But will it be Claude? Probably it goes through Microsoft through copilot. In which case it becomes a race to the bottom. I never need Fable even today: Opus suffices on 99%. So why do I need anthropic? Opus will be replicated in half a year by every lab.
In that sense I am super bullish on Microsoft as they have the vendor lock.
For anyone who has used the various tools, the harness still matters and copilot is pretty bad. While Anthropics models lose on cost compared to OpenAI, there is still something that works really well in Opus paired with the Claude desktop app when it comes to tool use and solving tasks.
I see what youre saying, but there is no way a TUI or GUI is anthropics 1.5T moat
Why are you pivoting the argument? You said why use Anthropic because copilot can use the models. For anyone who has used copilot and Claude, they are worlds apart and MSFT has been working on copilot for more than a year and they are light years behind and are not answering simple business use cases. I am not here defending the value of their moat but simply the harness still matters.