AI Inventory Starbucks Project Failed - Bad Technology Doesn't Work

Tying an entirely non-deterministic tool like LLMs to something like inventory is hilarious to me.

CEO’s love fads, seen it over and over. It’s just mind boggling how much money they waste but yet get rewarded by big fat bonuses at the end of the quarter. Having been a corporate drone for 40 years I’ve seen it all so am not surprised by these kind of things.

We need a YouTube Channel that documents all of the AI Failures

Spend ridiculous amounts on tokens, redesign the packaging with multiple types of packaging for each different type of thing in inventory and end up with results as quirky as RFID.... Why are we UNSOLVING things we already solved just to spend more money solving them in the most expensive way possible??? This is nuts. Maybe we could just improve the things we already have that work.

"normie selling to king normie" I laughed too hard on this

There's a company in Sweden who only does weird tests. The same lads who built the LLM vending machine that Anthropic and WSJ got to bonk until it gave out a free game console. They also ran attempts where different LLMs would run a cafe. They managed to build a machine which forgot to schedule staff to open the cafe, and ordered 6 giant cans of tomato which nothing on the menu used.

Not everything needs an AI solution

These AI projects are all about going public and siphoning index funds. Prove me wrong.

The best "talking the CEO down" scenario is when the CEO gets back from a trade show already sold on some product or service and you have to gently explain that you're already doing exactly the same thing but without the additional middle man.

The real problem I have with build things near the interstates is that at that point we should probably really consider improved rail infrastructure.