I'm hyper bullish on AI. I'm that crazy AI guy who's made it his life mission to infuse AI into every facet of life. When I find an interesting AI tool, I buy first and figure out use cases later. I spend hours daily with AI tools, using them not just for productivity, but for overall well-being and knowledge enhancement.

As someone who's both a power user and creator of AI tools (Solvemigo, PenPersona, MemoryPlugin, AskLibrary), I have a unique perspective on what works and what doesn't. Here's a deep dive into my AI toolbox - hope it helps you build yours!

Quick Stats

• Total monthly AI spend: ~$400
• Most used tool: Claude
• Best value for money: Claude
• Must-have stack: Claude, Perplexity, Rosebud, Cursor (or Cline) (and of course, MemoryPlugin and AskLibrary 😉)

Core Language Models

Claude

Cost: $20 per month

Use case: Writing, Coding, General Advice, Large Document Analysis, Experimentation and Prototyping

Claude is my favourite AI model, by far. It is super empathic, highly intuitive, can read in-between the lines like a mind-reader, is super creative, very polite, and it just "gets" me like no other AI tool. Claude is my primary AI tool, I use it for creative ideation, for coding, copywriting, working with large documents (200K context window is great), or just for fun.

Claude is less strict about following instructions compared to ChatGPT though, like if you ask it to follow a very specific process or output format, ChatGPT will do it, Claude will try to "intuit" your end-goal and figure out its own way to do it, which might not be the way you wanted it. There are some use cases especially API integration scenarios where this can be a problem, but largely it doesn't make much of a difference and often Claudes way ends up being equally good if not better.

ChatGPT

Cost: $20 or $200 per month

Use case: Coding, Chit-Chat, General Advice, Research, Automation

ChatGPT with GPT4 was a quantum leap for AI coding assistance, but it was leapfrogged by Claude with a larger context window (32K in $20 plan for ChatGPT vs. 200K for Claude) and just better understanding of requests and coding ability. But for basic things, ChatGPT is still very, very good, and it's great at following instructions to the letter. The Advanced Voice Mode is super cool and fun. I've been trying their Pro plan and there are times when the o1 Pro model can be seriously very impressive, and Deep Research can do some very good, well, research, it's great to be able to delegate theses things to Deep Research. It can produce even 100 pages worth of text in its output. Operator is also very cool, although very rough around the edges still.

With the Pro plan (the $200 one), I'm trying to figure out how the reasoning models and agentic capabilities can be a step change for my AI usage.

Gemini

Cost: $20 per month

Use case: Large document analysis, analysing large codebases

Gemini is the worst AI model of the top three, by far. Google AI Studio is better, but the Gemini app has pointless refusals to the most benign queries, short-term memory loss within the chat, where it will tell you, aggressively, that you haven't told it something when you have. No other AI model does this, and Gemini can get very aggressive. If you point out a mistake to ChatGPT, it'll be like, "oh, alright, I'll change it", Claude will be like "oh my god I'm so sorry, I apologise, let me fix my mistake", and Gemini will be like "YOU did not give me enough information so you cannot call me wrong, you stupid".

The only reason I use it is for the giant context window, and to stay on top of what features they're releasing.

Development Tools

Cursor

Cost: $20 per month

Use case: Coding

Cursor is an AI native IDE. It has this amazing autocomplete that can complete large sections of code, if you rename a variable, it can find all the instances of it in the file and you just keep pressing Tab Tab Tab and it'll keep moving your cursor to the right location and making the change. It also has a memory, it remembers your recent changes, like if you rename an export from a file and open another, it'll guide you to the place that variable might be imported and used, two Tabs and you're done. Copilot is very far away from this.

They also have a chat feature and an agentic coding feature called Composer. With the Chat, you can have a chat with various AI models beside your code, and this is the flow I use most often. Composer is their agentic flow that can automatically apply changes across files, but I found it to be less precise and so I have mostly stuck with the Chat feature.

