Hiring is difficult, but you shouldn't compromise on team members. Or it may suggest you need to look beyond years in a similar role/function.

Every hire should be someone we'd want as one of our first ten employees.

If you hire such people, your team will adapt, initiate, and problem-solve, and your company will grow. You'll stay nimble even as you scale, and you'll learn from your colleagues.

If you only hire for a specific role or someone who can execute the job, you'll become a cluster of optimizers, and talent will depart for a more fascinating company. A startup is continually changing, therefore you want individuals that embrace it.

As a leader, establishing ideal conditions for talent and having a real ideology should be high on your agenda. You can't eliminate attrition, nor would you want to, but you can hire people who will become your company's leaders.

In my last four jobs I was employee 2, 5, 3, and 5. So while this is all a bit self serving, you’re the one reading my writing — and I have some experience with who works out in the first ten!

First, we'll examine what they do well (and why they're beneficial for startups), then what they don't, and how to hire them.

First 10 are:

• Business partners: Because it's their company, they take care of whatever has to be done and have ideas about how to do it. You can rely on them to always put the success of the firm first because it is their top priority (company success is strongly connected with success for early workers). This approach will eventually take someone to leadership positions.
  

• High Speed Learners: They process knowledge quickly and can reach 80%+ competency in a new subject matter rather quickly. A growing business that is successful tries new things frequently. We have all lost a lot of money and time on employees who follow the wrong playbook or who wait for someone else within the company to take care of them.
  

• Autodidacts learn by trial and error, osmosis, networking with others, applying first principles, and reading voraciously (articles, newsletters, books, and even social media). Although teaching is wonderful, you won't have time.
  

• Self-scaling: They figure out a means to deal with issues and avoid doing the grunt labor over the long haul, increasing their leverage. Great people don't keep doing the same thing forever; as they expand, they use automation and delegation to fill in their lower branches. This is a crucial one; even though you'll still adore them, you'll have to manage their scope or help them learn how to scale on their own.
  

• Free Range: You can direct them toward objectives rather than specific chores. Check-ins can be used to keep them generally on course without stifling invention instead of giving them precise instructions because doing so will obscure their light.
  

• When people are inspired, they bring their own ideas about what a firm can be and become animated during discussions about how to get there.
  

• Novelty Seeking: They look for business and personal growth chances. Give them fresh assignments and new directions to follow around once every three months.

Here’s what the First Ten types may not be:

• Domain specialists. When you look at their resumes, you'll almost certainly think they're unqualified. Fortunately, a few strategically positioned experts may empower a number of First Ten types by serving on a leadership team or in advising capacities.
  

• Balanced. These people become very invested, and they may be vulnerable to many types of stress. You may need to assist them in managing their own stress and coaching them through obstacles. If you are reading this and work at Banza, I apologize for not doing a better job of supporting this. I need to be better at it.
  

• Able to handle micromanagement with ease. People who like to be in charge will suppress these people. Good decision-making should be delegated to competent individuals. Generally speaking, if you wish to scale.

Great startup team members have versatility, learning, innovation, and energy. When we hire for the function, not the person, we become dull and staid. Could this person go to another department if needed? Could they expand two levels in a few years?

First Ten qualities and experience level may have a weak inverse association. People with 20+ years of experience who had worked at larger organizations wanted to try something new and had a growth mentality. College graduates may want to be told what to do and how to accomplish it so they can stay in their lane and do what their management asks.

Does the First Ten archetype sound right for your org? Cool, let’s go hiring. How will you know when you’ve found one?

• They exhibit adaptive excellence, excelling at a variety of unrelated tasks. It could be hobbies or professional talents. This suggests that they will succeed in the next several endeavors they pursue.
  

• Successful risk-taking is doing something that wasn't certain to succeed, sometimes more than once, and making it do so. It's an attitude.
  

• Rapid Rise: They regularly change roles and get promoted. However, they don't leave companies when the going gets tough. Look for promotions at every stop and at least one position with three or more years of experience.

You can ask them:

• Tell me about a time when you started from scratch or achieved success. What occurred en route? You might request a variety of tales from various occupations or even aspects of life. They ought to be energized by this.
  

• What new skills have you just acquired? It is not required to be work-related. They must be able to describe it and unintentionally become enthusiastic about it.
  

• Tell me about a moment when you encountered a challenge and had to alter your strategy. The core of a startup is reinventing itself when faced with obstacles.
  

