If, by some miracle, all LLMs ceased working right this second, any developer who would no longer be productive should not have been a developer in the first place.
You don’t need a miracle, if Anthropic API is down due to technical issues you don’t have software development anymore. It’s insane how much we are delegating to 3rd parties. It’s not like having cloudflare down where your users cannot access your services. The AI tools used to investigate prod issues stop working, developers stop working. The AI support system that allowed the company to get rid of their support team stops working. In addition to all the issues that causes to customer facing products based on AI. The sales team cannot work anymore.
It’s like the industry is willingly introducing a common external risk to everything
Limiting token quotas would be fine. Encourage developers to use efficient models, plan the work first, and to not burn thousands of GPU hours on waste.
It's much like when developers would waste tons of money on AWS spinning up massive test VMs and leaving them running without care. Until the finance people cracked down on it.
If, by some miracle, all LLMs ceased working right this second, any developer who would no longer be productive should not have been a developer in the first place.
You don’t need a miracle, if Anthropic API is down due to technical issues you don’t have software development anymore. It’s insane how much we are delegating to 3rd parties. It’s not like having cloudflare down where your users cannot access your services. The AI tools used to investigate prod issues stop working, developers stop working. The AI support system that allowed the company to get rid of their support team stops working. In addition to all the issues that causes to customer facing products based on AI. The sales team cannot work anymore.
It’s like the industry is willingly introducing a common external risk to everything
True, but they will not want to work for you anymore, they'll want to work for company that provides it.
I'd happily work for a company that paid me the money they would have spent on LLMs.
2 replies →
Limiting token quotas would be fine. Encourage developers to use efficient models, plan the work first, and to not burn thousands of GPU hours on waste.
It's much like when developers would waste tons of money on AWS spinning up massive test VMs and leaving them running without care. Until the finance people cracked down on it.