Cursor Tab is amazing. I was using their Chat feature quite extensively until I discovered Cline. Cline gets expensive but it's worth the additional bucks.

Cline

Cost: API costs vary, close to a hundred dollars a month for extensive usage

Use case: Coding

Cline is like Cursor Composer, but way, way better. See the thing with Cursor is that they give you everything for $20 per month. Claude 3.5 Sonnet, their most used model, is not cheap. The only models more expensive are literally o1 and Claude 3 Opus. So what Cursor Composer does, is it feeds only chunks from files, whereas Cline feeds entire files instead. This is more expensive, and iterating on a single feature for an hour can easily rack up close to $10 in API usage costs. But the way I think about it is, if I were to do that coding myself, well, there's no way I can type at the speed of an LLM (I'm a 100 WPM on my best day, Claude 3.5 Sonnet is ~3600 WPM), and if I were to delegate or outsource it, firstly it'll take days instead of hours, it'll cost me hundreds if not thousands of dollars, there will be more misunderstanding and back and forth, timeline overruns, so that $10 is incredible value.

Cline can also automatically correct lint errors, it reviews the file after every change it makes and auto-corrects any issues, it can use the terminal, it can use a browser, it can connect with MCP tools.

Cline is now my go-to AI coding tool, with Cursor kept around for the Cursor Tab feature and simpler changes.

Replit

Cost: $180 per year

Use case: Coding, Prototyping, making small apps to automate simple but loathe-some tasks

Replit is a great product with a terrible company behind it. My support experiences with them have always been disappointing. They'll run you in circles, take ages to respond, and ultimately not give you any resolution for your issues. I've been overcharged, suffered data loss, double-charged, yeah.

The only reason I got Replit is because of Replit agent, which can create full-blown web apps, frontend, backend, database, the whole shebang, from a simple natural language prompt. It can take your idea from a prompt to production (just keep in mind they don't back up the databases and the agent might just run DROP TABLE users; (real story, happened with me)). The concept is very, very cool, and I used it to make some tools that I really wanted, but there would be no way I'd do it on my own. Like I made an app that reads my emails and shortlists interesting ones for me. I've also used it to fix bugs in my apps from my phone and push the fixes to production, as it integrates Claude.

I just wish they had 100x better support, and did basic things like DB backups when you're selling something as capable of building production apps.

Copilot

Cost: $10 per month

Use case: Coding

Copilot is super disappointing. They introduced something revolutionary back in the day, but then let their product stagnate and it's totally superseded by Cursor for me. I just recently bought a new subscription to try their agentic features, but Cline runs circles around it. Yes, Cline cannot automatically open a pull request on GitHub, and cannot work from my phone, but that's about the only thing where Copilot has a leg up anymore.

Knowledge & Research

Perplexity

Cost: $200 per year

Use case: Search, research, debugging sometimes

Google search has been going downhill for a long time now. Remember site:reddit.com anyone? I had to resort to doing that for pretty much every search query. Then I switched to Kagi, but these days, I end up using Perplexity for most of my searches. It can compile information from dozens of links, and present answers. If I am not looking for answers specifically but want the links instead, I still use Kagi, but most of the time, I am after answer. It's helped me solve obscure bugs that both ChatGPT and Claude couldn't figure out, but some random (but amazing) blog had covered. I keep the reasoning mode on for most of my searches. I'm naturally very curious and like to soak up random knowledge about things, so Perplexity is excellent for feeding my curiosity.

AskLibrary

Cost: $10-$20 per month

Use case: Learning, personal and professional growth

AskLibrary is my own product. I had this huge and growing book collection and I knew I'd never be able to read them all. I've also had moments where a book has changed my life, multiple times. I don't want to miss out on those insights, and with AskLibrary, now I don't have to. AskLibrary lets me upload my entire book collection and ask questions to it. The answers are super detailed and well rounded, and backed by citations. A single answer goes through a 3-stage retrieval process that processes hundreds of pages through an ensemble of AI models. I recently compared other generic "Chat with PDF" tools, and AskLibrary's answers are far, far superior.