• Tell me about a moment when you eliminated yourself from a position at work. They've demonstrated they can permanently solve one issue and develop into a new one, as stated above.
  

• Why do you want to leave X position or Y duty? These people ought to be moving forward, not backward, all the time. Instead, they will discuss what they are looking forward to visiting your location.
  

• Any questions? Due to their inherent curiosity and desire to learn new things, they should practically never run out of questions. You can really tell if they are sufficiently curious at this point.

People who see their success as being the same as the success of the organization are the best-case team members, in any market. They’ll grow and change with the company, and always try to prioritize what matters. You’ll find yourself more energized by your work because you’re surrounded by others who are as well. Happy teambuilding!

Offices without books are like bodies without souls.

After 10 years, my CEO sold his company for $30 million. I've shared many of his lessons on medium. You could ask him anything at his always-open office. He also said we could use his office for meetings while he was away. When I used his office for work, I was always struck by how many books he had.

Books are useful in almost every aspect of learning. Building a business, improving family relationships, learning a new language, a new skill... Books teach, guide, and structure. Whether fiction or nonfiction, books inspire, give ideas, and develop critical thinking skills.

My CEO prefers non-fiction and attends a Friday book club. This article discusses 5 books I found in his office that impacted my life/business. My CEO sold his company for $30 million, but I've built a steady business through blogging and video making.

I recall events and lessons I learned from my CEO and how they relate to each book, and I explain how I applied the book's lessons to my business and life.

Note: This post has no affiliate links.

1. The One Thing — Gary Keller

Gary Keller, a real estate agent, wanted more customers. So he and his team brainstormed ways to get more customers. They decided to write a bestseller about work and productivity. The more people who saw the book, the more customers they'd get.

Gary Keller focused on writing the best book on productivity, work, and efficiency for months. His business experience. Keller's business grew after the book's release.

The author summarizes the book in one question.

"What's the one thing that will make everything else easier or unnecessary?"

When I started my blog and business alongside my 9–5, I quickly identified my one thing: writing. My business relied on it, so it had to be great. Without writing, there was no content, traffic, or business.

My CEO focused on funding when he started his business. Even in his final years, he spent a lot of time on the phone with investors, either to get more money or to explain what he was doing with it. My CEO's top concern was money, and the other super important factors were handled by separate teams.

• Product tech and design

• Incredible customer support team

• Excellent promotion team

• Profitable sales team

My CEO didn't always focus on one thing and ignore the rest. He was on all of those teams when I started my job. He'd start his day in tech, have lunch with marketing, and then work in sales. He was in his office on the phone at night.

He eventually realized his errors. Investors told him he couldn't do everything for the company. If needed, he had to change internally. He learned to let go, mind his own business, and focus for the next four years. Then he sold for $30 million.

The bigger your project/company/idea, the more you'll need to delegate to stay laser-focused. I started something new every few months for 10 years before realizing this. So much to do makes it easy to avoid progress. Once you identify the most important aspect of your project and enlist others' help, you'll be successful.

2. Eat That Frog — Brian Tracy

The author quote sums up book's essence:

Mark Twain said that if you eat a live frog in the morning, it's probably the worst thing that will happen to you all day. Your "frog" is the biggest, most important task you're most likely to procrastinate on.

"Frog" and "One Thing" are both about focusing on what's most important. Eat That Frog recommends doing the most important task first thing in the morning.

I shared my CEO's calendar in an article 10 months ago. Like this:

CEO's average week (some information crossed out for confidentiality)

Notice anything about 8am-8:45am? Almost every day is the same (except Friday). My CEO started his day with a management check-in for 2 reasons:

• Checking in with all managers is cognitively demanding, and my CEO is a morning person.

• In a young startup where everyone is busy, the morning management check-in was crucial. After 10 am, you couldn't gather all managers.

When I started my blog, writing was my passion. I'm a morning person, so I woke up at 6 am and started writing by 6:30 am every day for a year. This allowed me to publish 3 articles a week for 52 weeks to build my blog and audience. After 2 years, I'm not stopping.

3. Deep Work — Cal Newport

Deep work is focusing on a cognitively demanding task without distractions (like a morning management meeting). It helps you master complex information quickly and produce better results faster. In a competitive world 10 or 20 years ago, focus wasn't a huge advantage. Smartphones, emails, and social media made focus a rare, valuable skill.

