cyndaquil
Lv 100 Bold natured
I use VS Code you can either use copilot pro and select claude as your AI model or use the claude code extensions.
AI (LLMs), Agents, MCP, Open Claw, Coding Assistants
- Thread starter Regular Developer
- Start date
-
- Tags
- chatgpt llm programming tech
cyndaquil
Lv 100 Bold natured
Holy fukk I just tried using the ADO mcp server.
all you software engineers who are building out mcps for jira/confluence etc deserve a motherfukking salute.
MCP + VSCODE = Time for Family
null
...
good news 
claude sonnet 4.6 is seems to be much better.
less chit-chat.
more admitting when wrong.
asks logical feedback / investigative questions aka "multi-step tasks".
"Anthropic claims its new Claude model showcases ‘human-level capability’ in multi-step tasks."
www.siliconrepublic.com
a bit argumentative but i just tell it to
claude sonnet 4.6 is seems to be much better.
less chit-chat.
more admitting when wrong.
asks logical feedback / investigative questions aka "multi-step tasks".
"Anthropic claims its new Claude model showcases ‘human-level capability’ in multi-step tasks."
Anthropic says new Claude Sonnet 4.6 is much better at computer use
The Claude Sonnet 4.6 shows 'major improvement' in computer use skills when compared to prior Sonnet models, Anthropic claimed.
a bit argumentative but i just tell it to
Opus is honestly the best I've worked with but it eats through tokens way too muchgood news
claude sonnet 4.6 is seems to be much better.
less chit-chat.
more admitting when wrong.
asks logical feedback / investigative questions aka "multi-step tasks".
"Anthropic claims its new Claude model showcases ‘human-level capability’ in multi-step tasks."
Anthropic says new Claude Sonnet 4.6 is much better at computer use
The Claude Sonnet 4.6 shows 'major improvement' in computer use skills when compared to prior Sonnet models, Anthropic claimed.
a bit argumentative but i just tell it to
Thread renamed
null
...
I've been looking at open claw: OpenClaw — Personal AI Assistant .
It's wild. I reinstalled a small GMKtec mini PC (linux) and will try to set up Open Claw (OC) on that.
From what I have read OC doesn't use MPC.
You interact with OC via a configured chat "Channel" on a messaging service, such as WhatApp.
OC figures out what to do using a configured LLM (such as ChatGPT) through use of formal output prompts.
OC performs the actions indicated by the LLM using configured "Skills".
You give OC full rights to perform actions on the host it is installed on.
OC inventor Peter Steinberger has been hired by Open AI.
It's wild. I reinstalled a small GMKtec mini PC (linux) and will try to set up Open Claw (OC) on that.
From what I have read OC doesn't use MPC.
You interact with OC via a configured chat "Channel" on a messaging service, such as WhatApp.
OC figures out what to do using a configured LLM (such as ChatGPT) through use of formal output prompts.
OC performs the actions indicated by the LLM using configured "Skills".
You give OC full rights to perform actions on the host it is installed on.
OC inventor Peter Steinberger has been hired by Open AI.
Last edited:
One of my colleagues just brought up open claw today, funny enough. Now I have to look into it, if 2 people are mentioning in quick time
I've been using Codex + VS Code a lot it's really boosted my productivity. Basically using it to speed up boring parts (sql, Python, pulling data).
Starting to get better at using plan mode instead of spamming commands. I built a few tools for it, already so want to start adjusting skills. And my biggest complaint they just fixed, saw the pop-up about using sub agents.
Don't even know if I need them, just happy to get experience trying to use it.
Starting to get better at using plan mode instead of spamming commands. I built a few tools for it, already so want to start adjusting skills. And my biggest complaint they just fixed, saw the pop-up about using sub agents.
Don't even know if I need them, just happy to get experience trying to use it.
null
...
any more specifics.
My use case is data science, but also do analysis and often have to do my own data engineering work to get the data.
I'vs used to streamline analysis, document things so I can run it again in the future (i'm ass at this). Eventually I think I can get it to automatically write reports and push info to people. Much better than dashboards, I hate dashboards more than anything.
The DE stuff I've been using to quickly set up my API pulls, transform the data as needed, dump it to dbt etc. Pretty basic but saved me a ton of time and makes it easier to get the real DE team to put it in the actual warehouse.
Then running ml models, iterating, testing out ideas. It was fast before simply using the AI tools, but super fast now with it plugged directly into my IDE.
One of my tools I basically:
Built a python tool that could connect to a db, then execute SQL scripts and then had a few options for exporting. Easy to do, but that allowed me to tell the agent to use that tool, and then just point it to code, or have it generate code and run it. I got tired of explaining what it needed to pull, so I turned it into a skill and added another tool that every week will go into the db rebuild it's own data dictionary and track how tables connect.. Then having it save any queries it runs for additional context.
Hoping to get it to the point where it understands what tables to pull from on it's own.
I'm not building apps or shipping code, but it's been a big help for me, can also run two at once and tab between projects when i get bored.
Ive gotten some good recognition off recent work I've done with it's help.
They're encouraging us to use AI and learn all of this so I'm diving all the way in.
Last edited: