I’m a copilot pro+ user that also is looking to move away, as the new pricing model means you “prepay” the 40$, but the pricing model is based on tokens, so you could use 40$ slowly, or super fast depending on what you are doing. Ask for more complex help, the 40$ wont get very far depending on the model. You still get tab completions for free, but that isn’t what most people look for nowadays.
where I can request help with coding, bug fixes etc, does anything like that exist outside github?
Practically it would be good old fashion google, which is slow and only is as effective as you can make it by finding answers yourself. You could ask for targeted help for any common AI provider, like ChatGPT/Gemini. But this isn’t as effective or native as having Copilot directly managing your codebase for you.
If you are learning or new, I actually suggest this more than any AI. Mainly because if you find the AI is wrong (which is a lot) burning more tokens pointing out issues is literally burning money to burn more money, IE your using AI to “solve itself” instead of using some secondary resource.
AI is a useful tool to do something fast, but it isn’t perfectly accurate. So when its wrong, not only do you need to notice but you need to know how to solve it without just asking AI again, because very very often asking it again gives you even more wrong answers. With how pricing is moving, this means just making an expensive problem more expensive.
So this brings me to learning how to find answers without AI. You wont necessarily need to do this, but you need to know how to find answers yourself and get unstuck without AI. Otherwise if you only know how to ask AI to solve the problem for you, you end up completely stuck when AI isn’t available or if its wrong.
TBH I see basically all providers moving toward either degradated models (where the model gets dumber, or less effective) on the path toward degredated models, or toward token based usage. You can always jump around, or chase free tiers but none of this is super effective long term.
“Peak AI” is over. You need to have skills behind you or just really deep wallets or know how to actually run models yourself, which is free beyond paying for the hardware and power.
If you want actually free and always free models run them locally using something like ollama or lm studio, hook it up to something like opencode and you run your model locally. The main flaw with this is the models you can run locally aren’t as good as other models like those provided from Anthropic, you also are constrained by what hardware you have locally. If you don’t have an Apple device, or a dedicated GPU your performance will be poor even with small models.
This might be a route to play around in, it’s a free route you can at least try, but I highly recommend just taking the foot off the gas for AI and actually learning the stuff.
The future will be made up of developers who know how to use AI, and developers who use AI and know what they are doing. Only one of these I think will stay employed long term, as the first group will just be seen as a money pit that requires extensive tokens to do their job. Most companies are not at this point where token usage needs to be optimized, but I know companies are starting to care, as money is not infinite.