Yesterday I had an intense discussion with a developer about some AI API.

They wanted to find an API that was free to use, and spent 3 days trying to find one.

I simply suggested they buy $10 worth of credits on whichever LLM provider they like (anthropic, openai) and they balked at me.

I told them about how them spending time to find the free tool is actually costing them a lot of money disguised as time.

Tried reasoning out with them but their logic was that they don’t want to end up paying a lot of money later when their product gets huge.

With $10 on any major LLM provider, they can tinker with the APIs for few months, get their product to work with it, and even go to production.

They will need to look for a cheaper alternative at a meaningful scale where they would either be making lots of revenue to pay for it, or raise their next round.

What they were not understanding was that all the value lies in being able to use the API.

Instead they were willing to spend time trying to build GPT2 from Youtube lectures and then use it on own hardware, to save $10 expense.

As a dev, if you earn 8LPA at 8 hour work days, your time is worth close $5/hr. If you spend 3 days on something, you’ve spent $120 (~10K INR). The hourly amount would be very high if they’re the ones setting up the company.

Just run the numbers on how much cost on API calls will you save if you do everything in house. It doesn’t make any sense at the beginning.

At scale, sure go for it if the numbers start making sense. But doing it for commercial purpose at the beginning doesn’t make sense.

I encourage you to learn about these things, but take it as an academic endeavour, not a commercial one where you have the ROI calculations in mind.

Just another painful reminder that nothing is free.