Hacker News Reader

Tried them, sonnet is still where it’s at. Not sure what is the special sauce at Anthropic.

do u mean that Claude 3.7 Sonnet is still better for coding tasks than gpt-4.1 family, o3, o4-mini?

which version of Sonnet? thinking/reasoning or raw/one-shot?

Not who you responded to, but I use Windsurf pretty much daily. I keep trying new models but for actual code changes I keep going back to Sonnet 3.5.

3.7 can be useful for planning sometimes, but when making code changes it’s really hard for me to keep it from making unnecessary changes.

You're soon going to find a rugpull whereby Windsurf will remove all Claude models. This is to be expected considering Windsurf has been bought by OpenAI.

If that does happen, good thing that Cursor is very similar. Being dependent on any one tool in this environment is folly. Though to be honest, in either case, I am personally dependent on Anthropic. Comparing Claude Code costs to Cursor/Windsurf using Sonnet, it certainly feels like I am temporarily VC subsidized, like early Uber.

It seems to me that all of this is going to be a giant rugpull in the end. "Oh, you are saving the cost of a $40/per hour contractor? Well, we'll just charge you $35.99."

Until then, we might be in the golden era of cheap building, maybe?

Reply on news.ycombinator.com