable 5 is part of that thought process. It came out the same week I was writing this, and I burned through my allotted amount in about 30 minutes. With Opus on the $20 Claude Pro plan, I have been able to finish some milestones in one sitting without hitting rate limits.

Codex limits have felt less aggressive to me, even though it is not the same tier as Claude Pro. I only ran out of Codex tokens once, and that was on a day when I had three implementations running at the same time throughout the day.

Where I Land Right Now

I do not think the answer is "Codex forever" or "Claude forever." For implementation-heavy work, Codex is still my default. For design-sensitive work, Claude is making a strong case.

My subscription mix may change as the models change. I may flip-flop. That is fine. The real lesson for me is to use the right tool for the right kind of work, instead of forcing one assistant to be good at everything.

References

Comments

Post a Comment

Popular posts from this blog

Running CockroachDB with Docker Compose and Minio, Part 2

This is my second post on creating a multi-service architecture with docker-compose. This is meant to be a learning exercise and typically docker-compose is used to set up a local development environment rather than a production-ready set up. I regularly, find myself building these environments to reproduce customer bugs. For a production-specific application, refer to your platform vendor documentation. At some later time, I will cover Kubernetes deployments that can be used as a stepping stone for a real-world application. Until then, let's focus on the task at hand. We're building a microservice architecture with CockroachDB writing changes in real-time to an S3 bucket in JSON format. S3 bucket is served by a service called Minio. It can act like an S3 appliance on premise or serve as a local gateway to your cloud storage. Let's dig in:
Read more

VirtualBox options to start VM in Normal, Detached and Headless Modes

Image
I've been an avid user of VirtualBox for a long time! Much of what I've learned in my career had something to do with VirtualBox. I do all my testing and new tutorials in VirtualBox, I use Hortonworks Sandbox VirtualBox image for my work. This is probably not a new feature but it is definitely new to me because I just noticed it, VirtualBox has an option to run in different modes, at least from the GUI perspective. I know there's a lot of untapped potential in using the VBox CLI but I haven't had much need for it yet. Back to the topic, you can now launch your VMs in headless and detached modes right from the GUI! Usually, when I launch a VM in normal way, the VM window just hangs there and is pretty annoying. I prefer to SSH to a running VM from a console instead of using the VM window. Now I can launch a VM in headless mode and it's business as usual. No extra windows, and I bet the memory toll on my machine is also smaller. When you click on start your VM, you no...
Read more

Digsby is bringing out a Linux and Mac client very soon

Hey all, my new favorite multi-protocol IM client is soon to become a multi-platform client. It is still in beta and they're already rolling out a client for the other two platforms. It is a great news to me because I literally fell in love with Digsby on Windows and every time I find myself using Kopete or Pidgin on Linux I yearn for Digsby. It is just so great at what it does. Hope the Linux client is not crippled and offers just as many features as Windows. Sign up for notification here . p.s. Ars did a nice and simple review of Digsby here . Talk to you soon.
Read more