Reflections on the closed company Hola! Dating app

I'll discuss my fuckups as an entrepreneur and CEO. All of them refer to the dating app Hola!, which I co-founded and starred in.

Spring 2021 was when we started. Two techies and two non-techies created a dating app. Pokemon Go and Tinder were combined.

Online dating is a business, and it takes two weeks from a like to a date. We questioned online dating app users if they met anyone offline last year.

75% replied yes, 50% sometimes, 25% usually.

Offline dating is popular, yet people have concerns.

• Men are reluctant to make mistakes in front of others.

• Women are curious about the background of everyone who approaches them.

We designed unique mechanics that let people date after a match. No endless chitchat. Women would be safe while men felt like cowboys.

I wish to emphasize three faults that lead to founders' estrangement.

This detachment ultimately led to us shutting down the company.

The wrong technology stack

Situation

Instead of generating a faster MVP and designing an app in a universal stack for iOS and Android, I argued we should pilot the app separately for iOS and Android. Technical founders' expertise made this possible.

Self-reflection

Mistaken strategy. We lost time and resources developing two apps at once. We chose iOS since it's more profitable. Apple took us out after the release, citing Guideline 4.3 Spam. After 4 months, we had nothing. We had a long way to go to get the app on Android and the Store.

I suggested creating a uniform platform for the company's growth. This makes parallel product development easier. The strategist's lack of experience and knowledge made it a piece of crap.

What would I have changed if I could?

We should have designed an Android universal stack. I expected Apple to have issues with a dating app.

Our approach should have been to launch something and subsequently improve it, but prejudice won.

The lesson

Discuss the IT stack with your CTO. It saves time and money. Choose the easiest MVP method.

2. A tardy search for investments

Situation

Though the universe and other founders encouraged me to locate investors first, I started pitching when we almost had an app.

When angels arrived, it was time to close. The app was banned, war broke out, I left the country, and the other co-founders stayed. We had no savings.

Self-reflection

I loved interviewing users. I'm proud of having done 1,000 interviews. I wanted to understand people's pain points and improve the product.

Interview results no longer affected the product. I was terrified to start pitching. I filled out accelerator applications and redid my presentation. You must go through that so you won't be terrified later.

What would I have changed if I could?

Get an external or internal mentor to help me with my first pitch as soon as possible. I'd be supported if criticized. He'd cheer with me if there was enthusiasm.

In 99% of cases, I'm comfortable jumping into the unknown, but there are exceptions. The mentor's encouragement would have prompted me to act sooner.

The lesson

Begin fundraising immediately. Months may pass. Show investors your pre-MVP project. Draw inferences from feedback.

3. Role ambiguity

Situation

My technical co-founders were also part-time lead developers, which produced communication issues. As co-founders, we communicated well and recognized the problems. Stakes, vesting, target markets, and approach were agreed upon.

We were behind schedule. Technical debt and strategic gap grew.

Bi-daily and weekly reviews didn't help. Each time, there were explanations. Inside, I was freaking out.

Self-reflection

I am a fairly easy person to talk to. I always try to stick to agreements; otherwise, my head gets stuffed with unnecessary information, interpretations, and emotions.

Sit down -> talk -> decide -> do -> evaluate the results. Repeat it.

If I don't get detailed comments, I start ruining everyone's mood. If there's a systematic violation of agreements without a good justification, I won't join the project or I'll end the collaboration.

What would I have done otherwise?

This is where it’s scariest to draw conclusions. Probably the most logical thing would have been not to start the project as we started it. But that was already a completely different project. So I would not have done anything differently and would have failed again.

But I drew conclusions for the future.

The lesson

First-time founders should find an adviser or team coach for a strategic session. It helps split the roles and responsibilities.

People are slowly recognizing entrepreneurship's downside.

Growing up, entrepreneurship wasn't common. High school class of 2012 had no entrepreneurs.

Businesses were different.

They had staff and a lengthy history of achievement.

I never wanted a business. It felt unattainable. My friends didn't care.

Weird.

People desired degrees to attain good jobs at big companies.

When graduated high school:

• 9 out of 10 people attend college

• Earn minimum wage (7%) working in a restaurant or retail establishment

• Or join the military (3%)

Later, entrepreneurship became a thing.

2014-ish

I was in the military and most of my high school friends were in college, so I didn't hear anything.

Entrepreneurship soared in 2015, according to Google Trends.

Then more individuals were interested. Entrepreneurship went from unusual to cool.

In 2015, it was easier than ever to build a website, run Facebook advertisements, and achieve organic social media reach.

There were several online business tools.

You didn't need to spend years or money figuring it out. Most entry barriers were gone.

