How do you actually use AI tools in daily life and projects?

[m4b]

I’m curious about real-world AI tool usage. What are your typical use cases (work, study, personal productivity, creative stuff, etc.)? Do you understand the differences between models (like GPT-3.5 vs GPT-4, Claude vs Gemini, etc.) across tools? Do you switch models based on your use case, or just stick with whatever’s default? Would love to hear how people experiment and what patterns you’ve noticed.

 

[akiratoriyama]

Personally find GPT the most useful of all of them. Helps with health related stuff (have to crosscheck though), finance related stuff (CC optimization strategy, to get a card or not); find it especially useful when it comes to troubleshooting PC/phone issues, especially if you go really technical (gemini can be somwhat better than gpt in this though).


2 novel uses I have seen of it are uploading entire whatsapp chat exports and asking about relationship dynamics between you and the other person; or the group. Also, blind spots analysis

 

[m4b]

2 novel uses I have seen of it are uploading entire whatsapp chat exports and asking about relationship dynamics between you and the other person; or the group.

TIL! Interesting use case; I'm going to use it.

blind spots analysis

What is blind spot analysis?

 

[akiratoriyama]

The prompt mentioned in this.

 

[HumorSimpson]

For coding I found Gemini better than chatgpt.

What about you @Kushagra_ ?

 

[m4b]

asking about relationship dynamics between you and the other person

Can you provide me with the exact prompt if you don't mind?

 

[akiratoriyama]

Can you provide me with the exact prompt if you don't mind?

I just ask it to analyse the chats i shared, and not to miss any of the chats. I find it sometimes skipping the chats in between. Might have been fixed with GPT 5 but was an issue till 4o.

 

[akiratoriyama]

For coding I found Gemini better than chatgpt.

What about you @Kushagra_ ?

Yep, most of my friends in CS prefer gemini.

 

[HumorSimpson]

Yep, most of my friends in CS prefer gemini.

Only thing is , I had to correct myself for the pronunciation as J-mi-nai instead of what I have been uttering J-mini all these years.

 

[akiratoriyama]

Only thing is , I had to correct myself for the pronunciation as J-mi-nai instead of what I have been uttering J-mini all these years.

I still call it Bard occasionally lol.

 

[m4b]

Yep, most of my friends in CS prefer gemini.

For coding I found Gemini better than chatgpt.

Have you tried Claude 4 yet? Much more powerful than Gemini for coding. Gemini, in case you want to plan a big project or a very complex and critical feature that needs more reasoning and context.
Anyway, do you use Gemini web UI for coding or any agentic tools?

 

[HumorSimpson]

Gemini web UI. No agentic tools.

Do you use agentic tools ? Which one ? Review ?

 

[54R4NG]

Being a Project Manager I am filled with daily operations requirement which use to be manual or limited with PBI reports in the past.
In 3 accounts I am actively using Copilot for daily reporting, transcription, project SLA assessments, finding waste and creating six sigma active agents for project management. there are many more use cases, but these are bare minimum which most of the active PM(s) do on a day-to-day operations level.

 

[m4b]

Gemini web UI. No agentic tools.

Do you use agentic tools ? Which one ? Review ?

It's all about prompt engineering. I see a LOTTT of folks prefer the cursor over windsurf, and cursor PR reviews are on point, efficient over windsurf and copilot. Most of the time it's Claude 4. Windsurfing is also gearing up. Personally, I use Cline for coding, its open source. Need to buy LLM models though. Cline doesn't cap, unlike Cursor. I have created workflows, rules, and memory banks to automate my work.

 

[m4b]

It's all about prompt engineering. I see a LOTTT of folks prefer the cursor over windsurf, and cursor PR reviews are on point, efficient over windsurf and copilot. Most of the time it's Claude 4. Windsurfing is also gearing up. Personally, I use Cline for coding, its open source. Need to buy LLM models though. Cline doesn't cap, unlike Cursor. I have created workflows, rules, and memory banks to automate my work.

ofc switching between Gemini for longer context tasks and Claude for shorter context tasks.

 

[HumorSimpson]

Cursor and windsurf are code editors . SO one has to choose the underlying LLM beneath them ?

 

[m4b]

Cursor and windsurf are code editors .

These are not just code editors. They have their own proprietary system prompts and guardrails, which are injected for your query. They provide many features such as agent for planning, agen for task etc.
Some of the leaked prompts: https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools

On the top of it you need LLM models to actually use it.

 

[HumorSimpson]

Wow thanks @m4b . Good info. Will explore in the weekend if possible.

 

[m4b]

They are not just code editors. They have their own proprietary system prompts and guardrails, which are injected for your query. They provide many features such as agent for planning, agen for task etc.
Some of the leaked prompts: https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools

On the top of it you need LLM models to actually use it.

These are like billion-dollar prompts for real. The next gen is promt engineering; how can you efficiently command an agent to automate work.

 

[m4b]

Wow thanks @m4b . Good info. Will explore in the weekend if possible.

You can use open source models instead of paying for a model for small use cases, though you need good computational resources to run them offline.
With Cline, you can integrate https://lmstudio.ai/ to run em.