The demo of how OpenAI embeddings work previewed by Slobodan on Day 7 was encouraging. It didn’t take long to build and it was answering most questions we were asking about Vacation Tracker correctly.

Now that we’ve established that building what we want is possible, It’s time for us to map out what the full prototype will look like. And when I say full, I mean someone should be able to come to our website, sign up, import their help desk or other data source, and embed the tooltip into their product. They should be able to manage tooltips, see what kind of questions people are asking and improve the training for their product so that Knowlo answers questions better over time.

This will be the subject of today’s update along with an update on our search for a freelancer to help us build the MVP.

But first, Let’s talk about our move to WordPress.

Webflow to WordPress

As those who’ve been following our journey know, we’ve been using Webflow for our very basic website. But we decided on Day 6 that we were going to move to WordPress because It’s simpler, cheaper and because we’re familiar with it. WordPress is also the most popular website platform in the world which means it has way more themes, plugins and developers available than Webflow.

we’ve been dealing with servers for a long long time. So one of our main criteria for this website is that we absolutely DO NOT want to deal with servers. We want a fully hosted solution where we can pay a little bit more and outsource that headache. For this reason, we signed up for WordPress.com instead of hosting our own WordPress installation.

It turns out that WordPress.com is even less expensive than what I wrote in Day 6. It’s actually $54 CAD which is about $40 USD. For reference, Webflow is costing us $58 USD per month due to some sneaky pricing tactics. And if I didn’t say so before, we’re tracking all expenses in USD because we’ll be charging our product in USD.

Let’s get started:

• Step 1: sign up to WordPress.com. It was a few steps but very simple.
• Step 2: finding a template. I went to Themeforest to find a WordPress template for our new site. There are lots of websites where you can find WordPress templates (WordPress.org and Templatemonster.com are two that come to mind) but I’ve been using Themeforest for years and I like their templates. I searched for SaaS (Software as a Service) WordPress templates, found one that I liked without overthinking it too much, and bought it for $40.69 USD.

• Step 3: importing the Themeforest template into WordPress.com. I asked CofounderGPT how to do it and was told that It’s not possible to import themes for the hosted version, but it gave me instructions on how to import the theme if I wanted to host WordPress myself on a server. I happen to know that it is possible to import themes from Themeforest into WordPress.com, I checked that before buying the template. So CofounderGPT was a little confused on that one, but the steps it provided to import the theme into WordPress actually worked. For the rest of the steps, I followed the instructions which were provided with the template.
• Step 4: installing all the required plugins. When you install the template, it will need some plugins. For example, your contact form on the Contact Us page of your website requires a plugin to work. When the developers who built the template put it together, they used these plugins to create some of the functionalities of the website. You’ll see in WordPress that there is a warning after you install the template that some plugins are missing. For now, I just installed those.
• Step 5: importing the demo data. If you see a template on Themeforest that you like, chances are you want it exactly as you’re seeing it, not a completely empty version of it. This is why You’ll want to import the demo data of the site you saw and liked when you bought the template. There are usually instructions in the template documentation about that.
• Step 6: making adjustments to the site. I worked with WordPress for a few years a long time ago and It’s changed a lot in the last 10 years! But the documentation for the template is quite good. There are also lots of tutorial articles on Google and videos on Youtube. I basically stripped the website down to the Homepage and Blog pages, simplified the homepage, imported our logo, added some of our content and did some adjustments to the settings of the website. I got it to about 90% of how I want it to look like.
• Step 7: get a freelancer to polish the website. That last 10% can get really annoying if you’re not super familiar with WordPress. You can end up spending countless hours trying to get it exactly how you want. So I decided to get a freelancer to help me finish it. There are lots of WordPress developers, that’s one of the advantages of going with that platform. Most people will indirectly know someone’s nephew who knows how to build WordPress sites. You can also go on some of the sites that we mentioned in Day 6, there are many people to choose from there who can help. I just reached out to someone that I’ve worked with for years who is a WordPress master. He’ll import all our content from Webflow and clean the site up in the next couple of days. He agreed to do it for $200 USD.

All of this took about 5 hours to do. First mistake almost corrected.