Most people can't focus anymore. Screens light up, notifications buzz, emails arrive, Instagram feeds... Many people don't realize they're interrupted because it's become part of their normal workflow.

Cal Newport mentions Bill Gates' "Think Weeks" in Deep Work.

Microsoft CEO Bill Gates would isolate himself (often in a lakeside cottage) twice a year to read and think big thoughts.

Inside Bill's Brain on Netflix shows Newport's lakeside cottage. I've always wanted a lakeside cabin to work in. My CEO bought a lakehouse after selling his company, but now he's retired.

As a company grows, you can focus less on it. In a previous section, I said investors told my CEO to get back to basics and stop micromanaging. My CEO's commitment and ability to get work done helped save the company. His deep work and new frameworks helped us survive the corona crisis (more on this later).

The ability to deep work will be a huge competitive advantage in the next century. Those who learn to work deeply will likely be successful while everyone else is glued to their screens, Bluetooth-synced to their watches, and playing Candy Crush on their tablets.

4. The 7 Habits of Highly Effective People — Stephen R. Covey

It took me a while to start reading this book because it seemed like another shallow self-help bible. I kept finding this book when researching self-improvement. I tried it because it was everywhere.

Stephen Covey taught me 2 years ago to have a personal mission statement.

A 7 Habits mission statement describes the life you want to lead, the character traits you want to embody, and the impact you want to have on others. shortform.com

I've had many lunches with my CEO and talked about Vipassana meditation and Sunday forest runs, but I've never seen his mission statement. I'm sure his family is important, though. In the above calendar screenshot, you can see he always included family events (in green) so we could all see those time slots. We couldn't book him then. Although he never spent as much time with his family as he wanted, he always made sure to be on time for his kid's birthday rather than a conference call.

My CEO emphasized his company's mission. Your mission statement should answer 3 questions.

• What does your company do?

• How does it do it?

• Why does your company do it?

As a graphic designer, I had to create mission-statement posters. My CEO hung posters in each office.

5. Measure What Matters — John Doerr

This book is about Andrew Grove's OKR strategy, developed in 1968. When he joined Google's early investors board, he introduced it to Larry Page and Sergey Brin. Google still uses OKR.

Objective Key Results

• Objective: It explains your goals and desired outcome. When one goal is reached, another replaces it. OKR objectives aren't technical, measured, or numerical. They must be clear.

• Key Result should be precise, technical, and measurable, unlike the Objective. It shows if the Goal is being worked on. Time-bound results are quarterly or yearly.

Our company almost sank several times. Sales goals were missed, management failed, and bad decisions were made. On a Monday, our CEO announced we'd implement OKR to revamp our processes.

This was a year before the pandemic, and I'm certain we wouldn't have sold millions or survived without this change. This book impacted the company the most, not just management but all levels. Organization and transparency improved. We reached realistic goals. Happy investors. We used the online tool Gtmhub to implement OKR across the organization.

My CEO's company went from near bankruptcy to being acquired for $30 million in 2 years after implementing OKR.

I hope you enjoyed this booklist. Here's a recap of the 5 books and the lessons I learned from each.

1. The 7 Habits of Highly Effective People — Stephen R. Covey

Have a mission statement that outlines your goals, character traits, and impact on others.

1. Deep Work — Cal Newport

Focus is a rare skill; master it. Deep workers will succeed in our hyper-connected, distracted world.

1. The One Thing — Gary Keller

What can you do that will make everything else easier or unnecessary? Once you've identified it, focus on it.

1. Eat That Frog — Brian Tracy

Identify your most important task the night before and do it first thing in the morning. You'll have a lighter day.

1. Measure What Matters — John Doerr

On a timeline, divide each long-term goal into chunks. Divide those slices into daily tasks (your goals). Time-bound results are quarterly or yearly. Objectives aren't measured or numbered.

Thanks for reading. Enjoy the ride!

Only two of the following three options can be achieved: consistency, availability, and partition tolerance

Someone told me that growing from 30 to 60 is the biggest adjustment for a team or business.

I remember thinking, That's random. Each company is unique. I've seen teams of all types confront the same issues during development periods. With new enterprises starting every year, we should be better at navigating growing difficulties.

As a team grows, its processes and systems break down, requiring reorganization or declining results. Why always? Why isn't there a perfect scaling model? Why hasn't that been found?

