$20/mo isn't a lot... Especially if you make money with coding.
Also 20$ per month is way less than what it costs them to run it. Eventually they will need to charge way more to cover their costs, and the people who can't code without an AI assistant will need to pony up :)
That only applies to regular ChatGPT use.
And developers actively using AI for coding can easily spend more than $20/mo for the API.
There are people spending $10-15/day in OpenAI API usage working through Cline.
If you’re not running the model locally, you’re sending your code to them for analysis. Now ByteDance has it.
The issue would be based on the terms of employment and the software license. There’s likely a provision that just says “don’t share” regardless of what the other party will use it for.
To OP’s point, if your company is paying for the sub, then sending the codebase data would be an approved use of the codebase as part of your job.
I hope your current and future employers never find this side of your personality :D
If it benefits you more than 1%, then you're in profit.
Of course, if you're in a job that doesn't actually care about performance, and performing won't lead to better salary at some point, then it may not matter.
I get making economic/stats based analysis like this, but is your boss going to notice that 1% to give you a raise they otherwise wouldn’t? Probably not…
Your company culture can be performance-minded and this still be true.
The real problem however is, I cannot simply share my employer's repos to be absorbed by any LLM out there. So I use only the tools that my employer provides and approves of. Currently that is Microsoft Copilot chat/RAG via my work account. It takes some copy/paste and adaptations of problems/solutions but it is much more efficient than using SO. It is also a great teacher that never gets tired of my plenty why/how questions.
In my view, the future is that LLMs can train on entire private code repos until it understands its ins and outs. Currently it would need to fit in the context window, hence you need to babyspoon it, as I understand things.
In this case it tricks you because it assumes that the LLMs increase productivity and launders that into the calculation. For me, and many people, LLM usage decreases my productivity.