| ▲ | sidrag22 4 hours ago | |
I was "Claude only" for well over a year. Kinda crazy how they seem to be gaining a LOT of public attention the last few months, yet i see this type of sentiment from other devs/myself. for me it started with their opencode drama, and openai's decision to embrace opencode in response. I didn't even know what opencode was prior to that drama, yet now here i am using opencode and a ton of crafted openai agents in my projects. Would love to have some claude agents in that mix, but i guess im stuck in Claude Code if i wanna even touch their models... I'd love to go back to just claude as i "trust" them more in a sorta less evil vibe manner, but if they are gonna prevent subscription usage to something people use to allow themselves more freedom, they gotta then close that gap with their own tools rather than pumping out stuff like this which scares me off given the past couple months. I totally understand why they are cutting off 3pa access to stuff like openclaw, where the avg user is just a power user in comparison to avg claude user or whatever. I haven't kept up a ton with their opencode issues, but I just know i can't get behind a company actively trying to make my potential usage of tokens less optimized to keep me locked into their ecosystem. Really just kinda hoping local models kill it all for devs after a few years, I'm not interested in perma relying on data centers for my workflow. | ||
| ▲ | lbreakjai 40 minutes ago | parent [-] | |
The most cost-effective way I found to use their models is to use them through a Github copilot licence. Github charges by request, not per token. Asking Opus with high effort to plan a feature in depth "costs" the same as asking it the number of "r" in strawberry. I've got a setup where GPT5-mini (Free on GH) talks to you to refine and draw the outline of your feature, calls a single Opus subagent to plan it in depth, then calls a single sonnet subagent to implement it. Github will only charge you for one opus request (With a 3x multiplier) and one sonnet request, whether they consume 50 or 500.000 tokens. I'm running this setup for 9 hours a day at work and I've barely consumed 40% of my monthly allowance. | ||