The Three Things Productive Organizations Must Have

Any company should be efficient and productive. Three items are needed:

First, it must verify that no two team members have conflicting information about the roadmap, strategy, or any input that could affect execution. Teamwork is required.

Second, it must ensure that everyone can receive the information they need from everyone else quickly, especially as teams become more specialized (an inevitability in a developing organization). It requires everyone's accessibility.

Third, it must ensure that the organization can operate efficiently even if a piece is unavailable. It's partition-tolerant.

From my experience with the many teams I've been on, invested in, or advised, achieving all three is nearly impossible. Why a perfect organization model cannot exist is clear after analysis.

The CAP Theorem: What is it?

Eric Brewer of Berkeley discovered the CAP Theorem, which argues that a distributed data storage should have three benefits. One can only have two at once.

The three benefits are consistency, availability, and partition tolerance, which implies that even if part of the system is offline, the remainder continues to work.

This notion is usually applied to computer science, but I've realized it's also true for human organizations. In a post-COVID world, many organizations are hiring non-co-located staff as they grow. CAP Theorem is more important than ever. Growing teams sometimes think they can develop ways to bypass this law, dooming themselves to a less-than-optimal team dynamic. They should adopt CAP to maximize productivity.

Path 1: Consistency and availability equal no tolerance for partitions

Let's imagine you want your team to always be in sync (i.e., for someone to be the source of truth for the latest information) and to be able to share information with each other. Only division into domains will do.

Numerous developing organizations do this, especially after the early stage (say, 30 people) when everyone may wear many hats and be aware of all the moving elements. After a certain point, it's tougher to keep generalists aligned than to divide them into specialized tasks.

In a specialized, segmented team, leaders optimize consistency and availability (i.e. every function is up-to-speed on the latest strategy, no one is out of sync, and everyone is able to unblock and inform everyone else).

Partition tolerance suffers. If any component of the organization breaks down (someone goes on vacation, quits, underperforms, or Gmail or Slack goes down), productivity stops. There's no way to give the team stability, availability, and smooth operation during a hiccup.

Path 2: Partition Tolerance and Availability = No Consistency

Some businesses avoid relying too heavily on any one person or sub-team by maximizing availability and partition tolerance (the organization continues to function as a whole even if particular components fail). Only redundancy can do that. Instead of specializing each member, the team spreads expertise so people can work in parallel. I switched from Path 1 to Path 2 because I realized too much reliance on one person is risky.

What happens after redundancy? Unreliable. The more people may run independently and in parallel, the less anyone can be the truth. Lack of alignment or updated information can lead to people executing slightly different strategies. So, resources are squandered on the wrong work.

Path 3: Partition and Consistency "Tolerance" equates to "absence"

The third, least-used path stresses partition tolerance and consistency (meaning answers are always correct and up-to-date). In this organizational style, it's most critical to maintain the system operating and keep everyone aligned. No one is allowed to read anything without an assurance that it's up-to-date (i.e. there’s no availability).

Always short-lived. In my experience, a business that prioritizes quality and scalability over speedy information transmission can get bogged down in heavy processes that hinder production. Large-scale, this is unsustainable.

Accepting CAP

When two puzzle pieces fit, the third won't. I've watched developing teams try to tackle these difficulties, only to find, as their ancestors did, that they can never be entirely solved. Idealized solutions fail in reality, causing lost effort, confusion, and lower production.

As teams develop and change, they should embrace CAP, acknowledge there is a limit to productivity in a scaling business, and choose the best two-out-of-three path.

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.

Sundar Pichai unveiled Simplicity Sprint at Google's latest all-hands conference.

To boost employee efficiency.

Not surprising. Few envisioned Google declaring a productivity drive.

Sunder Pichai's speech:

“There are real concerns that our productivity as a whole is not where it needs to be for the head count we have. Help me create a culture that is more mission-focused, more focused on our products, more customer focused. We should think about how we can minimize distractions and really raise the bar on both product excellence and productivity.”

The primary driver driving Google's efficiency push is:

Google's efficiency push follows 13% quarterly revenue increase. Last year in the same quarter, it was 62%.

Market newcomers may argue that the previous year's figure was fuelled by post-Covid reopening and growing consumer spending. Investors aren't convinced. A promising company like Google can't afford to drop so quickly.

Google’s quarterly revenue growth stood at 13%, against 62% in last year same quarter.