Notion AI

Cost: $96 per year

Use case: Knowledge management, documentation, simple autocomplete, chatting with my Notion workspace

Notion AI is, well, AI integrated into Notion. It does completions, like you can tell it to convert some information into a table, or finish the rest of the writing, or do some math, etc., you can search for pages in natural language or query your databases in natural language, you can have AI generated columns for your databases, etc.

TBH for me, not worth it, although it has improved by leaps since it first came out.

Personal Enhancement

Rosebud

Cost: $110 per year

Use case: Journaling

I've read so much about how great journaling is, it's often placed alongside things like meditation, but seeing a blank page and writing about my feelings or day in it, it's not very inspiring. I never could get a hang of it, until I found Rosebud. Rosebud is an AI-powered journaling app that you can talk to, and it remembers the things you tell it. I recently crossed a 365 day streak on Rosebud. Now there is a record of my days and how my feelings on things have evolved going back a year, and it feels like the journal "understands" me. I usually journal while taking a walk at the end of the day, and it helps me decompress, and I equate it to a mini therapy session every day. It's truly an amazing app.

Rewind

Cost: $20 per month

Use case: Memory enhancement

I loathe how low fidelity human memory is, and how unreliable it is. We will have cybernetic memory enhancements soon, but in the meantime, Rewind keeps a snapshot of everything I do on my computer, 24x7. So I can press a hotkey and find exactly what I was doing on a specific day, or I can ask it questions about when I did something specific.

It's abandonware at this point though, and I might switch to an alternative soon (screenpipe). I did order their pendant though and that's the only reason I keep my subscription going at this point. But it's definitely a superpower to have a tool like this and it's helped me countless times. Does it help me every day? Nah, and the novelty wore off quickly. But whenever I need to remember what I did for some specific task, I do use it.

Utilities

TypingMind

Cost: $99 one-time, $10 per month for cloud storage, which is optional, but if you are using it extensively, you will want

Use case: General AI chat, trying new models

TypingMind is how I started with my extensive AI usage. It's an API frontend for ChatGPT, Claude, Gemini, and almost any other model that's compatible with the OpenAI API. I've largely switched over to the Claude and ChatGPT first-party apps these days, as they offer more features, a better experience especially on mobile, and I got anxious watching the cost spike up in my chats, it deterred me from using AI quite as extensively. I calculated not that long ago that ChatGPT Plus, Claude Pro, vs API usage ChatGPT and Claude subscriptions (the $20 ones) offer value equivalent to hundreds of dollars worth of API credits.

If I hit my Claude limits and need to continue a chat, I export the data and switch to TypingMind, or if I want to try a new model, I use TypingMind too.

MemoryPlugin

Cost: $35 per year

Use case: Productivity, streamlining AI workflows

MemoryPlugin is also my own product. It's long term memory for more than half a dozen AI tools at this point. It lets your favourite AI tools remember things about you across chats and over time. The memory can be shared across all your AI tools, organised into Buckets to group and separate related memories, and there's new features in the works that will allow dramatically more information to be remembered.

Having to give AI tools all my context in each new chat gets tiring fast, it's slow, error-prone, and annoying. But now I don't need to do that anymore.

Conclusion

The AI landscape is evolving rapidly, and there are a lot of bullshit "courses" on AI tools out there. No, these AI tools aren't going to make you a millionaire overnight, but they can help you stay ahead of the curve in a rapidly evolving world with immense uncertainty looming about the future. The only way you can make the most of AI and be better off because of AI, rather than worse off, is by using these tools extensively, figuring out what you can delegate, what you can now do that you previously couldn't, what you can do faster, how you can be better, how you can get things done faster or cheaper or better. The possibilities are endless. So I leave you with this question: What's in your AI toolbox, and how are you cultivating it to improve your life?