Now that we’re half way through 2025, I’ve zeroed in on a set of AI tools and platforms that genuinely augment my work and, ones that align with the way I think, learn, and stay in control. Here’s a candid, personal breakdown of what’s working, when I use each tool, and why some just don’t stick.
OpenAI ChatGPT
This is my current go-to tool. It’s effectively replaced Google for quick information searches, explanations, and solving problems on the fly. I work with it like a real assistant: ask a question, get a solid response, and iterate when needed.
It’s always open on my desktop and on my phone. That constant availability is a key part of its value. One major bonus I didn’t foresee until recently is it’s phenomenal for learning new languages. You can carry on full conversations, in context, in real time, in the language you’re learning. That kind of immersive experience right from your pocket was a game-changer.
When I need to learn something fast, clarify a topic, or troubleshoot, ChatGPT is my first stop.
Anthropic’s Claude
Claude is my development “deep-dive” companion. It’s not a replacement for ChatGPT in that it’s a different animal altogether.
With a much larger token limit, Claude can handle long-form thinking without losing context. When I’m architecting a complex component, walking through logic, or drafting multi-step flows, I go to Claude. It lets me dump everything into one conversation and stay inside that mental space.
I use it exclusively for coding conversations and planning, not for quick facts or research. That clarity of purpose is helpful: ChatGPT for bite-sized questions; Claude for major architecture, building on thoughts, iterating, and coming up with something that’ll work for you.
Cursor
Cursor is now a regular in my day-to-day, both professionally and personally.
At work: I use it extensively with React.js, Next.js, Nest.js, and TypeScript. It fills boilerplate, suggests refactorings, and answers questions based on the code in my repository. I do turn off autocomplete though because I like to click around as I’m reading and throwing in autocomplete when I click somewhere is absolutely distracting.
At home: I tap into it for mechanical, repetitive tasks with things like Java, PHP, or SQL routines. Need to seed a database or convert a long list into structured arrays? Cursor handles that for me. Recently I used with Claude to build a new feature into one of my pet projects and it absolutely helped with a lot of the more repetitive tasks and made good use of patterns already in place to make good decisions.
The secret is feeding it context, keeping a developer diary that says what I’ve done, what I want, and how it responded. It sounds tedious, but the more precise and technical the documentation, the better Cursor works. With some guardrails, it really accelerates those routine tasks, but you always have to remind it to update the ever-growing documentation and then reference it back.
OpenAI Sora
Sora has been a delightful experiment. It’s how I dip my toes into visual creativity.
I’m playing with short‑form video and image generation just to visualize ideas that float in my head during the week. It’s not as structured as my coding tools, but that’s the point: no pressure and just exploration.
Sora is hit or miss (same with any generative model), but when you’re detailed with prompts, it can surprise you. It’s my creative playground for testing the boundaries of what’s possible visually.
Suno.ai
This is my unofficial replacement for Spotify. Suno isn’t just a playback service, it composes music tailored to my prompts. Describe the mood, the instruments, the tempo; Suno gives you a custom track.
I’m already curating playlists that match my working style, focus times, or even morning routines. It feels like I’m composing my own background music. There are some copyright and licensing questions bubbling under the surface, but so far, personally, it works. If legal obstacles don’t derail it, I fully expect to make the switch.
Google Gemini
I wanted to like it more, especially since it overlaps with my Google Workspace, but I ran into a serious user experience friction point: chat‑history control.
In Gemini, either chats are retained for a fixed period or not at all. There’s no granular, user-controlled cleanup. You can’t just hit “delete” on a thread when you’re done with it. That’s a strange design choice, given that every other AI tool I use gives me agency over my data.
I tried it, paused, and walked away. Full ownership and control are non-negotiable for me. Until that changes, I can’t make Gemini part of my regular toolkit. There are times when you want to be reminded of what you searched for, and other times when you want to forget something that either didn’t work or something that you no longer need.
Why It Works & Where I Draw the Line
| Tool | What I Use It For | Why It’s Great | What Holds It Back |
|---|---|---|---|
| ChatGPT / API | Quick answers, language learning | Instant access, always available, excellent for language practice | Conversation depth capped vs. Claude |
| Claude | Complex dev questions and planning | Long-context capacity, ideal for architectural thinking | For me, not suited for quick lookups |
| Cursor | Code generation and developer context | Boosts workflow with boilerplate and refactor suggestions | Needs structured logs to stay useful |
| Sora | Imagery and video concept creation | Fun creative outlet with surprisingly refined output | Inconsistent results, needs detailed prompts |
| Suno.ai | Personalized music | Custom-tailored soundtracks for different moods | Still navigating copyright concerns |
| Gemini | Workspace integration trials | Potential within Google ecosystem | No user control over history = deal-breaker |