Knowlo Prototype

Now we get to the fun part. When Slobodan was showing me how we can use embeddings to build Knowlo, we talked about what we needed if we wanted to actually pilot this with a few companies like Vacation Tracker. Luckily for us, we have some friends who also have SaaS products and we can easily get 2-3 companies to test this out for us.

Slobodan and I discussed piloting this in our company and a few others to start with. But in order for us to do that, we’ll need a lot more than the simple prototype we’ve built so far. So we made a list of all the pages we thought we needed and asked CofounderGPT if we were missing something:

So with this detailed prompt, the answer I got back from CofounderGPT was that it thinks we also need onboarding documentation/tutorials, integration instructions, help/support page, a feedback mechanism and a settings page.

Since this is a really crude pre-MVP version of the product that we want to build and test out, I think that the only two things that we absolutely need are some basic documentation and the ability to gather feedback from customers. After all, we’re rushing to get something out quickly so that we can create a feedback loop with potential customers as fast as possible. we’ll need as many ways as possible for them to share feedback with us. Besides these two items, we’ll tell CofounderGPT to put the other suggestions on hold for now.

we’re ready for CofounderGPT to write out what each of these pages will look like. we’re basically going to write some specifications but without the code this time. we’ll write the specs, define what the pages will look like and then send all that over to Slobodan and our freelancer to write the code with CofounderGPT.

Writing specs (again)

Let’s get CofounderGPT to explain what each of those pages will look like and how it will behave, here is the prompt I used:

With this very detailed prompt, CofounderGPT created a pretty clear outline of what the Sign up page should look like and how it should work:

Now we rinse and repeat for all the other pages we outlined above, including the two new suggestions from CofounderGPT. Not all pages are created equally though. Some are more complicated than others. For example, the import page and the tooltips pages need to be thought through and discussed with Slobodan before we finalize how they are going to work with CofounderGPT. So for now I just made a draft specification for those pages.

I won’t go through every page of the prototype in this post, but I do want to show how CofounderGPT and I worked together on a more complex page than just a signup page.

As we were working through all this, I realized that the “Analytics, reporting and training” page is trying to do too many things in one page. So I decided that “training” is important enough to pull out into a separate page, mainly because the training page is a way that users will be able to provide feedback to Knowlo so that it keeps getting better.

I don’t clearly see how this page will work so Let’s explore with CofounderGPT:

Without skipping a beat, CofounderGPT wrote the specs on how it thought this page should work. The specs were as detailed as all the previous ones but here is a summary of how the training page will work:

I spent about 2 hours working with CofounderGPT and we outlined what all the pages of the prototype should look like. The reason we did all this is so that we can create mockups to add to our specs. Then we’ll pass everything along to Slobodan so that he and CofounderGPT can write all the code that we’ll need to get started.

After writing out all those pages upon pages of text, where do you store all these documentation to share with the technical team? There are several ways to do this easily. The most crude way is to just put it all into a Google Docs or Microsoft Word document and send it by email to whoever you need to. We use Confluence in Vacation Tracker so I just created a new project and put everything into that project in Confluence. But Confluence is overkill to start, It’s way too complicated when you’re just testing an idea out. I hear Notion is great and not expensive. So it may be worth checking out if you’re looking for a tool to put your documentation into.

Freelancer update

We outlined the kind of freelancer we were looking for in Day 6. Slobodan and I reached out to a few people in our network and did a few interviews. We actually used one or two questions that were suggested by CofounderGPT. Finally, we’ve decided to go with someone that we’ve worked with for years in the past. The main reason is a combination of trust, seniority, and availability. For now we’ve agreed to explore this idea together and that there will be an availability of 5-10 hours per week to help us with this project.

Scoreboard

Time spent today: 10h

Total time spent: 51h

Investment today: $81.04 (for WordPress.com and the template)

Total investment: $288.54

Users: 0

Revenue: $0

What’s Next?

In tomorrow’s post, we’ll create mockups for all the pages we planned out today and we’ll find a template that we can use for our product which matches all the specs/mockups we created. And hopefully, we’ll make some progress moving our website to WordPress.