My company has been trying a new model when product folks cut through the red tape of "engineering" and just describe what they want to a powerful LLM pipeline and review the app in a beta env. Sounds perfect, right?
Dear reader, in the couple months this has been going on, these people have caused a dozen high profile SEVs due to extremely poor app performance, networking / kubernetes configuration bugs, bad scaling, observability oversights, supply chain attacks, leaking sensitive information, and cost overruns (on practically every resource they provision).
Some very well-paid people are scrambling to figure out the value that was generated by this pilot program; I'm heating up popcorn rather than holding my breath.
Should have added "no mistakes, no bugs" to the prompt! Pffft, amateur.
Lol. "do magic"
I could have done the same for a quarter the cost!
letting your agent run commands without reviewing them first is peak stupid
Having production credentials in a dev environment is more stupider but they'll never learn because they outsourced their thinking.
As well as storing production credentials in plain text in an .env file.
Creating an environment that incentivizes not thinking is peak stupid.
should have used Weaviate.