Google isn't alone. In my recent essay regarding 2025 programmers, I warned about the economic downturn's effects on FAAMG's workforce. Facebook had suspended hiring, and Microsoft had promised hefty bonuses for loyal staff.

In the same article, I predicted Google's troubles. Online advertising, especially the way Google and Facebook sell it using user data, is over.

FAAMG and 2nd rung IT companies could be the first to fall without Post-COVID revival and uncertain global geopolitics.

Google has hardly ever discussed effectiveness:

Apparently openly.

Amazon treats its employees like robots, even in software positions. It has significant turnover and a terrible reputation as a result. Because of this, it rarely loses money due to staff productivity.

Amazon trumps Google. In reality, it treats its employees poorly.

Google was the founding father of the modern-day open culture.

Larry and Sergey Google founded the IT industry's Open Culture. Silicon Valley called Google's internal democracy and transparency near anarchy. Management rarely slammed decisions on employees. Surveys and internal polls ensured everyone knew the company's direction and had a vote.

20% project allotment (weekly free time to build own project) was Google's open-secret innovation component.

After Larry and Sergey's exit in 2019, this is Google's first profitability hurdle. Only Google insiders can answer these questions.

• Would Google's investors compel the company's management to adopt an Amazon-style culture where the developers are treated like circus performers?

• If so, would Google follow suit?

• If so, how does Google go about doing it?

Before discussing Google's likely plan, let's examine programming productivity.

What determines a programmer's productivity is simple:

How would we answer Google's questions?

As a programmer, I'm more concerned about Simplicity Sprint's aftermath than its economic catalysts.

Large organizations don't care much about quarterly and annual productivity metrics. They have 10-year product-launch plans. If something seems horrible today, it's likely due to someone's lousy judgment 5 years ago who is no longer in the blame game.

Deconstruct our main question.

• How exactly do you change the culture of the firm so that productivity increases?

• How can you accomplish that without affecting your capacity to profit? There are countless ways to increase output without decreasing profit.

• How can you accomplish this with little to no effect on employee motivation? (While not all employers care about it, in this case we are discussing the father of the open company culture.)

• How do you do it for a 10-developer IT firm that is losing money versus a 1,70,000-developer organization with a trillion-dollar valuation?

When implementing a large-scale organizational change, success must be carefully measured.

The fastest way to do something is to do it right, no matter how long it takes.

You require clearly-defined group/team/role segregation and solid pass/fail matrices to:

• You can give performers rewards.

• Ones that are average can be inspired to improve

• Underachievers may receive assistance or, in the worst-case scenario, rehabilitation

As a 20-year programmer, I associate productivity with greatness.

Doing something well, no matter how long it takes, is the fastest way to do it.

Let's discuss a programmer's productivity.

Why productivity is a strange term in programming:

Productivity is work per unit of time.

Money=time This is an economic proverb. More hours worked, more pay. Longer projects cost more.

As a buyer, you desire a quick supply. As a business owner, you want employees who perform at full capacity, creating more products to transport and boosting your profits.

All economic matrices encourage production because of our obsession with it. Productivity is the only organic way a nation may increase its GDP.

Time is money — is not just a proverb, but an economical fact.

Applying the same productivity theory to programming gets problematic. An automating computer. Its capacity depends on the software its master writes.

Today, a sophisticated program can process a billion records in a few hours. Creating one takes a competent coder and the necessary infrastructure. Learning, designing, coding, testing, and iterations take time.

Programming productivity isn't linear, unlike manufacturing and maintenance.

Average programmers produce code every day yet miss deadlines. Expert programmers go days without coding. End of sprint, they often surprise themselves by delivering fully working solutions.

Reversing the programming duties has no effect. Experts aren't needed for productivity.

These patterns remind me of an XKCD comic.

Programming productivity depends on two factors:

• The capacity of the programmer and his or her command of the principles of computer science

• His or her productive bursts, how often they occur, and how long they last as they engineer the answer

At some point, productivity measurement becomes Schrödinger’s cat.

Product companies measure productivity using use cases, classes, functions, or LOCs (lines of code). In days of data-rich source control systems, programmers' merge requests and/or commits are the most preferred yardstick. Companies assess productivity by tickets closed.

Every organization eventually has trouble measuring productivity. Finer measurements create more chaos. Every measure compares apples to oranges (or worse, apples with aircraft.) On top of the measuring overhead, the endeavor causes tremendous and unnecessary stress on teams, lowering their productivity and defeating its purpose.

