pull down to refresh
Is it me or does Anthropic seem to be gaining steam as the frontrunner for AI agents
idk, Gemini 3 really closed the gap and Grok is in the mix. They've been the frontrunner, but seems to be trending less so than moreso.
Coincidence that neither are using custom chips instead of Nvidia GPU's for training?
I think that both Gemini (as a generic chat & image model, with phone integration) and Grok (as a search & ELI5 model) beat GPT and Claude 9 out of 10 times. Google did a great job with Gemini 3 (it's much better for normies than GPT 5.x) and honestly Grok 4.1 is my go-to as a search engine stand-in right now.
Since May last year though, whenever I want to vibe code something, Claude (+ Code tooling) has been consistently improving and reducing the need for me to correct it (and beating the competition). I suspect that this is because Anthropic prioritizes optimization for tool calling, extensibility and self-checks, whereas both Grok and Gemini are a bit more generic. Gemini 3 has beaten Claude 4.5 Opus on some standalone coding questions I've thrown at it on arena.ai, but from a process perspective, I think that Anthropic has something going for them.
fwiw, I like Anthropic the company about as little as I like what Google has become, but that doesn't mean that their product can't be good.
Following concepts, not specific LLMs. The models are just the implementation, as are the tools.
What's interesting about clawdbot is that it is built with Skills, not MCP. This is something we saw coming for a few months now and it gives Anthropic a big moat.