As we moved from a week full of testing capabilities and limitations, we tried to make the most of our limits. The week proceeded fairly well, until we ran into the weekend.

Having decided to tackle the brunt of the work on the weekend, my limitations didn’t really bother me throughout the week. We utilized Gemini mostly in the beginning of the week, setting it up to interact with our Google suite of tools. Towards the middle of the week, we began creating a load of transfer documents to maintain project coherency between AI tools. The end of the week, however, we were met with our biggest limitation yet:

How do we get iterative work done within free tier limits?

Unfortunately the answer is simple. We upgrade.

But I’m not going to wantonly throw money at an AI tool if I’m really not using it. Working on the website this weekend demanded it.

We ran into our chat limitations almost immediately into building the first few page blocks for the website. We barely touched on anything too complicated. What we were left with was a half hour of work and a whole day ahead of us.

But I had already spent money to reserve the domain name, so what was a little more? So I took things a step further.

I upgraded GPT to a plus subscription. GPT seemed to have the most severe limitations for free users. But Gemini remains functional through multiple iterations of images and project ideas that immediately were put into incubation for Project Daedalus to expand upon later. We have yet to hit any kind of rolling limit with Gemini even after building out concept art for the entire website. But upgrading GPT to plus has also allowed me to make the most of Gemini as well.

Functionally, I use Gemini to track financials, do research, and do iterative image generation from art I can throw together. I’ve been building all the logos and symbols based on one core piece of artwork that I created many years prior in GIMP.

GIMP I feel will be a very important tool for assisting image generation later. Especially when I incorporate more hand drawn work into this. But for now, I’m just maintaining that the entire design was born from a single original piece of artwork.

Enough about GIMP. More about DeepSeek. After uploading some project transfer documents as an experiment, DeepSeek immediately wanted to jump into building. This is good. I can use this later. For now I discovered the laser sharp focus of DeepSeek. And it didn’t require an upgrade because I really haven’t been using it much…

Yet.

Claude is the big test. It’ll be experimental when I have version 1 of the website built. I’m going to take the final architecture document that I generate and upload it into Claude to attempt to build website number 2 with as few errors and iterations as possible. We will push Claude’s free tier limitations when I get to that point.

But why then upgrade GPT?

Because I’ve used it the most frankly. I’ve built a lot of architecture with it. I’ve created a lot of artifacts. I’ve generated a lot of usable content and reference documents.

ChatGPT just felt like it took the most chains off of my workflow.

And that’s exactly what I needed. My entire workflow exploded since my decision to upgrade. I have the entire website skeleton built and ready to receive some organs and flesh. I’ve got all my Gemini image generations fully prepared for the website. I’ve locked in reminders and project focuses. I’ve refined some other project ideas that were in their incubating phase and moved those to active development.

A lot of projects still remain in phase 0 and phase null. But that’s okay.

But speaking of projects. Since I didn’t work physically on the website this week, what else could I have done?

We expanded Project Daedalus to include over 63 project ideas for development.

The idea generation was mostly done with Gemini as a limit test.

We never hit it and generated a ton of projects.

These projects were iterated upon over the week until we developed a full coding curriculum to push these ideas from a dormant place of incubation into reality. But Daedalus isn’t just covering project ideas. It covers active projects and ongoing experimentation as well.

I’ll be learning as I go and I certainly shouldn’t get bored, but with these new open limits and defined work spaces, we should be able to move into week 4 and get much more done. I don’t foresee the next week containing a lot of planning. There may be some organization, but I think I can rest assured knowing that next week is all about development.

That means the blog post may end up being shorter as well. That’s okay. I look forward to it.

I also look forward to continuing this experiment to see where the journey leads. AI provides a suite of new tools that make the interactive digital playground easier to access.

All we really need is a little bit of ingenuity.

And just an ounce of determination.

(Pictured below is the initial GIMP artwork of the Theseus logo that started it all)