Everyone wanted a side gig to escape the 95.

Small company applications have increased during the previous 10 years.

2011-2014 trend continues.

2015 adds 150,000 applications. 2016 adds 200,000. Plus 300,000 in 2017.

The graph makes it look little, but that's a considerable annual spike with no indications of stopping.

By 2021, new business apps had doubled.

Entrepreneurship will return to its early 2010s level.

I think we'll go backward in 5 years.

Entrepreneurship is half as popular as it was in 2015.

In the late 2020s and 30s, entrepreneurship will again be obscure.

Entrepreneurship's decade-long splendor is fading. People will cease escaping 9-5 and launch fewer companies.

That’s not a bad thing.

I think people have a rose-colored vision of entrepreneurship. It's fashionable. People feel that they're missing out if they're not entrepreneurial.

Reality is showing up.

People say on social media, "I knew starting a business would be hard, but not this hard."

More negative posts on entrepreneurship:

Luke adds:

Is being an entrepreneur ‘healthy’? I don’t really think so. Many like Gary V, are not role models for a well-balanced life. Despite what feel-good LinkedIn tells you the odds are against you as an entrepreneur. You have to work your face off. It’s a tough but rewarding lifestyle. So maybe let’s stop glorifying it because it takes a lot of (bleepin) work to survive a pandemic, mental health battles, and a competitive market.

Entrepreneurship is no longer a pipe dream.

It’s hard.

I went full-time in March 2020. I was done by April 2021. I had a good-paying job with perks.

When that fell through (on my start date), I had to continue my entrepreneurial path. I needed money by May 1 to pay rent.

Entrepreneurship isn't as great as many think.

Entrepreneurship is a serious business.

If you have a 9-5, the grass isn't greener here. Most people aren't telling the whole story when they post on social media or quote successful entrepreneurs.

People prefer to communicate their victories than their defeats.

Is this a bad thing?

I don’t think so.

Over the previous decade, entrepreneurship went from impossible to the finest thing ever.

It peaked in 2020-21 and is returning to reality.

Startups aren't for everyone.

If you like your job, don't quit.

Entrepreneurship won't amaze people if you quit your job.

It's irrelevant.

You're doomed.

And you'll probably make less money.

If you hate your job, quit. Change jobs and bosses. Changing jobs could net you a greater pay or better perks.

When you go solo, your paycheck and perks vanish. Did I mention you'll fail, sleep less, and stress more?

Nobody will stop you from pursuing entrepreneurship. You'll face several challenges.

Possibly.

Entrepreneurship may be romanticized for years.

Based on what I see from entrepreneurs on social media and trends, entrepreneurship is challenging and few will succeed.

Are you an idiot at fundraising?

Humans undervalue what they don't grasp. Consider NASCAR. How is that a sport? ask uneducated observers. Circular traffic. Driving near a car's physical limits is different from daily driving. When driving at 200 mph, seemingly simple things like changing gas weight or asphalt temperature might be life-or-death.

Venture investors do something similar in entrepreneurship. Most entrepreneurs don't realize how complex venture finance is.

In my early startup days, I didn't comprehend venture capital's intricacy. I thought VCs were rich folks looking for the next Mark Zuckerberg. I was meant to be a sleek, enthusiastic young entrepreneur who could razzle-dazzle investors.

Finally, one of the VCs I was trying to woo set me straight. He insulted me.

How I learned that I was approaching the wrong investor

I was constructing a consumer-facing, pre-revenue marketplace firm. I looked for investors in my old university's alumni database. My city had one. After some research, I learned he was a partner at a growth-stage, energy-focused VC company with billions under management.

Billions? I thought. Surely he can write a million-dollar cheque. He'd hardly notice.

I emailed the VC about our shared alumni status, explaining that I was building a startup in the area and wanted advice. When he agreed to meet the next week, I prepared my pitch deck.

First error.

The meeting seemed like a funding request. Imagine the awkwardness.

His assistant walked me to the firm's conference room and told me her boss was running late. While waiting, I prepared my pitch. I connected my computer to the projector, queued up my PowerPoint slides, and waited for the VC.

He didn't say hello or apologize when he entered a few minutes later. What are you doing?

Hi! I said, Confused but confident. Dinin Aaron. My startup's pitch.

Who? Suspicious, he replied. Your email says otherwise. You wanted help.

I said, "Isn't that a euphemism for contacting investors?" Fundraising I figured I should pitch you.

As he sat down, he smiled and said, "Put away your computer." You need to study venture capital.

Recognizing the business aspects of venture capital

The VC taught me venture capital in an hour. Young entrepreneur me needed this lesson. I assume you need it, so I'm sharing it.

