Andrew EadesI listened to the Vergecast interview with the Claude Code guy. It was annoying that the interviewer was so all-in on AI that he never questioned some of the things being said. One example of using Claude he gives is that he has a spreadsheet that all his teams update with project status and he prompts Claude to check the spreadsheet for status updates and send reminders via Slack if people haven’t updated. Except it gets at least one name wrong!
Leaving aside that he is tracking project status in a spreadsheet instead of an AI-coded tool, how have we got to the point where we are prompting an LLM do something like this? And it doesn’t do it right? Yet this is unquestionably taken as proof of how amazing AI is! This is the best example the creator of Claude Code could come up with FFS.
The real thing here is that people desperately want text-to-action. LLM’s are probably part of the solution but not the entire solution. It would be trivial to write a spreadsheet function that did this job 100% correctly but we are settling for a half-working thing because we are too lazy to start.
It’s like when before the browser address bar was merged with the search field, we all went to google.com and “searched “ for the website because it was easier than typing or mistyping a URL. So instead of going directly to a site, we all did this hop from google.com even though it was less efficient.