My general sentiment when it comes to AI, is that I am behind and everyone else has already made life changing improvements with it.
I scroll LinkedIn, Reddit and Twitter (only for the AI posts, I promise) content regularly. And on the one hand, the inspiration for AI related capabilities is incredible, on the other, it fuels a constant sense that I am an AI laggard.
So with that being my persistent mindset for the past few months, I spent a few hours this weekend in Claude Code.
Prior to this I'd never used it. I'd gone only as far as to click the Code switch at the top of the desktop client. No further.
This weekend however, I ended up spending 8+ hours playing around with building an idea I had for an app that could ingest a basic level of employee data, and help architect a job level framework.
Here is the lifecycle I went through, and the rollercoaster of emotions/focuses I had in that 8-hour period.
Getting it setup.
I spent a few minutes setting up a Github account (a space where the outputs from Code are stored), connecting it to Claude, setting up a project and installing various bits of coding stuff necessary to get started.
I say stuff because I don't know what most of it is (Node.js? NPM? Claude terminal?).
Initial frustrations
Some of this experience still felt a bit inaccessible to those who don't know anything about coding (me), or don't have the persistence to push through obstacles (not me, I am stubborn about an outcome, especially when I feel stupid like with this).
For example, I started coding into Github, but then there was no way to 'preview' the web app in Code (like all the guides suggested there would be). So I ended up having to download all the code, and start coding locally (into a folder), and actually run it locally, too. This took me probably an hour to sort out and even chatting with Claude chat about it didn't seem to resolve the issue for me. Annoying.
Then I ended up working in Claude Terminal instead of Claude Code, and I started to feel like this guy below.
AI Giddiness
Once I was past that and was actually interacting with Code in a way that I could see in my browser, the giddiness set in.
Don't get me wrong, I've played around plenty with Lovable, Replit and Google AI Studio, but there was a whole other level of intuition that seemed to come from my work with Claude code.
As someone who has always fantasised about being a tech entrepreneur (but was never technical), this kind of tool genuinely tears down the biggest barrier between normal humans and tech founders. It still blows my mind.
I'm in over my head
The thing that comes from building with AI though is that you quickly realise the limitation isn't the tool, but you as the prompter.
Once I came to this realisation, the shine of the basic prototype I'd built started to wear off. I realised it didn't have the maturity of something that would actually work at the level I needed it to. So I went back to the drawing board.
I sat down with Cowork and the Project Management Skill and started to more properly map out what I actually wanted this tool to be. I got it to ask me questions and after about half an hour we had built a whole product roadmap that would take the tool from the basic concept I had made so far, into something that more closely resembled a workable tool that would displace a substantial (and manual) piece of work.
I am a big fan of MCP's and I use ClickUp as my project management software. So I setup a project for this tool build and got Claude to actually build the roadmap into this project for me (its honestly been one of the most game changing things for me this year).
Claude wrote way better product requirements than I ever could, and I just needed to review and adjust them before putting them into code to create the feature.
AI had now replaced the need for a product manager AND a developer.
Token optimisation
I pay $200 a month for the Max plan because I've been a power user of Claude Cowork for the past couple of months. So I knew I would have plenty of firepower available (i.e. tokens) to have a crack at building with Code. But then a few hours in, I got the dreaded 'you're at 90% of your usage' flag. To put it in perspective I haven't switched away from Opus 4.6 (the latest and greatest Claude model) ever in the time I've had it. I am always using the best model no matter what I'm working on.
At this point I was working on my two screens with four windows open.
ClickUp with the features I was building
Claude chat in the desktop app, to chat through issues I was facing or ask questions about something.
A browser with the app I was building.
Terminal window where I was prompting and engaging with Claude Code for the build.
It became clear to me at this stage that I would need to come up with a more optimised way of working that would reduce my token burn, especially if I was going to be building in a way that I had now admittedly become addicted to.
The utter coincidence at which, while this issue was occurring for me in realtime, a newsletter I follow closely in this space published a piece on this very matter. Well worth a sub if you don't already.
Where are we going as a function
All of this got me thinking about where we are as a function and where we're going.
If someone like me can pick up claude code and fumble their way through building an app to an (admittedly) decent level, pretty much anyone could now be building whatever they want for their teams and employees. We're no longer constrained by budgets, or procurement cycles, we're constrained by effective scoping and prompting.
While thinking about this, I came across a post that summed up the direction I think we're going really well.
So yeah, I had a pretty crazy weekend of fun that helped ease some of my AI shame. Although I'm sure there'll be another feature released shortly that will spike it again. I think the only antidote to that is just carving out a bit of time to play with these tools as they come to market.
I'd love to hear from you though - what are the imposter syndrome voices in your head telling you you're behind on AI and how could you ease them?