Jared Tyler

Here's a story about how my latest project gave me the exhilarating rush of cash flow, only to deliver a heart-wrenching setback as I watched it crumble in a single weekend. The idea: I was inspired by all the GPTs out there that took a user's image and created a new image based on the original using DALL-E. And with the holiday's approaching, I had the perfect idea: transform family photos into pixar-style animated Christmas photos. Execution: I used GPT-4V and DALL-E to generate the photos. I used @TallyForms to collect the user photos and integrated it with @make_hq to generate the images and send them to the user upon completion. The workflow worked great. But I was running into one issue: GPT-4V seems to be significantly more limited in describing photos of people than ChatGPT and that was preventing me from generating accurate photos. I wasn't going to let that stop me, so I decided I'd generate the photos myself with ChatGPT and worry about automation later. I used some stock photos to generate a few example photo transformations and then I uploaded it to Etsy. I threw a budget of $5 at Etsy ads (which is pay per click) and waited... Results: Within 24 hours I had 4 paying customers at $7.99 each! I was absolutely thrilled! Thoughts were racing through my mind about automating this process and bringing in thousands as I scale it up. But I needed to focus on getting those 4 customers their generated images first! I got to work with ChatGPT and a custom GPT I made for this task. And it worked... kind of. It was actually quite reliable with the stock photos, but with these more complex customer photos, the results were less than impressive. Unfortunately I began to learn that even though DALL-E is great at generating images that closely resemble a prompt, it is far from perfect. It didn't really matter how detailed my prompts were, it always messed something up. And the perfectionist in me did not want to deliver a product to the customer that didn't really represent the original photo. I worked tirelessly at prompting and tweaking and editing until I had good-enough results. I got the customers their photos and I shut the listing down. It wasn't feasible for me to spend hours upon hours generating a single photo. My takeaways: 1. I should have started with a simpler MVP. The Tally form with Make integration was pretty simple, but still took a number of hours to setup. But it didn't matter because I ended up just putting it on Etsy and generating the photos manually anyway. This is how I should have started. 2. Test more rigorously during the initial stages. I could have saved some headache by convincing myself that this tech setup wasn't worth the time by trying some images that were more complex than stock photos. 3. I should have started weeks earlier. I started this in beginning/mid-December, but had the idea weeks earlier. I should have executed right away considering that I only had until December 25th to execute this with the Christmas theme. 4. Etsy is actually a decent way to get some initial traction for digital products like this. I was shocked at how fast I got orders without a reputation on Etsy. It's super disappointing that I had to shut this idea down when it was just gaining traction. But I learned a lot and I think this idea is still valid. It could definitely be executed with something like stable diffusion and Dream Booth or something similar, but I didn't think that was worth pursuing halfway through December. I'm very excited to take these learnings with me as I #buildinpublic for my next project in 2024!

Whoever builds an app using AI to make understanding technical specs easier will make MILLIONS. Currently digging through an 800+ pages of the PCIe spec to find one key piece of info. A chat interface that knows everything about this spec and could point me to the location in the spec that answers my question would be incredible. The current ‘ask a PDF’ solutions don’t really cut it when it comes to super technical stuff and the need to look at diagrams and tables

Life is not guaranteed, don’t let anything stop you from doing what you want to do

Lately been using n8n for prototyping backends. It’s pretty great at connecting things together quickly and running through a workflow with different APIs. I’ve found that you still need a bit of code in between for more complex applications and you also need to have a developer mindset and understanding of APIs, JSON, etc. But I like the tool. Also used Bubble for full stack in the past. Works well, but I felt like I was fighting the tool at times

What was your experience like with no-code? I want to write an article about it. Would like to get different point of views from the community. If you've used a no-code tool before, please let me know.

First time using @supabase. Awesome tool! One dumb complaint: The data type int8 is 8 bytes. In my world of low level computing, an int8 is typically an 8 BIT integer. I was quite confused why I couldn’t have an integer larger than 8 bits for a good minute haha. Guess I need to get out of my low level world once in a while

Just listened to this MFM episode. They talked about why Open Source is game changer. I couldn’t agree more. As someone who works for a company where security is our #1 priority, we will always take an open source tool over closed source. We can ensure the code isn’t malicious and that no data will be leaked. That’s a huge deal for: defense companies, hospitals and healthcare, law firms, and even education. I can see a future where all big SW products have a competitive open source alternative

AWS Lambda functions are awesome. Easy to integrate into whatever system you’re working with + I get to use python. But handling packages and dependencies in Lambda sucks. It’s seriously a pain and often doesn’t work without odd workarounds

@_david__wright_ I like to do woodworking when I have spare time. Don’t get to it as much as I’d like though

What's your hobby? Do you have a passion that isn't your side hustle or code project?

