Which AI tool have you been using the most lately?
Get dedicated servers deployed in minutes here. Miami, Los Angeles and NYC area. Unmetered bandwidth. 24/7 data center support. Providing hosting since 2006.
Personal: Github Copilot (usually whatever latest Claude is added) for coding and Perplexity/ChatGPT for anything non-coding. Not paying out of pocket for any, thankfully.
Work is Claude via VertexAI + Github CoPilot in whichever IDE / terminal of choice.
There's access to all of Anthropic models via VertexAI, but will normally use Sonnet for coding / code understanding, and Haiku for Atlassian (Jira / Confluence), and Context7 MCP "agents"
Personal is:
Claude, Gemini Pro via Antigravity + Google AI Pro
Kimi, Minimax via Kilo Code in Windsurf IDE
Z.AI GLM Lite via Claude Code
Trial of GPT 5.3 Codex via Opencode and ChatGPT Plus
Whichever one is responsible for buying up all the global NVMe stock and ruining the LE market for the rest of us. 😂
Jokes aside, mostly ChatGPT. It basically acts as my junior sysadmin intern. I ask it to write obscure iptables rules or Docker configs, and then I spend 10 minutes fixing its hallucinations.
@kako1talk said: Jokes aside, mostly ChatGPT. It basically acts as my junior sysadmin intern. I ask it to write obscure iptables rules or Docker configs, and then I spend 10 minutes fixing its hallucinations.
Nice, I called them hiccups, but hallucinations has a more nicer sound to it. I think I will borrow this
I used ChatGPT during 3 and 4, but 5.0 really broke a lot of shit, 5.1 broke what I fixed for 5, so by the time 5.2 came out, I gave up. Just Claude right now. We have Gemini for work, but fuck that. Every once in awhile I'll troll through Grok to see what it can do, but Claude writes mostly-usable code.
@WSS said: I used ChatGPT during 3 and 4, but 5.0 really broke a lot of shit, 5.1 broke what I fixed for 5, so by the time 5.2 came out, I gave up.
Hated 5.x so much, it actually made me angry.
Claude seems really good with code. But I sent it a picture of my dried out dying cactus and it kept trying to gaslight me into thinking it was a mushroom. Even after I told it it was definitely a cactus. It kept breaking the news like I had a bad illness: "I don't know how to break this to you.. but that looks like a mushroom". It was not a bloody mushroom!
@WSS said: I used ChatGPT during 3 and 4, but 5.0 really broke a lot of shit, 5.1 broke what I fixed for 5, so by the time 5.2 came out, I gave up.
Hated 5.x so much, it actually made me angry.
Claude seems really good with code. But I sent it a picture of my dried out dying cactus and it kept trying to gaslight me into thinking it was a mushroom. Even after I told it it was definitely a cactus. It kept breaking the news like I had a bad illness: "I don't know how to break this to you.. but that looks like a mushroom". It was not a bloody mushroom!
Claude is good when you tell it with no uncertain terms to not guess or overcorrect when told that it's wrong. I'm still learning it- only been using it about two weeks as my primary. Grok is brilliantly braindead. It's either all in, or restricted to the point of uselessness after raping Will Stancil, I guess.
OpenAI has done itself a huge disservice with the 5.x branch. 5.x wouldn't even assist me with study because ESV is copyright (even though the copyright is very wide), and absolutely refused to assist me despite acknowledging that I was correct and it was wrong.
Recently got yeeted out of perplexity...was using a ahem grey market plan. And just realized I never needed it. Had a google workspace plan (to use custom email domain)...and that includes the fancy version of gemini chat and notebooklm. Could have just used that all along...
Currently busy making my own S3 server. Cause minio seems to be dying and figured may as well try my own rather than figure out what else exists
Use Claude code with glm is much better than opencode+glm.
Personally use codex. Sadly I am not allowed to use any ai tools in work environments.
I was using CC initially for about a month. It's solid but didn't like some of the choices the anthropic lock in forces on you. e.g. the hardcoded websearch that overrides MCP. They both do the job I guess.
ReliableSiteHosting
Hosting Provider
Comments
Gemini.
Opencode and GLM
Currently busy making my own S3 server. Cause minio seems to be dying and figured may as well try my own rather than figure out what else exists
Use Claude code with glm is much better than opencode+glm.
Personally use codex. Sadly I am not allowed to use any ai tools in work environments.
Action and Reaction in history
Random rotation until one impresses me
Insert signature here, $5 tip required
Personal: Github Copilot (usually whatever latest Claude is added) for coding and Perplexity/ChatGPT for anything non-coding. Not paying out of pocket for any, thankfully.
Work: Azure OpenAI and Gemini
Head Janitor @ LES • About • Rules • Support
Work is Claude via VertexAI + Github CoPilot in whichever IDE / terminal of choice.
There's access to all of Anthropic models via VertexAI, but will normally use Sonnet for coding / code understanding, and Haiku for Atlassian (Jira / Confluence), and Context7 MCP "agents"
Personal is:
Gemini / Github Copilot
NotebookLM is awesome for stuying. Creates some cool podcasts out of your notes.
Ympker's VPN LTD Comparison, Uptime.is, Ympker's GitHub.
Gemini when I want to be lied to constantly and then gaslit about it.
Whichever one is responsible for buying up all the global NVMe stock and ruining the LE market for the rest of us. 😂
Jokes aside, mostly ChatGPT. It basically acts as my junior sysadmin intern. I ask it to write obscure iptables rules or Docker configs, and then I spend 10 minutes fixing its hallucinations.
Mostly ChatGPT.
Host-C | Storage by Design | AS211462
“If it can’t guarantee behavior under load, it doesn’t belong in production.”
I used ChatGPT during 3 and 4, but 5.0 really broke a lot of shit, 5.1 broke what I fixed for 5, so by the time 5.2 came out, I gave up. Just Claude right now. We have Gemini for work, but fuck that. Every once in awhile I'll troll through Grok to see what it can do, but Claude writes mostly-usable code.
My pronouns are like/subscribe.
Hated 5.x so much, it actually made me angry.
Claude seems really good with code. But I sent it a picture of my dried out dying cactus and it kept trying to gaslight me into thinking it was a mushroom. Even after I told it it was definitely a cactus. It kept breaking the news like I had a bad illness: "I don't know how to break this to you.. but that looks like a mushroom". It was not a bloody mushroom!
Michael from DragonWebHost & OnePoundEmail
Claude is good when you tell it with no uncertain terms to not guess or overcorrect when told that it's wrong. I'm still learning it- only been using it about two weeks as my primary. Grok is brilliantly braindead. It's either all in, or restricted to the point of uselessness after raping Will Stancil, I guess.
OpenAI has done itself a huge disservice with the 5.x branch. 5.x wouldn't even assist me with study because ESV is copyright (even though the copyright is very wide), and absolutely refused to assist me despite acknowledging that I was correct and it was wrong.
My pronouns are like/subscribe.
Recently got yeeted out of perplexity...was using a ahem grey market plan. And just realized I never needed it. Had a google workspace plan (to use custom email domain)...and that includes the fancy version of gemini chat and notebooklm. Could have just used that all along...
I was using CC initially for about a month. It's solid but didn't like some of the choices the anthropic lock in forces on you. e.g. the hardcoded websearch that overrides MCP. They both do the job I guess.