Parker Rex
All videos
Parker Rex

Every “Top AI Coding Tool” List Is Wrong — Here’s What Actually Works

May 4, 2025

Watch on YouTube@parkerrex

I spent a week and a lot of money trying every AI coding assistant, then boiled it down to what actually moves the needle. The takeaway: pick tools by where you are in your workflow, and don’t let AI replace your core skills.

Key takeaways#

  • AI tools are multipliers, not replacements. They fit different parts of the software delivery cycle.
  • If you’re learning, experiment with tools to feel the surface, but for production SAS, you’ll eventually need a lean, coherent flow.
  • Favor a minimal, effective setup: Cursor (for Q&A and exploration) and Warp; Taskmaster is optional if you want to build the core skill faster.
  • Writing prompts well is the real skill. Your speed and correctness come from prompt craft and understanding the process, not just the tool.
  • The trend is toward a one-stop shop in the long run, with ideation and QA becoming the differentiators.

Tool alignment with the software lifecycle#

  • Ideation and PRD
  • Use AI to brainstorm problems, shape requirements, and draft PRDs or one-pagers.
  • Architecture and implementation
  • Tools like Taskmaster can help bridge PRD to concrete implementation and maintain context across steps.
  • QA and testing
  • QA remains a critical skill; AI should augment you, not replace rigorous testing and validation.
  • Deployment and maintenance
  • Automation and tooling support scale, but human oversight is still key for correctness and product fit.
  1. Define the problem and create a PRD/one-pager (don’t skip this step).
  2. Draft the PRD/FAQ and supporting docs (six-pager, PR FAQ) to set guardrails.
  3. Use Cursor for questions and rapid validation of ideas, staying within your existing stack (TypeScript primarily; Python for agents when needed).
  4. Use Warp for quick code generation and exploration, but don’t rely on it to learn fundamentals.
  5. Iterate: test, refine requirements, and QA thoroughly.

Tips:

  • Focus on prompt-writing as your core lever. The actual coding speed comes from how you prompt versus which tool you use.
  • If you’re already competent, you can layer tools to accelerate, but guard against over-reliance (muscle growth matters).

Skill-building vs shortcuts#

  • Shortcuts can accelerate you, but they can also atrophy core skills if you lean on them too early.
  • Cursor is your “new raw-dogging” approach: ask smart questions, stay aligned with your TypeScript perspective, and watch where you’re learning.
  • Regular practice with prompts, problem framing, and QA will keep your skills sharp even as AI handles repetitive tasks.

End-state and mindset#

  • The endgame is moving toward a one-stop tool with strong ideation and QA workflows.
  • Testing will be paramount: you’ll need solid guardrails (PRDs, six-pagers, FAQs) to ensure outputs meet real needs.
  • Adoption is inevitable; the question is how you structure your own workflow to maximize learning and output.

Community and next steps#

  • Consider joining the group for AI, product, and code integration; weekly workshops and a collaborative environment help you level up faster.
  • The group currently offers half-off access and a focus on practical applications across product, marketing, and code.

Actionable takeaways#

  • Pick tools based on your current stage:
  • Early learning: experiment with Cursor + Warp to gauge what’s possible.
  • Building production SAS: define a clear PRD/FAQ, then use tools to support execution, not replace it.
  • Invest time in prompt engineering. It’s the most scalable skill you’ll develop.
  • Create the project scaffolding early: PRD, six-pager, PR FAQ, and clear acceptance criteria before heavy coding.
  • Don’t overlook QA. Make it a core part of your process from the start.
  • Cursor - AI-powered code editor for Q&A and exploration
  • Warp - Quick code generation and terminal exploration
Transcript

