I’m a big fan of Claude from Anthropic for tasks where I need AI assistance with my writing. I don’t have AI create content for me—I dictate using my voice and then have it clean things up. I’ve also been using Claude for coding through Cursor, which has defaulted to Claude 3.5.
Last week, while I was at the SANS OSINT Summit, Claude 3.7 was released. I was excited to try it since they claimed the biggest improvements were in math and coding capabilities. While I don’t need it for math, I definitely can use the coding improvements. What’s also really cool is they now have a “thinking mode” similar to ChatGPT models. The thinking mode is supposedly even better at coding, so I was eager to test it.
Last night, I had a late-night coding session using Claude 3.7 with thinking mode to write some Python code. First impressions: it’s much quicker than I expected. When I use the O-1 Pro from ChatGPT (I currently have the $200/month plan for deep research), it’s noticeably slow. Claude seems much more responsive in comparison.
The quality was impressive too. The first small standalone program I asked Claude to write was fascinating—after providing the code, it immediately said, “Wait a second, I see a problem here” and fixed it before I could even respond. When it finally declared “That should be good,” I ran the code and it worked perfectly on the first try.
I also needed help dealing with changes to the Slack API for getting notifications, and Claude quickly provided solutions across the board for that project and other portions I was working on.
Overall, I’m super impressed with Claude 3.7’s thinking mode. It’s definitely my new default in Cursor for coding. If you haven’t had a chance to play with it yet, it’s making something that was already amazing even better.
