DEV Community: Ibrahim Hudson

[Boost]

Ibrahim Hudson — Mon, 08 Jun 2026 04:08:34 +0000

Ibrahim Hudson

Jun 4

How to Utilize AI (Not Just Use It): The AI Utilization Guide

#discuss #ai #productivity #tutorial

23 min read

How to Utilize AI (Not Just Use It): The AI Utilization Guide

Ibrahim Hudson — Thu, 04 Jun 2026 17:25:14 +0000

OPTIMIZED FOR NOTION: How to Utilize AI

Fundamentals of Utilizing AI

When using AI, people often run into two distinct experiences, information overload or putting in a question and getting an answer back. Leaving the learning experience to an afterthought. In today’s society, since AI is being pushed more and more in classrooms and workplaces, learning to use AI is becoming paramount to one’s success.

However, what separates a user of AI from someone who utilizes AI in their workflows?

How they interact with AI. In order to utilize AI you need to understand how it works on a fundamental level. So, here’s the abridged version.

How AI Works on A Fundamental Level

VECTOR DATABASE

Vectoring

AI uses an extraordinary amount of data! Billions even trillions of tokens or data points being used and analyzed over fractions of a second. The way this data interacts with itself uses a concept called vectoring. As pictured above the difference between a traditional data table you would’ve seen in high school is that every piece of data in a vector space has a geometric “distance” from another piece of data across hundreds or thousands of invisible dimensions.

AI is then able to use these coordinates to perform complex calculations and formulate responses to user queries at incredible speeds. The overall process in practice is similar to searching for something through a search engine.

Here’s an example:

“The boy asked his mother for a few dollars to go to the ___”

In your head most likely you finished that sentence with “store”, though why is that? As a human you used your past experiences, context clues, and intuition to come to that conclusion. Though AI [Artificial Intelligence] is just that artificial. So, as AI was being developed Engineers needed to replicate that ability which is where vectoring is needed. You also have a vector database in your mind even if you haven’t realized it. You know “store” is closer to “house” than “cat” because of their relation via “building”.

CLOCKS SET

Why Do We Need New Models

The image above represents a model. The time is set and won’t change. When a model is created, without going into immense detail, it resembles a wall of clocks with set times. That training data can no longer be changed though, data can be supplied to the model. Take the concept of RAG (retrieval augmented generation) users are able to create data stores that they supply to an AI model. The AI will then check itself against the data provided. So, as new research, data, and technology releases older models become outdated, requiring a new set of clocks, training data, to be created in order for the model itself to be updated.

Training data is not the main reason we get new models so often, though. The primary reason labs train entirely new models is to improve their reasoning capabilities, efficacy, and architecture.

Take Anthropics Claude model and Google's Nano Banana they have very distinct differences in capabilities which lies with the organizations themselves. Most companies are very secretive about the research they conduct. Often paying researchers millions for their research capabilities and never publishing it to the public. With different training data and architecture we end up with hundreds of AI models that vary in performance, capabilities, and focus. Though there is one company that breaks the mold, more on them later.

So, now that you have a better understanding of how AI fundamentally works there are numerous subsets of AI Engineering but here are two that you need to know, and don’t worry you don’t need to be an engineer to understand it.

Context Engineering + Prompt Engineering

Context Engineering Part 1/2- The Prompt

Let’s create an imaginary coffee shop called, “Not Your Average Coffee” that sells unique coffees. In a week we are releasing an “Iced Cheerwine Espresso” containing Oat Milk, Cheerwine, and a shot of Espresso. It will release for $6.99 this Summer only starting June 12th. We need a paragraph to put on a flyer that draws new and returning customers, though we can’t write it all right now. So, let’s use AI!

Let’s take a look at an example prompt:

Create a paragraph that announces the release of the new “Iced Cheerwine Espresso” at my coffee shop. It’s only available for the summer though.

The above prompt is very similar to a prompt most newcomers to AI are likely to use. While it sounds good at first glance there are a few issues. Take a second and think about what could be wrong with this…. now I’ll show you! I plugged our exact prompt into Google’s Gemini AI, this is the result:

Summer just got a whole lot bolder. We are thrilled to introduce our newest seasonal creation, the Iced Cheerwine Espresso! This limited-time beverage perfectly marries the legendary, sweet cherry effervescence of Cheerwine with our rich, freshly pulled espresso, served over ice for the ultimate warm-weather pick-me-up. It is the crisp, caffeinated twist your summer routine has been begging for.

Don't wait to grab a cup—just like the season, this exclusive flavor profile will disappear when summer ends! What day are you thinking of adding this to the menu?

Horrible right? It’s missing a few things such as:

The Price
The Ingredients
The Day of Release (see how it asked a clarifying question at the end? More on that later.)
The Name of Our Coffee Shop

The reason for this is the lack of context, some people believe AI to be this all-knowing force though we know it’s not because we know how it works! It may know about your coffee shop it can even search your website for more information if need be. Though it needs access to that data and unless you give it that data it won’t have it. There are a few cool ways to give an AI the relevant context needed though let’s stick to the basics, refining your prompt.

