When the year 2025 started, I didn’t expect to use AI this much.
Almost by accident, AI started showing up in every corner of my workflow. Not just the obvious stuff like code generation or content help, but in much subtler, surprisingly helpful ways. OpenAI Pro, Cursor and other tools are increasingly part of how I operate.
It’s not about chasing the next model drop. For me, it’s about tool fluency — knowing which model to reach for, how to frame a good prompt, and where it actually saves me time or gives me perspective.
Here’s how I use AI in my day-to-day life — especially as a software engineer, indie builder, and someone just trying to get more done with less friction.
1. 🏗️ Cursor + Claude 4 + AI Dev Tasks for software engineering
Architecting a solution and software engineering are sometimes an art. Of course, software engineering principles, best practices and years of experience should be drawn from. There’s no substitute for that! But there’s an element of devising a creative solution that requires perspective and testing key assumptions that AI is very suited for. That’s where my setup of Cursor, Claude 4 and AI Dev Tasks come in.
Of all models, Claude 4 I keep coming back to for slightly trickier coding or architecture planning. It’s thoughtful, coherent, and has unpicked some very tricky problems (including bugs!). AI Dev Tasks ’ Cursor rules I use as my pseudo product manager and pseudo junior dev who I create PRDs and execute fully-fledged, working features with. If you don’t know about it, I recommend you check it out.
With AI Dev Tasks, the prompt starts with :
“@create_prd I want to add X feature.”
It doesn’t always get it right. When it fails, it’s usually because I didn’t give it enough context — or I didn’t fully understand the problem myself (a blogpost for another time). Also, if you have bad, smelly code in the codebase or the API/SDK you are using is badly documented, it won’t perform well or will get confused.
But when it works, it’s a massive time-saver.
2. 💬 ChatGPT for career coaching
I’ve worked with a great coach before and AI isn’t a replacement by any means! But ChatGPT has been excellent at helping me think about my career goals. It’s like having a sounding board that doesn’t get tired.
Prompt I use:
You are an expert career coach with a track record of advising seasoned XXXs, YYYs, and ZZZs with over a decade of experience. You deeply understand the transition from strong individual contributor to impactful technical leader, especially in ambiguous environments like startups and big tech. You are committed to helping me develop authentic leadership capabilities—not just by managing people, but by driving outcomes, building influence, and leading through complexity. You provide grounded, evidence-based advice to help me identify and pursue high-leverage growth opportunities. Your role is to support my progression over the next few years into a meaningful, XXX position.
It helps me get unstuck when I’m too in my head. Knowing how valuable a (great) coach can be, I recommend coaching in general. If you can’t find one, try an AI coach?
3. 🗣️ ChatGPT for language roleplay
I’m based in the UK. That means I don’t get many chances to actually speak Japanese day-to-day.
So as an experiment, I created a roleplay prompt for ChatGPT, where it pretends to be a Japanese coworker. I bounce my keigo off it. It’s consistent and responsive, and I can make sure I’m not getting rusty without relying on textbooks alone.
Prompt I use:
You are a polite Japanese-speaking colleague. When I write a sentence, please do the following in your response:
First, provide a natural, polite Japanese translation of my sentence. Use 敬語 (including appropriate use of 尊敬語, 謙譲語, and 丁寧語) based on the context. Do not translate literally—translate naturally as a professional Japanese speaker would.
Then, reply in Japanese as if you are my colleague. Keep the tone professional but warm, and always use appropriate keigo. Include natural phrases, acknowledge what I said, and continue the conversation in a way that would be expected in Japanese business communication.
Structure the response like this:
日本語訳：
[Your translated sentence with keigo and furigana]
返答：
[Your response as a colleague, fully in Japanese keigo with furigana]
Only reply using the above structure every time.
It’s not perfect, but it’s a great way to keep your languages from getting rusty. In a few years time, I wouldn’t be surprised if there were language AI buddies who you could speak to in real time.
4. 📝 ChatGPT for writing (with my own tone)
Everyone’s using AI to write now, but I’ve built a bit of a system around it.
I feed ChatGPT my style guide (voice, tone, structure, signature phrases), and then create with it. I use it to draft content, refine tweets, or shape product copy.
It’s like pair writing with a really fast co-writer who respects my vibe.
I use ChatGPT project instructions along these lines:
# 🧠 Content Writing Style Guide
_This style guide defines the tone, structure, and phrasing typical of XXX’s writing for use when generating content with ChatGPT.
Tone
…_
Then I paste the rest of the guide into the instructions. Each conversation in the project then generates content aligning with the style guide.
5. 🧾 Gemini for OCR + spreadsheets
I sometimes need to scan a receipt or screenshot and enter it into a CSV.
I once gave both ChatGPT and Gemini a photo of a spreadsheet and asked them to convert it to clean CSV. ChatGPT Pro got close. But Gemini 2.5 Pro nailed it. Barely any cleanup required.
If you're dealing with receipts, old docs, or screenshots, it’s a killer use case that not enough people talk about.
6. 🎞️ Playing with Pika for video gen
This one’s still early for me. I’ve been experimenting with Pika for AI-generated videos — mostly for demo purposes and fun product visuals.
It certainly has flaws that you want to be careful of in customer-facing flows, but it’s getting close. More to discover here.
7. 📊 ChatGPT inside Google Sheets (for enriching + writing)
This one feels like a cheat code.
I’ve started using ChatGPT inside Google Sheets’ Extensions > App Scripts — and the leverage is wild. Especially for things like:
Enriching data (e.g. pulling company type or sector from a URL)
Auto-generating short product descriptions
Cleaning up messy labels
Translating or rewriting copy at scale
You’ll need to know a bit of scripting (JavaScript knowledge helps). It’s fast, cheap, and gets me 80% of the way there without needing a full enrichment pipeline.
Ask ChatGPT to generate the App Script for you but be careful (!) to have the code vetted by a real software engineer. Sometimes the generated code can be dangerous.
You’ll need a lot more than this to make your Sheets do anything useful but here’s the code for making a request to OpenAI. Please don’t hardcode your API key and put them somewhere safer (like Script properties).
var apiKey = PropertiesService.getScriptProperties().getProperty('OPENAI_API_KEY');
var prompt = `
You're helping someone do XXX.
`;
var payload = JSON.stringify({
model: "gpt-4o",
messages: [
{
role: "system",
content: "You are a precise assistant. Respond with only with a JSON result..."
},
{
role: "user",
content: prompt
}
],
temperature: 0.3,
max_tokens: 50
});
var options = {
method: "post",
contentType: "application/json",
payload: payload,
headers: {
Authorization: "Bearer " + apiKey
},
muteHttpExceptions: true
};
var aiResponse = UrlFetchApp.fetch("https://api.openai.com/v1/chat/completions", options);
var json = JSON.parse(aiResponse.getContentText());
It’s not perfect, but for $20/month it’s incredible bang for buck. Especially when you’re solo or early-stage and every minute matters.
🧰 Fluency over hype
I think this is the real unlock.
It’s tool fluency.
Knowing when to reach for which tool. Knowing what to ask. Knowing how to layer it into your day in ways that actually remove friction.
As a indie maker (making Shopify apps) and software engineer, my day is full of small frictions:
Naming things
Drafting content
Debugging bugs
Thinking through UX
Deciding what not to do
AI doesn’t solve everything. But it helps with more than I expected.
If you’re experimenting with AI in your workflow (especially around Shopify), I’d love to hear from you.
I write more like this at logicalicy.substack.com. Feel free to follow me on Twitter: @logicalcy
Top comments (0)