Recently moved to Windsurf from vscode, Microsoft keeps degrading copilot AI, and I am finding it’s ‘filters’ their agent is using (for all llms) are now so limiting it’s almost unusable due to the tweaks they have made causing code to be terrible and more mistakes than for instance the same instructions with same LLM’s in Windsurf.
Just touching base what people are using as I’ve found Windsurf is not only 1/4 of the price for the same code ,but the output in cfscript/lucee 7 a lot cleaner.
If you are looking to get/pay for Windsurf, you can also use this link to get 250 extra requests if you decide to subscribe to pro (which i did)… #enjoy
The new llm comparisons they released recently also allow devs to compare two llms with the same task to compare how they work for particular tasks.
Windsurf seem to have a lot more LLM’s lately also.
I’m really loving the way windsurf IDE works (based on Monaro like vscode) but their approach and UI is much better than vscode.
What are people finding with IDE’s is great for coding agents?
Mainly using Claude Sonnet 4.5 (thinking code) , chatGPT 5.2 medium (for docs) and Kimi K 2.5 (scafolding and repetitive code) atm…
what are people using this month (as this seems to change each month)
I’ve been using Google’s Antigravity since it launched late last year… I already have a Pro Gemini/AI/Google One plan (came free with my phone) so I can take advantage of the extra quota, but even the free tier is pretty generous apparently. So if anyone wants to try it out and not spend any $$ it’s not a bad way to go.
This agentic IDE was foundational to me bringing a site up from Lucee 5.4 to 6.2, and move from a series of automated custom install steps (with a very old school directory structure etc) to using the lucee container. Really amazing what I’ve been able to accomplish in a short time.
Also I just throw exceptions and other errors into it and it works out the root causes in CFML and other infra code/setup easily (I’m more of an infra person than CFML)
CascadeThis is a massive cost saving when used correctly, reduced my spend by 75% using plan, then cascade in this matter, and using GPT5.1-codex for bulk changes, or general questions is a no brainer as it’s free, and still has enough reasoning to do most tasks.
Being able to have it plan in Claude Opus, then execute using Claude Sonet or even GPT for renaming or simple tasks means when you ‘Implement’ the plan it runs through the tasks top to bottom with context and can do quite a bit of work, conversion of libraries from python or other languages for instance in ONE single credit. #nice
