'Availability is just as important as capability' is worth taking seriously from someone in your position — a Google Champion Innovator saying this about Google's own product is exactly the kind of signal that gets dismissed internally as a procurement complaint rather than a strategic warning. The habit formation risk you describe is the actual long-term concern: 8 hours/day of Claude Code means that when you're choosing tools for personal projects, the activation energy for switching is higher than whatever capability advantage Gemini might have. This is how platform preferences calcify before capability gaps even emerge. The EU regional endpoint gap feels like it has a tractable fix — what's your read on whether this is a Google organizational prioritization problem or a technical constraint that's harder to resolve than it appears?
This hits close to home. The regional endpoint gap is wild when you think about it. Google's own model garden serves Claude 4.6 to EU customers faster than it serves Gemini 3. That's not a good look.
I've been dealing with a similar kind of friction but from the tooling side. I actually like Claude Code as a coding agent, but I don't want to be locked into one model either. I ended up setting up claudish to route Claude Code through OpenRouter so I can swap between Gemini 3.1 Pro, GPT-5.4, and Claude depending on the task: https://reading.sh/claude-code-how-to-run-any-model-gpt-5x-gemini-3-1-stealth-inside-it-e67e957e53c3 Takes about 10 minutes and then the provider@model syntax lets you just pick whatever's available. Kind of sidesteps the whole vendor lock-in problem you're describing.
Curious whether you've tried gemini-cli with any non-Google models, or is it strictly Gemini-only?
'Availability is just as important as capability' is worth taking seriously from someone in your position — a Google Champion Innovator saying this about Google's own product is exactly the kind of signal that gets dismissed internally as a procurement complaint rather than a strategic warning. The habit formation risk you describe is the actual long-term concern: 8 hours/day of Claude Code means that when you're choosing tools for personal projects, the activation energy for switching is higher than whatever capability advantage Gemini might have. This is how platform preferences calcify before capability gaps even emerge. The EU regional endpoint gap feels like it has a tractable fix — what's your read on whether this is a Google organizational prioritization problem or a technical constraint that's harder to resolve than it appears?
This hits close to home. The regional endpoint gap is wild when you think about it. Google's own model garden serves Claude 4.6 to EU customers faster than it serves Gemini 3. That's not a good look.
I've been dealing with a similar kind of friction but from the tooling side. I actually like Claude Code as a coding agent, but I don't want to be locked into one model either. I ended up setting up claudish to route Claude Code through OpenRouter so I can swap between Gemini 3.1 Pro, GPT-5.4, and Claude depending on the task: https://reading.sh/claude-code-how-to-run-any-model-gpt-5x-gemini-3-1-stealth-inside-it-e67e957e53c3 Takes about 10 minutes and then the provider@model syntax lets you just pick whatever's available. Kind of sidesteps the whole vendor lock-in problem you're describing.
Curious whether you've tried gemini-cli with any non-Google models, or is it strictly Gemini-only?
I think gemini-cli only supports gemini models currently. However, Antigravity has a wider selection