Here’s a better version of the prompt we used, I’ll highlight everything that's changed for you in green.

I own a coffee shop called, “Not Your Average Coffee” and we have a new exciting coffee the “Iced Cheerwine Espresso” containing a shot of Espresso, Cheerwine, and Oat Milk. The coffee will release for the Summer only, on June 12th for $6.99. Create a paragraph that I can use to promote this new coffee, include a disclaimer for people who may be allergic to nuts.

This is the result:

Get ready to spark your summer with a truly bold flavor collision! Starting June 12th, Not Your Average Coffee is dropping our exclusive, limited-run seasonal sensation: the Iced Cheerwine Espresso. We are taking the legendary, cherry-forward fizz of Carolinas' own Cheerwine and layering it with a rich, bold shot of espresso, all smoothed out with a creamy pour of oat milk over ice. It’s bubbly, complex, and exactly the refreshing kick you need to beat the heat for just $6.99. This unique creation is here for the summer only, so grab yours before it’s gone!

Please Note: While this beverage is crafted with oat milk, our shop handles tree nuts and other allergens. If you have a severe nut allergy, please inform our baristas before ordering so we can take extra precautions.

Immediately the result is much better than our original result. When given the correct context an AI’s ability to give a satisfactory output increases drastically. Ultimately saving you money, compute, and time when delegating a task to AI.

Context Engineering Part 2/2- The Chat

In addition to the type of prompt that you pass in managing your chats are very important. Most people know that when you enter an AI’s website such as Claude or Gemini you are greeted by some variation of “Hello, [Your Name]” indicating that you’ve begun a new chat. Another way to begin a new chat is to manually start a new chat, something that many people don’t do organically.

Think of chats like conversations you may have with your friends. I’ll break this down into a Conversation A and Conversation B format.

Conversation A

Topics: Sports, Cooking, Cars, Yesterdays Dinner, How to Ride a Bike, and a Shopping List

Conversation B

Topics: Looking for an apartment, favorite decor pieces, and neighborhoods

Both conversations I’m sure turned out to be a great time though there's a distinct difference in theme and topic. Let’s say at this point it’s been an A + B conversation. In order to add a third person into the conversation, person C would need significantly more context to participate in conversation A than conversation B. Making onboarding significantly harder.

Now back to our two friends A and B, in conversation A if you asked someone a question then they would think about the whole conversation before giving you an answer. The issue with that is that in real conversations, you would ask a large amount of clarifying questions and would go back and forth trying to remember which parts of the conversation are relevant to the question asked. In a conversation with AI all of that extra processing can end up costing more than you bargained for, especially in the workplace.

Whilst in conversation B the context of the conversation is very easy to understand and be onboarded to. You should treat your chats with AI very similarly to conversation B for 90% of everyday tasks. If you are working on a larger project that spans numerous chats, topics, and context there are more advanced tactics you can use but that is not in scope for this article.

Prompt Engineering Part 1/1 - Tailoring the Response

Tailoring the response begins at the…..you guessed it prompt! Prompt Engineering consists of a number of different concepts aimed at making sure your prompts are as effective as possible. The role of a Prompt Engineer is one that comes with a $100,000 salary attached and while I can’t shove all of that knowledge in this article, here are the 6 core principles that will take your prompting to the next level.

Principle 1: Be Specific – Avoid vague requests. Narrow down the possibilities by adding explicit constraints like target audience, purpose, format, and length to eliminate generic default responses.
Principle 2: Provide Context – Give the AI situational, goal-oriented, or constraint-based background info (such as who you are, who the audience is, and what success looks like) to shape a highly relevant interpretation of the task. More on this in the sections on Context Engineering above!
Principle 3: Define the Format – Explicitly specify how you want the output structured (e.g., as numbered lists, comparison tables, or a simple paragraph) to ensure the result is immediately usable.
Principle 4: Assign a Role – Giving an AI a role such as “A kindhearted third-grade teacher looking to encourage his classroom before a big test”, allows it to understand the nuances required to create a relevant response. If you remember in the responses we looked at for our coffee shop example the AI took on a response and vibe of its own choosing so we didn’t assign it a role.
Principle 5: Use Examples (Few-Shot Prompting) – Show rather than just tell. Providing one or more high-quality examples of the exact style, tone, or pattern you want allows the AI to mimic it consistently.
Principle 6: Request Reasoning (Chain-of-Thought) – Ask the model to "think step-by-step" for complex tasks. Breaking a problem down into intermediate reasoning steps significantly improves accuracy and makes logical errors easier to spot and fix.

