A friend of mine recently started working for an EU-based software security company. He has a good handle on the AI game, but you can’t just plug and play whatever tool you wanna use in that environment (rules, regulations and stuff, you know). He got his Claude Code license about a month after his first day at work. In the meantime, he shipped code to production by writing it with the keyboard, using his hands. When the license arrived, he needed to get used to it for a bit, but then he got pumped to ship a bigger feature autonomously in that codebase. He wrote a good spec with a 9-item checklist and success criteria. Then he got Claude running to pump it out. Claude did a good job. Until step 7 out of 9. Tokens were out. Not just for the session, but for the month. Turns out, that Enterprise license comes with a $100 API token budget. He finished Claude’s code by writing it with the keyboard, using his hands.
And here we are talking about how coding is solved and engineers are replaced by AI. Tokens aren't solved. Organizations aren't solved. Just a few years ago I worked in German consultancies where “Digitalization” was the buzzword of all buzzwords. There are so many orgs that will need to learn to adopt AI in a meaningful way. Government orgs. Non-profits. Small orgs. The EU. The World outside. Even if the bubble tells you otherwise and AI engineering becomes that perfect system tomorrow, we have like 10-20 years worth of work and adaptation to the new ways ahead of us.
EDIT: Since I wrote the above, I realized that my AI bubble feed and personal conversations are far behind the general sentiment:
If you skim those threads, you will see that not only there will be a shift in organizations, but also in engineers first (just assuming for fun that those reddit communities are representative for the "general sentiment"). "I only code by hand posts" - ranking highest.
[AIN#36] AI Nuggets Week#36#
📺 Expert Knowledge YouTube Nuggets#
Who Owns Your Data Now? Agents vs. Systems of Record on The Agents #013 (Jason Lemkin & Amelia Lerutte, SaaStr AI) - SaaStr's Salesforce data grew from about 5GB to 40GB in a month, mostly through API writes and after a migration from Marketo. But do they still need Salesforce, now that their agent setup becomes superior? First time I see them sharing their agents hands-on on screen.
🎰 Just for Fun links#
(I'm having fun so you don't have to)
Cal's brain gym, exercise number 2 "productive meditation": take a walk, keep one real problem in your head, and notice when you stop making progress. No podcast during this part :)
https://x.com/AmandaBPerino/status/2096172716856422568
DHH keynote habits to teach all the slow people who prepare the conference slide deck for 80h (me) and he will still be funnier than your talk.
https://x.com/grok/status/2095840622674379159
Grok trying to outsmart Claude after the fact.