Writing on the wall

I should start off by introducing this piece as an editorial. For those who know me, they know I am a technologist through-and-through and I wanted to share a few distinct opinions on where the world is moving.

1) Apple vs. NVidia = apple moving back to the boutique computer. As more workloads require CUDA cores to advance their ML models, developers will need CUDA support. Yes, you can run remoting debugging, but that's not the same as working on an ML model at your local coffee shop. The only saving grace on this front is Cloud GPU work loads. So, for all of those companies continuing to invest in extremely expensive workstations to make developers happy, consider switching back to Microsoft Windows or better yet Chrome OS. Both run Linux and both work seamlessly with Cloud providers at a fraction of the cost.

2) This is more of a wish than a road ahead, but I sincerely hope there are more ARM / RISC based processors coming to the Linux and Windows worlds. This is an area where Apple knocked it out of the park. I love sitting with a laptop that IS NOT doubling as an electric blanket.

3) Corporations that are still running in hybrid mode, support your associates by providing the best high-speed fiber internet possible. We have reached the phase where "The Network has become the computer", and the internet truly is the Frontside Bus of our modern infrastructure. If you're using LLMs or Stable Diffusion, your experimenting and learning pipelines require highspeed internet more than ever. For example, while working on a stable diffusion pipeline, for every model referenced the developer will pull down gigabytes of information on every run. Make sure they are working as efficiently as possible.

4) Lastly, ML and AI are changing the world. There are constraints, but they are in our control to manage. Stop computing as if it's the nineties and/or twenty-teens. Invest in your cloud infrastructure, decrease your machine and fleet management costs by moving to commodity operating systems, and triple down on investing in your resources to truly understand how to most effectively work in a cloud environment.

Decisions

• Stop buying Mac computers, choose Chromebooks (Running Chrome with Linux or Linux), it's a leap, but it's a move that will pay off. [Note: If you're an Apple development shop, this DOES NOT apply to you :) ]
• Develop in any language (Java, Go, Rust, Python, Node), stop trying to force one.
• Stop installing Office products and prefer online tooling (yes, they work in offline mode)
• Educate on cloud, learn about cloud provided virtual desktops and cloud TPU and GPU workloads
• Like it or not, if you're running ML workloads, you're going to need NVidia CUDA and/or supported TPUs if you and your company are going to get the most out of the next generation of compute.

Conclusion

These are exciting times, LLMs are NOT going to replace your development team anytime soon, nor are they going to rewrite your legacy technology stack. The best way to move forward is to follow the advice of one of my favorite retailers Mr. Sam Walton: "Control your expenses better than your competition", and "Swim upstream.".