Here we go again with Claude AI vs Gemini. This time I was inspired by a recent post by Nathan Warren at Exponential View. This one: “Claude 4 – The first universal assistant?”, and I thought I would try its AI assistant capabilities out a little bit. Here’s a slice of Warren’s post on why he was giving this a go:
The promise of a truly universal AI assistant – one that bridges thinking and doing, research and execution – is becoming a practical reality. I got early access to Anthropic’s new model, Claude 4 Opus, and had a chance to run it through its paces while preparing for annual leave. So my key question was: can Opus 4 help me get more done in one day than I could do alone or even with other LLMs?
Claude’s AI assistant abilities are leveled up by adding ‘connectors’ - there are connectors for Gmail, Google Drive and Calendar, Notion, GitHub, Zapier, and more. Notion AI is so damn great already that I don’t feel the need for Claude or any other tool to enhance it, but I was curious to see how the Gmail and Google Drive connectors might work, partly thanks to Warren’s mention of using it to tackle his to-do list by searching in Gmail.
My Gmail task for Claude was summarizing emails, specifically any AI newsletter emails, as I subscribe to a big number of those, so if Claude or Gemini could do a good/great job on this it would be nice time saver. My prompt to Claude was as simple as it gets (and I’ll want to refine this if this becomes a regular task):
Find any AI related newsletter emails in my Gmail Inbox and summarize them
You can see Claude starting to respond to the prompt in the screencap at the top of the post. Then it worked further and offered up some quite good results:
I gave Gemini the same prompt and it did well, but it offered notably less deep responses:
Although it still did well on identifying good sources (AI newsletters):
The ‘less deep’ difference between Gemini and Claude is more significant when it comes to asking for research and summaries of Google Drive content, with Claude again on the deeper side. Here’s the simple prompt I gave to both of them for the Google Drive research:
Tell me about cyber risk methodologies and in particular quantitative versus qualitative methodologies based on what I have on this area in Google drive
There’s a lot of good source material for this prompt, years worth of notes and saved artifacts from training I’ve done, projects I’ve worked on, and more:
Gemini chose to look at one document and then share its response. It’s a good source, but still just one source. When I added a follow-up prompt asking it to leverage more resources in Drive, it chose one additional document.
Claude looked at more than 30:
Here are a few screenshots of Claude working through the Cyber Risk sources and drilling down within Security > GRC > Cyber Risk:
Claude is the clear winner for me here. I feel somewhat confident that Gemini can and will improve though. I hope it does, because as good as Claude is I’’m not sure I love the idea of the access into the Google Workspace that these connectors give it. Ideally Gemini will improve and it already has all its hooks into Google Workspace and I’ll probably turn off Claude’s connectors.
As always, I’d love to know what you’re up to with Claude, Gemini, AI assistants in general.