Most people view venture money from an entrepreneur's perspective, he said. They envision a world where venture capital serves entrepreneurs and startups.

As my VC indicated, VCs perceive their work differently. Venture investors don't serve entrepreneurs. Instead, they run businesses. Their product doesn't look like most products. Instead, the VCs you're proposing have recognized an undervalued market segment. By investing in undervalued companies, they hope to profit. It's their investment thesis.

Your company doesn't fit my investment thesis, the venture capitalist told me. Your pitch won't beat my investing theory. I invest in multimillion-dollar clean energy companies. Asking me to invest in you is like ordering a breakfast burrito at a fancy steakhouse. They could, but why? They don't do that.

Yeah, I’m not a fine steak yet, I laughed, feeling like a fool for pitching a growth-stage VC used to looking at energy businesses with millions in revenues on my pre-revenue, consumer startup.

He stressed that it's not necessary. There are investors targeting your company. Not me. Find investors and pitch them.

Remember this when fundraising. Your investors aren't philanthropists who want to help entrepreneurs realize their company goals. Venture capital is a sophisticated investment strategy, and VC firm managers are industry experts. They're looking for companies that meet their investment criteria. As a young entrepreneur, I didn't grasp this, which is why I struggled to raise money. In retrospect, I probably seemed like an idiot. Hopefully, you won't after reading this.

Deep learning-based generative art is being researched. As usual, self-learning is better. Some models, like OpenAI's DALL-E 2, require registration and can only be used online, but others can be used locally, which is usually more enjoyable for curious users. I'll demonstrate the Stable Diffusion model's operation on a standard PC.

Let’s get started.

What It Does

Stable Diffusion uses numerous components:

• A generative model trained to produce images is called a diffusion model. The model is incrementally improving the starting data, which is only random noise. The model has an image, and while it is being trained, the reversed process is being used to add noise to the image. Being able to reverse this procedure and create images from noise is where the true magic is (more details and samples can be found in the paper).

• An internal compressed representation of a latent diffusion model, which may be altered to produce the desired images, is used (more details can be found in the paper). The capacity to fine-tune the generation process is essential because producing pictures at random is not very attractive (as we can see, for instance, in Generative Adversarial Networks).

• A neural network model called CLIP (Contrastive Language-Image Pre-training) is used to translate natural language prompts into vector representations. This model, which was trained on 400,000,000 image-text pairs, enables the transformation of a text prompt into a latent space for the diffusion model in the scenario of stable diffusion (more details in that paper).

This figure shows all data flow:

The weights file size for Stable Diffusion model v1 is 4 GB and v2 is 5 GB, making the model quite huge. The v1 model was trained on 256x256 and 512x512 LAION-5B pictures on a 4,000 GPU cluster using over 150.000 NVIDIA A100 GPU hours. The open-source pre-trained model is helpful for us. And we will.

Install

Before utilizing the Python sources for Stable Diffusion v1 on GitHub, we must install Miniconda (assuming Git and Python are already installed):

wget https://repo.anaconda.com/miniconda/Miniconda3-py39_4.12.0-Linux-x86_64.sh
chmod +x Miniconda3-py39_4.12.0-Linux-x86_64.sh
./Miniconda3-py39_4.12.0-Linux-x86_64.sh
conda update -n base -c defaults conda

Install the source and prepare the environment:

git clone https://github.com/CompVis/stable-diffusion
cd stable-diffusion
conda env create -f environment.yaml
conda activate ldm
pip3 install transformers --upgrade

Download the pre-trained model weights next. HiggingFace has the newest checkpoint sd-v14.ckpt (a download is free but registration is required). Put the file in the project folder and have fun:

python3 scripts/txt2img.py --prompt "hello world" --plms --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

Almost. The installation is complete for happy users of current GPUs with 12 GB or more VRAM. RuntimeError: CUDA out of memory will occur otherwise. Two solutions exist.

Running the optimized version

Try optimizing first. After cloning the repository and enabling the environment (as previously), we can run the command:

python3 optimizedSD/optimized_txt2img.py --prompt "hello world" --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

Stable Diffusion worked on my visual card with 8 GB RAM (alas, I did not behave well enough to get NVIDIA A100 for Christmas, so 8 GB GPU is the maximum I have;).

Running Stable Diffusion without GPU

If the GPU does not have enough RAM or is not CUDA-compatible, running the code on a CPU will be 20x slower but better than nothing. This unauthorized CPU-only branch from GitHub is easiest to obtain. We may easily edit the source code to use the latest version. It's strange that a pull request for that was made six months ago and still hasn't been approved, as the changes are simple. Readers can finish in 5 minutes:

• Replace if attr.device!= torch.device(cuda) with if attr.device!= torch.device(cuda) and torch.cuda.is available at line 20 of ldm/models/diffusion/ddim.py ().

