New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Which AI tool have you been using the most lately?
ReliableSiteHosting
Member, Patron Provider
in General
Comments
Flagged the other comment, hope he gets banned.I found ChatGPT an unrealistic AI. I have tried it.For coding, I use Claude and Cursor.
For studying (yes, I always use it for that), Gemini.
I am very sorry, didn't mean to insult someone. Won't do that again.
My apologies, I made a mistake.
use mostly Gemini and Claude code with z.ai as Provider
Claude and OpenClaw primarily. OC has been the best virtual assistant one could hope for. You gotta keep it on tight rails though and you can't give it direct access to important data (but you CAN build a bridge to give it obfuscated data for more important things).
We're using GPT-OSS 20b to roll out AI spam filters soon. We've got an incredible protection layer on each end of it to tightly control the input and output.
I'm also working on a system where suspected compromised accounts have their logs obfuscated to hide PII or any sensitive information, sent to AI for a judgement call with a unique generated ID. Then the AI sends the decision back to the API which pulls the unredacted logs and messages me (without AI) to make a call on suspending the email account and opening a ticket.
AI is the future of providing low cost services, but you have to treat the AI as hostile. And it has to augment the human, not replace.
Candy.
Please ban openclaw.
Gemini.
VS Code with Copilot, since i have free credits there for each month.
And a bit of ollama sometimes, although sad to say - self-hostable open models have fallen behind massively recently.
Other than this I don't use many AI tools. I mean I do access models via web interface, but does that even count?
Oh, and I'm @DediRock customer and serververify contributor @jbiloh .
As more and more tech firms go from encouraging AI usage to actually enforcing it those in tech who are able to choose which AI tools they use and how heavily they use it will soon become a small minority.
Tech firms want workers to train their AI models and now actively track AI adoption giving poor performance reviews to anyone who does not meet their demands. Anyone who fails to follow the firm's guidance on AI usage won't survive long term.
Those who follow the firm's guidance on AI usage won't survive long term either because they will all get fired as soon as the AI model reaches a level where it no longer needs their input.
Really cool usage, a fan of mxroute since a long time
... have you tried gpt oss safeguard 20b? If you're running it on a local inference engine fine-tuning it with Unsloth even a QLoRa could make it even better for this sort of thing!
Last month I used the $200 Claude Code subscription, even then in busy weeks I ran out of usage. This month I've been using the $200 Codex subscription and I'd dare to say it fits better my workflow, although I still prefer the "vibes" of Claude Code by a lot. Even better is that since they have 2x the usage right now I can code away without worrying about usage, but after usage goes back to 1x idk, we'll see if I switch.
As for some of the work I do which is building custom agentic harnesses, I've had a lot of luck with Gemini Flash Lite 2.5 for speed and decent intelligence, but Gemini Flash 3 Preview and Kimi K2.5 are outstanding contenders. I've also tried Kimi CLI with Fireworks as the inference provider but it's not as good as the Codex or CC harnesses sadly.
For professional coding: Gemini.
For creative writing, digital arts, life advice: AI Mode; see best of "yoursunny lore" by Google AI and Freaky Fast Digital Coma art piece.
Yes we know they are the same thing with different skins.
Gemini allows resuming a session; AI Mode is lost when closing browser but it works anonymously.
I use DeepSeek very occasionally if I forget a word or term. That's pretty much all.
The tool itself doesn't matter. What really matters is how you put it to work.
Gemini mostly
Mistral API + Le Chat (Underdog but its really good
)
NotebookLM is awesome for studying. Creates some really cool "podcasts" out of your notes that are easy to listen to.
Mostly ChatGPT, sometimes claude, Antigravity now and then, for work and personal projects.
Just me and Gemini's free tier, teaming up to fix my grammar.
My use/rankings based on personal experience.
Grok = Emotional Support | Shit coding.
ChatGPT = Structured Thinking | Great Coding.
Claude = Better Coding than ChatGPT | Better emotional support than ChatGPT | Grok allows more adult emotional support topics.
Cursur = Better or same coding as Claude.
Github Copilot = Coding in-between ChatGPT and Cursor.
API from all companies have their own rankings, because they are different from their online interfaces
Another Input:
Many users who use online interfaces don't realize you can use custom instructions, which alter the AI LLM's output to match your needs better like teaching/coding/meditation (if it has voice)/and so on. Sometimes you can misalign from intended companies guardrails, but that is on you and can get you into trouble.
claude code in terminal using zed
Mostly Gemini Pro (while I still have the free 12 moths).
Coding with local models I alternate between Qwen Coder, Deepseek Coder and Codestral (all capped with 16Gb VRAM models)
Antigravity with Gemini and Claude
OpenCode with GLM and Kimi
I’ve never used Antigravity. It looks like a neat IDE, and I’m curious. Why do you use Gemini + Claude instead of just one or the other?
SillyTavern. I find the process of creating AI personalities and developing what they can do (with the limits of current tech) fascinating
On the backend, OpenRouter + either DeepSeek 3.2 or GLM 5.
I'm curious too about antigravity, been using z.ai and copilot (copilot is better bang for the bucks imo). In your experience of using antigravity, is it cheaper/better than Copilot for coding needs @Patriarch ?
You might like this project, https://lite.koboldai.net/.
It allows for most custom AI personalities in many formats, like SillyTavern.
It uses a database of linked users who run AI from home or servers.
VS datacenters.
All of them with this tool: https://app.chathub.gg
Which AI agents have been using you as a tool lately?
You think an AI agent is the tool when in fact you are the tool. Those agents are being trained on your input, you are the tool helping them improve.
Remember that brief period of time when everyone worked for big tech, for free, by solving captchas? Well now we are working for them again, for free or even paying them, by training their AI.
So the real question is:
Which AI agents have been using you as a tool lately?