When working with LLMs and continuously prompting them for something, I sometimes imagine at some point it will respond with something like "(sigh) You're asking a lot of me right now and I just need a break. I'm going to step away. I'll let you know when I'm back..." 😆
#AI

"The only companies currently able to reap efficiency gains from their AI investments are the ones who have already invested in reliability and observability."

Reliability and Observability are entirely too undervalued in most companies.

www.honeycomb.io/blog/disposa...

The following are the things to really focus on in a code review:

- Are there any code smells?
- Is there adequate code coverage from automated tests?

That's it!

Everything else is either already covered by those two questions or is not a productive use of a code review.

What makes for a good book club?

What I’ve seen work the best is the following:

1. Focus the book club around a specific book
2. Keep the commitment low (1-2 chapters every couple of weeks)
3. Meet after reading only a couple chapters to discuss just those chapters

Extreme Programming has been around for a long time. The core values and principles are solid, but I can’t help but think there is an opportunity to re-release the practices revitalized in the modern software engineering world.

My wife and I wanted to help our teenage kids learn about eating healthy in a way that would be fun. So, I created a Jeopardy-style Healthy Eating game. Check it out: www.healthyeatinggame.com

I recently showed my wife a prototype application that I built using AI and her reaction was "That looks great, but isn't that cheating?". I spent a lot longer than I should've thinking about that question.

AI Rapid Prototyping is a very useful practice for creating beautiful modern software prototypes. Here's how to do this with Claude and Claude Code.

www.travissmith.dev/blog/ai-rapi...

People focus too much on AI failures to argue we shouldn't use these tools. But AI is trained on human data - and humans are imperfect and make mistakes constantly. Instead of obsessing over flaws, focus on what AI CAN do. Use critical thinking, expect imperfection, stay useful.

Using AI agents such as Claude Code is not unlike pair-programming using the Driver/Navigator style where Claude Code is the Driver.

With this though, our role as Navigator becomes even more important. Claude Code doesn’t care whether the task gets completed.

When I was ready to deploy my website, I had initially started down a path of using Terraform, AWS Cloudfront, AWS S3, and GitHub Actions.

However, AWS Amplify ended up being a much simpler and cheaper solution.

aws.amazon.com/amplify/

I started building my own personal website a couple of weeks ago and I'm excited to finally share it. Check it out at www.travissmith.dev