• Replace if attr.device!= torch.device(cuda) with if attr.device!= torch.device(cuda) and torch.cuda.is available in line 20 of ldm/models/diffusion/plms.py ().

• Replace device=cuda in lines 38, 55, 83, and 142 of ldm/modules/encoders/modules.py with device=cuda if torch.cuda.is available(), otherwise cpu.

• Replace model.cuda() in scripts/txt2img.py line 28 and scripts/img2img.py line 43 with if torch.cuda.is available(): model.cuda ().

Run the script again.

Testing

Test the model. Text-to-image is the first choice. Test the command line example again:

python3 scripts/txt2img.py --prompt "hello world" --plms --ckpt sd-v1-4.ckpt --skip_grid --n_samples 1

The slow generation takes 10 seconds on a GPU and 10 minutes on a CPU. Final image:

Hello world is dull and abstract. Try a brush-wielding hamster. Why? Because we can, and it's not as insane as Napoleon's cat. Another image:

Generating an image from a text prompt and another image is interesting. I made this picture in two minutes using the image editor (sorry, drawing wasn't my strong suit):

I can create an image from this drawing:

python3 scripts/img2img.py --prompt "A bird is sitting on a tree branch" --ckpt sd-v1-4.ckpt --init-img bird.png --strength 0.8

It was far better than my initial drawing:

I hope readers understand and experiment.

Stable Diffusion UI

Developers love the command line, but regular users may struggle. Stable Diffusion UI projects simplify image generation and installation. Simple usage:

• Unpack the ZIP after downloading it from https://github.com/cmdr2/stable-diffusion-ui/releases. Linux and Windows are compatible with Stable Diffusion UI (sorry for Mac users, but those machines are not well-suitable for heavy machine learning tasks anyway;).

• Start the script.

Done. The web browser UI makes configuring various Stable Diffusion features (upscaling, filtering, etc.) easy:

V2.1 of Stable Diffusion

I noticed the notification about releasing version 2.1 while writing this essay, and it was intriguing to test it. First, compare version 2 to version 1:

• alternative text encoding. The Contrastive LanguageImage Pre-training (CLIP) deep learning model, which was trained on a significant number of text-image pairs, is used in Stable Diffusion 1. The open-source CLIP implementation used in Stable Diffusion 2 is called OpenCLIP. It is difficult to determine whether there have been any technical advancements or if legal concerns were the main focus. However, because the training datasets for the two text encoders were different, the output results from V1 and V2 will differ for the identical text prompts.

• a new depth model that may be used to the output of image-to-image generation.

• a revolutionary upscaling technique that can quadruple the resolution of an image.

• Generally higher resolution Stable Diffusion 2 has the ability to produce both 512x512 and 768x768 pictures.

The Hugging Face website offers a free online demo of Stable Diffusion 2.1 for code testing. The process is the same as for version 1.4. Download a fresh version and activate the environment:

conda deactivate  
conda env remove -n ldm  # Use this if version 1 was previously installed
git clone https://github.com/Stability-AI/stablediffusion
cd stablediffusion
conda env create -f environment.yaml
conda activate ldm

Hugging Face offers a new weights ckpt file.

The Out of memory error prevented me from running this version on my 8 GB GPU. Version 2.1 fails on CPUs with the slow conv2d cpu not implemented for Half error (according to this GitHub issue, the CPU support for this algorithm and data type will not be added). The model can be modified from half to full precision (float16 instead of float32), however it doesn't make sense since v1 runs up to 10 minutes on the CPU and v2.1 should be much slower. The online demo results are visible. The same hamster painting with a brush prompt yielded this result:

It looks different from v1, but it functions and has a higher resolution.

The superresolution.py script can run the 4x Stable Diffusion upscaler locally (the x4-upscaler-ema.ckpt weights file should be in the same folder):

python3 scripts/gradio/superresolution.py configs/stable-diffusion/x4-upscaling.yaml x4-upscaler-ema.ckpt

This code allows the web browser UI to select the image to upscale:

The copy-paste strategy may explain why the upscaler needs a text prompt (and the Hugging Face code snippet does not have any text input as well). I got a GPU out of memory error again, although CUDA can be disabled like v1. However, processing an image for more than two hours is unlikely:

Stable Diffusion Limitations

When we use the model, it's fun to see what it can and can't do. Generative models produce abstract visuals but not photorealistic ones. This fundamentally limits The generative neural network was trained on text and image pairs, but humans have a lot of background knowledge about the world. The neural network model knows nothing. If someone asks me to draw a Chinese text, I can draw something that looks like Chinese but is actually gibberish because I never learnt it. Generative AI does too! Humans can learn new languages, but the Stable Diffusion AI model includes only language and image decoder brain components. For instance, the Stable Diffusion model will pull NO WAR banner-bearers like this:

V1:

V2.1:

The shot shows text, although the model never learned to read or write. The model's string tokenizer automatically converts letters to lowercase before generating the image, so typing NO WAR banner or no war banner is the same.

