@steipete Cool! I'm non technical & I built something similar in Hermes. Whichever model I'm using can read the skill. It can create, move, update, erase, triage Bugs. I don't touch it all. I call it Raid. I like to be in the loop, so I'll say pull bug 234. Then it starts working on it.
@cjzafir Haven't tried this one yet. I'm curious as well because I despise everything Grok related. I find the chat app to be unusable in comparison to chatgpt i figured their coding models would be similar. They will likely jack the price up after a few months anyway ...
@llmdevguy What's your take? It feels like v4 Flash is a much better model for coding. It's faster, cheaper, follows directions better and feels closer to GPT 5.5 than v4 Pro. In my experience v4 Pro has an issue with disregarding instructions not to patch unless approved.
@Jmoon_174@ArtificialAnlys@deepseek_ai V4 Pro is cool but super slow and doesn't follow directions as well as V4 Flash. Coding with V Flash imo is faster, follows instructions better, and honestly better at coding imo. Unless it is something really complex... only drawback to v4 Flash.
@joewill1082630@ArtificialAnlys@deepseek_ai Not to users but I have created a bug free integration into hermes for an automated Bug triage app. Pretty cool, built it in 36 minutes. With one iteration. I was considering open sourcing it but I'm just a non technical guy, I figured people wouldn't really care about it.
@ArtificialAnlys@deepseek_ai The point people are missing is models like V4 Pro remaining affordable will likely put pressure on the western "frontier" models to lower prices as market share continues to be lost. Keep up the pressure Deepseek!
@OpenRouter V4 flash is cool but it takes probably quadruple the amount of inference to get something simple done whereas gpt 5.5 just gets it in one try. I still use it but knowing that it will take a lot of iterations to get things done.
@jahooma@AIVibecode@deepseek_ai This was on my idea list at number 5 just in case my current idea fails. Good job! You beat me to it. If done right, you should be able to capture decent market share
@laozhang2579 I've been using DSv4Pro + Hermes & today tried GPT5.5 + Hermes for the first time. Wow, what a difference. You notice it right away. All of the bugs v4Pro couldn't get, GPT5.5 instantly 1 shot. DSv4Pro is cheap but you pay for it in frustration and opportunity cost.
@itsolelehmann I use DeepSeek V4 Pro on Hermes with no auxiliary models, and the $30/month figure is definitely user dependent. Iโm past that in a week with over 1 billion tokens.
@llmdevguy@dhtikna What did you think of Mimo v2.5 Pro? I was a fan of v2 Pro but something about the training or tuning of v2.5 I found to be off. Switched to Deepseek v4 Pro and so far has been much better for my use cases.
@WatcherGuru This is going to significantly reduce options volume imo. Most small traders only trade options to avoid PDT. Now with $2500 margin and 4:1, 2:1 overnight, a small trader can make decent trades with no theta sticking to $10-$50 names. About time!