I spent a week and a lot of money on tokens trying out every different AI coding assistant. Now, if you've seen any of these videos on my channel before, you know that I'm full-time building SAS, building community, and also figuring out this AI marketing space. So, I need as much leverage as I can possibly get matters to me. So, it was really frustrating when I spent a week trying different ones just to realize I should have stuck to what I know. Let's talk about it. So, I made a nice deck for our coding workshop this week. And this pretty much sums up the whole video, and I think it depends where you're at. But in my case, I am competent in the text stack that I use. So, I'm a TypeScript stan, but I'm also now scaling up in Python because it's necessary if you want to build agents. It's just every everything in there, it's just better, right? Even the frameworks, they don't even exist in Typescript yet. So, what I what I'd say with just tools in general though is being realistic on where you're at. So, if you are just getting into this, then I think it makes a lot of sense. If you're trying to learn how to code and you got the dopamine from using these tools, then that's great. If you're trying to build production SAS software that will be for paid users, then I think at some point you'll inevitably make the leap from the left to the right side. I think at the peak of it, this can also be a little too extreme where it's like you don't want to use Vim just to say you use Vim, but it does make you learn because there's just a lot of stuff in there. What's a buffer? How are these motions working? How can I move quicker? how can I start writing bash scripts to help me with random things? And there's practices that you can put into place as well around that where you don't want to be using AI for everything. But on the tool side, what I did find was each one kind of maps to your traditional process for the software development life cycle. And I think a lot of the vibe coders that came into this space, they don't realize that they're just building toys for the most part or internal tools. And there are definitely outliers where you can build something that has utility and does build value. So, I'm not going to hate on that. But on the majority, I think you'll end up having to kind of learn all these steps. in that AI is just a multiplier on the existing steps that you take to build software because we've gone from lean to or we've gone through agile we've had or sorry waterfall and lean and agile and the Google 5day sprint and then shape up and shape up is the one that I resonate with the most the Spotify pod system all those but at the end of the day you're still doing this and I think that a lot of the people will just focus on ideiation or sorry just focus on the PRD and maybe the idea isn't that great because they didn't focus on a problem, they focused on the solution. And I talk about this a lot in our community. I even made a whole course about this. Yes, I made a course, but this is like four years ago and I felt so cringe. But made a whole thing about that because when we were scaling to our peak was 73 million in GMV per year, you have to be super diligent about what it is you're building and then you shape it up and you make the PRD. There's other ways of doing this, but each one of the tools I find maps to one of these. So in your case with taskmaster kind of helping bridge the gap between PRD and implementation and with client you're kind of doing the job of the architect in the case that it's keeping the context of what's going on in its head. Whereas you normally in real life would have an architect that kind of knows what the tech stack is and know we have GraphQL over here and we've got this other thing over here and we need to kind of think how this puzzle fits fits together. But if you just want the the end state is it basically that I still think that if you're going to pick a tool or other coding assistants that taskmaster and cursor are the way to go and then taskmaster is honestly optional if you want to get better at the core skill that it's replacing. And I find that when I'm working in Python cuz I'm a junior in there that I don't want to use a tool. I just want to have cursor. I just want to be asking questions. I just want to be focused on how do I think about this through the lens of my TypeScript skills so that I can get better and then I can bring this in to help. I probably would move quicker to get to the end state, but I wouldn't be learning. There'd be no muscle growth. It'd just be atrophy honestly. And so I think that when you're think when you're looking at all the different tools, they all fit different places. It comes back to intent, but I would just be focused on getting better at writing prompts and then only using cursor and warp. And I have another video on my channel, but this is sort of the new flow that I'm going through which is just it's funny now like it used to be that you were raw dogging code if you didn't use an assistant. Now everyone does, right? Like it's just kind of the understood you that you're going to have this year. But in my case, I think the new raw dogging is like you just use cursor. It's like, yeah, kind of. Like, it's it's really good if you know how to write prompts, if you know how to take the existing steps that we've done for a while and then build prompts around them. So, that's the gist of the of the whole video. I'll cut it there that you just want to pick the tool when you need it. I think if you're competent in the area, then use a bunch of the different tools. Hopefully someone can prove me wrong and say that no, we figured this out. But I think endgame is actually that you'll have a one-stop shop like a Devon or something and then maybe the only real skill is in the ideation and then in the QA because testing is going to be so so important because you'll just put in your shaped up idea and your PRD and the expected outputs maybe even the Amazon six pager and a PR FAQ all of those supporting documents and And QA will be I mean QA honestly used to be kind of for like the worst engineers. Sorry if you're watching this, but that's at least how our team worked and that was the vibe I got from meeting other teams in San Francisco. But it's definitely a vibe shift. So I hope you find this video helpful. And if you're interested in joining the group that we have, you know, because AI it's either you use it or it's going to come after you. We share a bunch of stuff in there. It's half off right now. We do weekly workshops. I'm definitely not the smartest person in that group, which is is fantastic because you you want to have people that are smarter in your group. And we cover product, we cover marketing, we cover code. And I hope to see you in there. And I'd like to give you a cheers as you like the video and subscribe. And I'll see you in the next one.