When’s google gonna pull through with the Google sheets/docs/slides AI integrations? Microsoft is giving them a run for their money

@elvissun Well done! Very interesting move sending out an email with stripe notifications to push demand and help with social proof. Never would have thought to do that!

Yesterday I hit the $1000/day milestone with PressPulse. Here's how I did it and 2 big lessons I learned: ✨ The story I launched in October as an AI that finds PR/backlink opportunities for you and steadily grew its MRR to $400. Last week, my second-ever user churned. Digging deeper, he mentioned it was because he didn't have time to pursue the opportunities our AI finds for him. He'd pay 10x if we could just solve the problem for him. So, I decided to run a totally different experiment: a fully done-for-you service, priced at $237 for a DA40+ backlink. I put together a quick form and Stripe payment link, then sent it to 300 people who had previously tried my AI but didn't convert. The result was astonishing; 2 people paid within 5 minutes, and 2 more shortly after. 5 hours later, I sent another email titled "One spot left," with a Stripe screenshot of the 4 payments. I hesitated on the second email but it ended up being a great decision: - The last spot was taken within 5 minutes. - I got 4 more inquiries after raising the price to $347 in the email link. - The real-time social proof gave users the confidence to trust me. Without the screenshot, some might think I was bluffing about the sale. Overall, the 5 sales and one annual subscription (another experiment I tweet'd about last week) brought in $1365 in one day. ✨ Lesson 1- The power of an email list Russell Brunson often talks about the power of an email list. I didn't fully grasp it until I experienced it firsthand. You don't own the traffic on Google Ads because you have to pay for every single click. You don't own the traffic on Twitter, Instagram, TikTok. Your reach depends on the algorithm overlord. But you own your email list. You can email them anytime you want, and it's guaranteed to show up in their inbox. In this case, I had a list of 300 people who had previously tried my SaaS. They went through the free trial but didn't sign up. However, they are problem-aware and actively seeking a solution in this space. I was able to put together an intake form in a couple hours and sell it to these people thanks to my email list. This wouldn't have been possible without my own traffic, without spending a dime and on my own schedule. That's the power of an email list. Even an email list of 300 people had the purchasing power of $1000/day. Imagine 30k. Biggest lesson today - Build an email list and nurture the people on it. ✨ Lesson 2 - The power of stacking offers I first heard of the idea of stacking offers in Justin Welsh's newsletter. I didn't remember the details, but the concept struck a chord with me. Because this approach not only maximizes the lifetime value of each customer but also introduces efficiency in customer acquisition costs (CAC). In my case, a free trial of my core product became the lead magnet for the premium, done-for-you service. This synergy between offerings not only enhanced the appeal of our suite of solutions but also paved the way for more sustainable, cost-effective growth: I could potentially reach a much lower or even negative CAC on my core SaaS product. Second lesson, think about other ways you can solve the same problem, and sell to the same users.

@d4m1n Ooo this should be interesting. Can’t wait to see the outcome

I KNOW this is a problem. 📈 191 people have pressed on "buy" 📉 20 orders made it through Posthog showed me this. And of course, if you are set to pay $87 and all of a sudden have to pay $105, you'll drop. Of course, there are other factors, but I bet this is the biggest one! CAN'T WAIT TO SEE THE RESULTS 🔥

@rokasdam @supabase Just started using supabase the other day. Didn’t even know about this AI function. Great idea, thanks!

Seed your Supabase database with AI 🤖 💚 A common problem when starting a new project is empty database. In order to combat that we can use Supabase AI to seed the database with initial set of data. How to do it? - Go to Supabase dashboard - Go to SQL Editor - Ask Supabase AI to create a table - Ask Supabase AI to create sample data based on that table That simple! 🤯 #buildinpublic

Oh man this is cool

@TheJackForge Agreed. But you’re forgetting the most important of all: scrum masters

@carrd is honestly so underrated. $49 a year, ~$4/mo, and you get can build 25 sites. For landing pages, why not use it? It’s so easy and so affordable. My next project will be shipped on a Carrd landing page

The mac virtual display on the Apple Vision pro looks incredible. Makes me want one

@MaxPrilutskiy Been pretty rough to start. Can’t seem to find my focus after a full day of work, but I’m making progress

so it's january 31st 1/12 is now gone how are your yearly goals doing? 😏

My job involves extremely low level programming with no operating system. I asked ChatGPT to give me code to read a segment of memory given start and end address values. It took some serious convincing for it to do that. It just kept saying you should be careful when referencing memory directly. I KNOW THAT. Just give me the code man

AWS Lambda can be a real pain when it comes to adding additional libraries. Note to future self, for the requests python library: “pip install requests “urllib3<2” -t . —upgrade” It’s the only way I can get the requests library to work on Lambda. Frustrating, but glad it works

@matifanger A weird complicated mix of both. That’s what my employer seems to think anyways

Agile or Waterfall methodology?

@malenchi_alex LP!