I can also ask the model to draw a gorgeous woman:

V1:

V2.1:

The first image is gorgeous but physically incorrect. A second one is better, although it has an Uncanny valley feel. BTW, v2 has a lifehack to add a negative prompt and define what we don't want on the image. Readers might try adding horrible anatomy to the gorgeous woman request.

If we ask for a cartoon attractive woman, the results are nice, but accuracy doesn't matter:

V1:

V2.1:

Another example: I ordered a model to sketch a mouse, which looks beautiful but has too many legs, ears, and fingers:

V1:

V2.1: improved but not perfect.

V1 produces a fun cartoon flying mouse if I want something more abstract:

I tried multiple times with V2.1 but only received this:

The image is OK, but the first version is closer to the request.

Stable Diffusion struggles to draw letters, fingers, etc. However, abstract images yield interesting outcomes. A rural landscape with a modern metropolis in the background turned out well:

V1:

V2.1:

Generative models help make paintings too (at least, abstract ones). I searched Google Image Search for modern art painting to see works by real artists, and this was the first image:

I typed "abstract oil painting of people dancing" and got this:

V1:

V2.1:

It's a different style, but I don't think the AI-generated graphics are worse than the human-drawn ones.

The AI model cannot think like humans. It thinks nothing. A stable diffusion model is a billion-parameter matrix trained on millions of text-image pairs. I input "robot is creating a picture with a pen" to create an image for this post. Humans understand requests immediately. I tried Stable Diffusion multiple times and got this:

This great artwork has a pen, robot, and sketch, however it was not asked. Maybe it was because the tokenizer deleted is and a words from a statement, but I tried other requests such robot painting picture with pen without success. It's harder to prompt a model than a person.

I hope Stable Diffusion's general effects are evident. Despite its limitations, it can produce beautiful photographs in some settings. Readers who want to use Stable Diffusion results should be warned. Source code examination demonstrates that Stable Diffusion images feature a concealed watermark (text StableDiffusionV1 and SDV2) encoded using the invisible-watermark Python package. It's not a secret, because the official Stable Diffusion repository's test watermark.py file contains a decoding snippet. The put watermark line in the txt2img.py source code can be removed if desired. I didn't discover this watermark on photographs made by the online Hugging Face demo. Maybe I did something incorrectly (but maybe they are just not using the txt2img script on their backend at all).

Conclusion

The Stable Diffusion model was fascinating. As I mentioned before, trying something yourself is always better than taking someone else's word, so I encourage readers to do the same (including this article as well;).

Is Generative AI a game-changer? My humble experience tells me:

• I think that place has a lot of potential. For designers and artists, generative AI can be a truly useful and innovative tool. Unfortunately, it can also pose a threat to some of them since if users can enter a text field to obtain a picture or a website logo in a matter of clicks, why would they pay more to a different party? Is it possible right now? unquestionably not yet. Images still have a very poor quality and are erroneous in minute details. And after viewing the image of the stunning woman above, models and fashion photographers may also unwind because it is highly unlikely that AI will replace them in the upcoming years.

• Today, generative AI is still in its infancy. Even 768x768 images are considered to be of a high resolution when using neural networks, which are computationally highly expensive. There isn't an AI model that can generate high-resolution photographs natively without upscaling or other methods, at least not as of the time this article was written, but it will happen eventually.

• It is still a challenge to accurately represent knowledge in neural networks (information like how many legs a cat has or the year Napoleon was born). Consequently, AI models struggle to create photorealistic photos, at least where little details are important (on the other side, when I searched Google for modern art paintings, the results are often even worse;).

• When compared to the carefully chosen images from official web pages or YouTube reviews, the average output quality of a Stable Diffusion generation process is actually less attractive because to its high degree of randomness. When using the same technique on their own, consumers will theoretically only view those images as 1% of the results.

Anyway, it's exciting to witness this area's advancement, especially because the project is open source. Google's Imagen and DALL-E 2 can also produce remarkable findings. It will be interesting to see how they progress.