Macro productivity measurements make sense. Amazon's factory-era management has done it, but at great cost.

Google can pull it off if it wants to.

What Google meant in reality when it said that employee productivity has decreased:

When Google considers its employees unproductive, it doesn't mean they don't complete enough work in the allotted period.

They can't multiply their work's influence over time.

• Programmers who produce excellent modules or products are unsure on how to use them.

• The best data scientists are unable to add the proper parameters in their models.

• Despite having a great product backlog, managers struggle to recruit resources with the necessary skills.

• Product designers who frequently develop and A/B test newer designs are unaware of why measures are inaccurate or whether they have already reached the saturation point.

• Most ignorant: All of the aforementioned positions are aware of what to do with their deliverables, but neither their supervisors nor Google itself have given them sufficient authority.

So, Google employees aren't productive.

How to fix it?

• Business analysis: White suits introducing novel items can interact with customers from all regions. Track analytics events proactively, especially the infrequent ones.

• SOLID, DRY, TEST, and AUTOMATION: Do less + reuse. Use boilerplate code creation. If something already exists, don't implement it yourself.

• Build features-building capabilities: N features are created by average programmers in N hours. An endless number of features can be built by average programmers thanks to the fact that expert programmers can produce 1 capability in N hours.

• Work on projects that will have a positive impact: Use the same algorithm to search for images on YouTube rather than the Mars surface.

• Avoid tasks that can only be measured in terms of time linearity at all costs (if a task can be completed in N minutes, then M copies of the same task would cost M*N minutes).

In conclusion:

Software development isn't linear. Why should the makers be measured?

Notation for The Big O

I'm discussing a new way to quantify programmer productivity. (It applies to other professions, but that's another subject)

The Big O notation expresses the paradigm (the algorithmic performance concept programmers rot to ace their Google interview)

Google (or any large corporation) can do this.

1. Sort organizational roles into categories and specify their impact vs. time objectives. A CXO role's time vs. effect function, for instance, has a complexity of O(log N), meaning that if a CEO raises his or her work time by 8x, the result only increases by 3x.

2. Plot the influence of each employee over time using the X and Y axes, respectively.

3. Add a multiplier for Y-axis values to the productivity equation to make business objectives matter. (Example values: Support = 5, Utility = 7, and Innovation = 10).

4. Compare employee scores in comparable categories (developers vs. devs, CXOs vs. CXOs, etc.) and reward or help employees based on whether they are ahead of or behind the pack.

After measuring every employee's inventiveness, it's straightforward to help underachievers and praise achievers.

Example of a Big(O) Category:

If I ran Google (God forbid, its worst days are far off), here's how I'd classify it. You can categorize Google employees whichever you choose.

The Google interview truth:

O(1) < O(log n) < O(n) < O(n log n) < O(n^x) where all logarithmic bases are < n.

O(1): Customer service workers' hours have no impact on firm profitability or customer pleasure.

CXOs Most of their time is spent on travel, strategic meetings, parties, and/or meetings with minimal floor-level influence. They're good at launching new products but bad at pivoting without disaster. Their directions are being followed.

Devops, UX designers, testers Agile projects revolve around deployment. DevOps controls the levers. Their automation secures results in subsequent cycles.

UX/UI Designers must still prototype UI elements despite improved design tools.

All test cases are proportional to use cases/functional units, hence testers' work is O(N).

Architects Their effort improves code quality. Their right/wrong interference affects product quality and rollout decisions even after the design is set.

Core Developers Only core developers can write code and own requirements. When people understand and own their labor, the output improves dramatically. A single character error can spread undetected throughout the SDLC and cost millions.

Core devs introduce/eliminate 1000x bugs, refactoring attempts, and regression. Following our earlier hypothesis.

The fastest way to do something is to do it right, no matter how long it takes.

Conclusion:

Google is at the liberal extreme of the employee-handling spectrum

Microsoft faced an existential crisis after 2000. It didn't choose Amazon's data-driven people management to revitalize itself.

Instead, it entrusted developers. It welcomed emerging technologies and opened up to open source, something it previously opposed.

Google is too lax in its employee-handling practices. With that foundation, it can only follow Amazon, no matter how carefully.

Any attempt to redefine people's measurements will affect the organization emotionally.

The more Google compares apples to apples, the higher its chances for future rebirth.