Matt Asay

Short answer: None of them. You’re better off starting with the least-expensive model that can do the job. The reliability of your own company's data, workflows, and integrations plays a far bigger role than the model you choose.

The reliability of your company's data, workflows, and integrations plays a far bigger role than the model you choose.

When a tech company gives away a successful product, you can be sure they're seeking profit from another part of the stack.

When a tech company gives away a successful product, you can be sure they’re seeking profit from another part of the stack.

When a client overpaid for AI compute and then had nowhere to resell it, I realized the cloud resale market had quietly disappeared.

Open source software gives us a paradigm for open-weight models, reducing dependence on the model providers and expanding what application companies can build.

Even hardware companies like Nvidia know they can sell more products to more people if they give software developers what they want.

Install Python package dependencies without the packages, learn the ins and outs of making standalone Python apps, and install Python on Windows the easy way.

AI makes it too easy to create slide decks, reports, blog posts, and software that nobody cares about and nobody needs. Use it when it matters.

Windows 3.1, modems, CompuServe, AOL, and bulletin board systems made writing software fun. You could even earn a little profit.