Ask HN: Cheapest way to use LLM coding assistance?

[iamlucaswolf]

I would like to toy around with LLM assistance (Chat/Copilot) for my coding side projects. Given that I will only use this sporadically (20hrs per week max, with variation between weeks), I was wondering if there was a cheaper way to access these tools than paying for an OpenAI/GitHub Copilot subscription.

Has someone here done the math on self-hosting on the cloud/running locally/using the APIs directly/just paying for the subscriptions?

Some additional details on my specific situation (but also interested in general considerations):

- I’m living in Germany (26ct/kWh, no access to Claude) - I already have access to a box with 64GB RAM, no GPU though)

[b-mmxx]

Cody was mentioned already. Continue [0] is another tool that you could give a shot, load local llm etc.

[0] https://github.com/continuedev/continue

[AtomicOrbital]

checkout the cody extension https://github.com/sourcegraph/cody available for various editors like vscode

[iamlucaswolf]

Thanks, that seems to be exactly what I was looking for!

Any insights on the business model? Is this just to advertise the pro tier or do they monetize usage in another way (e.g. selling training data)?