19 comments

[ 3.1 ms ] story [ 54.7 ms ] thread
Anthropic didn't make the cut in our evaluation (data usage concerns). They have also been the shadiest of the companies

They lost me when they expired my money and then tried to take more without asking

I don't think it's Microsoft that favors it. It's likely customers. Claude wipes the floor with all the GPTs in GitHub Copilot (in my experience).
In some ways it makes sense to pave the way for Claude to protect the brand of VS Code. On the other hand, it’s a bit of a head-scratcher since it seems like VS code was built as a loss-leader to sell Microsoft cloud products. Perhaps enterprise ChatGPT, co-pilot and GitHub can make up the difference even if the community tier favors Claude.

Edit: maybe Cursor forced this and Microsoft is taking its choice to open license VS code on the chin. Will be interesting to see the strategy with Visual Studio going forward.

Claude was the gold standard for coding but I have had a lot of success with GPT-5. Nowadays I pretty much always default to GPT-5.
Claude/Anthropic is more focused on productivity (Coding, Spreadsheets, Reports). ChatGPT seems more focused on general-purpose LLM (Research, Cooking, Writing, Image Generation).

Makes sense that MS would partner with Anthropic since their tool-use for productivity (Claude Code) seems superior. I personally rarely code with ChatGPT, almost strictly Claude.

This needs that archive link that bypasses the paywall. I had to read it on my Apple News+ subscription to avoid the paywall.
Anthropic doesn't allow me to use my phone number across my personal and business logins. I simply can't use Claude where I need it, even if I'm willing to pay. I don't understand why they add so much friction when everyone else just allows me to do work.
Claude Sonnet 4 is the best for generating code for me so far, albeit needing some investment in instruction files and prompt files when using GitHub Copilot.
Auto feels like a way for them to slightly push people towards paid models more. If they really favored Anthropic, claude would be the included free model, right?
Personally I found gpt-5 to be a bit better than sonnet-4. At least in cursor. Claude is still more reliable and competent at tool calling, but I found gpt-5 to be better in token efficiency and a lot better at instruction following.
This is Microsoft subsidizing Claude inference costs -- if you look at how they charge models against your allotment, Gemini, GPT-5 and Claude 4 Sonnet all cost the same, despite Claude 4 Sonnet being more expensive than the other two. Not really sure I understand the economics here, especially since there's not really a clear winner between GPT-5 and Claude 4 Sonnet for coding (if anything I think GPT-5 puts up a better showing).
I use deepseek on the daily, the other day I tried to use claude and was surprised when after like 5 messages (had decent amount of code though) I got "limited"
Embarrassingly dumb question ...

is Claude Code just a plug-in for an existing editor. Or it is the entire editor itself?

ChatGPT seems to be really good at analysis but continues to be bad at coding.

It constantly loses existing code. It the end of an intense coding session you’ll find your 20 feature application now has 3 features. That alone makes it not worth using. I can’t be bothered constantly working to identify and prevent losing existing features.

It is however very powerful for doing debugging and diagnostics with throwaway code.

It’s also useful to analyse stuff with ChatGPT and give that to Gemini or Claude to do the coding.

Doesn't Microsoft own a lot of OpenAI? Did they sell?
I have to say I prefer Claude for code and diagnosis. ChatGPT in CoPilot often replies with essentially, "these are the things I would be looking at if I wanted to troubleshoot this", to which my muttering frustration is, "yeah obviously - no sht - now do it.", whereas Claude just does it (e.g. looks through the code, provides analysis as to where the problem is or the nature of the problem, and returns a practical and immediately useful response). You could say that's a prompting problem (and you'd be right) but if Claude is lower drag than ChatGPT in this context and others, that's a value proposition.

However, I really do* prefer ChatGPT for generalized but as-applied conversation about architecture decisions, pros/cons of different approaches, patterns. Claude can do this too, but I do prefer ChatGPT for this.

Microsoft has a dozen vertical Copilots to build, so picking the model with the best capability today makes sense. If Claude Code is stronger for dev productivity, using it in VS Code just raises the bar for everything else they ship