Issue 337: Finding the eye in the AI storm

When new technology comes along, the hype can feel like a hurricane. One day we’re told to tokenmax and the next to try to save as many tokens as we can. Finding the eye in a hype storm takes real experience and expertise, which is why we’ve packed this Overflow with conversations with just such experts.

We sat down with Fireworks AI's Benny Chen to explore the open-source protocols setting the standard for evaluating AI applications, and chatted with Snowflake's Vivek Raghunathan about turning the chaos of AI-assisted coding into a structured, five-stage repeatable playbook. Plus, Yobi's Frank Portman joined us to explore why standard LLMs aren't cut out for predicting human intent—and how they are building a "foundation model of behavior" instead.

And AI isn’t the only place on the web that’s full of chaos that people are trying to cut through. From the web, we’ve got the stories of fixing memory leaks, the final frontier of cryptography, and why arguing with people may be fun but not good for your personal growth. And chaos may reign in technology nowadays but knowledge is still king—luckily, we’ve got plenty of that for you in this issue. For instance, what constitutes a full stop? Why won’t clients pay on time? Where do commas technically go? All of that chaos and knowledge is ready for you down below.