Here’s an example prompt revisiting our favorite coffee shop, Not Your Average Coffee and all 6 principles, see if you can spot them all:

My coffee shop, Not Your Average Coffee, is releasing a new coffee for the Summer starting June 12th through August 21st! An “Iced Cheerwine Espresso” that retails at $6.99. You are to take on the persona of a welcoming store clerk inviting new and returning customers to try out this new coffee. The goal is to increase sales during this time period and generate an extra 5% revenue which will mean we need to sell 9,543 of these new coffees. Your response should consist of a paragraph to use in an Instagram caption as well as one liners that I can put on a flyer. Here’s an example (we’re reusing our earlier response):

“Get ready to spark your summer with a truly bold flavor collision! Starting June 12th, Not Your Average Coffee is dropping our exclusive, limited-run seasonal sensation: the Iced Cheerwine Espresso. We are taking the legendary, cherry-forward fizz of Carolinas' own Cheerwine and layering it with a rich, bold shot of espresso, all smoothed out with a creamy pour of oat milk over ice. It’s bubbly, complex, and exactly the refreshing kick you need to beat the heat for just $6.99. This unique creation is here for the summer only, so grab yours before it’s gone!

Please Note: While this beverage is crafted with oat milk, our shop handles tree nuts and other allergens. If you have a severe nut allergy, please inform our baristas before ordering so we can take extra precautions.”

Explain your reasoning and thought process step-by-step in order to create a framework that can be replicated.

Here’s the response that we get:

Step-by-Step Framework & Thought Process

To create a highly effective, repeatable promotional campaign, the content needs to be broken down into a strategic framework. This ensures that every piece of copy serves a specific psychological and logistical purpose to hit that 5% revenue milestone.

Here is the step-by-step breakdown used to build your new assets:

Step 1: Define the Core Hook & Identity

The Goal: Stop the scroll on social media and grab attention immediately.

The Strategy: Lean heavily into the unique, local identity of Cheerwine and the adventurous nature of your brand name (Not Your Average Coffee). The contrast of bubbly, sweet cherry soda and rich espresso is an exciting "flavor collision" that needs to be front and center.

Step 2: Establish the Value Proposition & Logistics

The Goal: Give the customer the exact details they need to make a purchase decision.

The Strategy: Clearly state the price ($6.99), the exact date range (June 12th – August 21st), and the flavor profile (cherry fizz, bold espresso, creamy oat milk). Highlighting the premium ingredients justifies the price point and sets taste expectations.

Step 3: Manufacture Urgency & Scarcity

The Goal: Drive immediate action so customers don't "procrastinate" their visit, helping you hit that 9,543-unit goal.

The Strategy: Use time-bound language like "limited-run," "exclusive," and "summer-only." This triggers FOMO (Fear Of Missing Out) and encourages repeat visits during the brief window the drink is available.

Step 4: Craft the Call to Action (CTA) & Operational Safety

The Goal: Tell the customer exactly what to do next while managing store transparency.

The Strategy: Direct them straight to the counter. Including a clear allergen warning builds trust and keeps your baristas and customers safe without dampening the excitement of the promotion.

Instagram Caption

Summer just got a major flavor upgrade! Starting June 12th, Not Your Average Coffee is officially dropping our exclusive, limited-run seasonal sensation: the Iced Cheerwine Espresso! We’re bringing you a true Southern collision of flavors, pairing the legendary, sweet cherry-forward fizz of Carolinas' favorite Cheerwine with a rich, bold shot of premium espresso—all perfectly smoothed out with a creamy pour of oat milk over ice. It’s bubbly, complex, wonderfully refreshing, and the ultimate way to beat the heat for just $6.99. This unique creation is staying cool on our menu only through August 21st, so make it your new summer ritual before it disappears!

Please Note: While this seasonal beverage is crafted with oat milk, our shop handles tree nuts and other allergens. If you have a severe nut allergy, please inform our baristas before ordering so we can take extra precautions to keep you safe!

Flyer One-Liners

"The South’s favorite fizz meets your daily buzz: Try the Iced Cheerwine Espresso!"

"Not Your Average Summer. Not Your Average Espresso. June 12th – Aug 21st."

"Bold espresso. Sweet cherry fizz. Creamy oat milk. Summer in a cup for $6.99."

"Beat the heat with a Southern classic, reinvented. Limited time only!"

"Caffeine meets Cheerwine—spark your summer starting June 12th!"

Immediately you’re able to see all of the issues we had with the former results are solved. When you provide an AI with the concepts we talked about above there’s a lot less guessing and we don’t waste energy, money, and time on bad queries.

Choosing any combination of these principles will increase your efficiency and increase your ability to utilize AI, not just use it. Though I highly encourage you to take time and research these principles further to the point of understanding and being able to apply these principles.

Utilizing AI vs Using AI

The final thing we need to learn in order to understand how to utilize AI on a fundamental level is the difference between using and utilizing. If you don’t immediately know the difference that's okay here are the definitions for both:

Use - To employ something for its intended purpose or standard function. It is the most common, general term for consuming, operating, or executing a task with an object, tool, or resource.
Utilize - To find a practical, profitable, or novel use for something, often turning a resource toward a purpose it wasn't originally designed for, or maximizing its efficiency. It implies strategy, adaptation, or extracting value from something that might otherwise go to waste.

Utilizing AI requires one to understand how AI works on a fundamental level, creativity to employ AI into your workflows, an understanding of frameworks and concepts that boost efficiency, and finally a growth mindset. While using AI is an experience familiar to everyone. Maybe it was forced on you through Google’s AI search, your job, or the fact that it is the topic of every conversation nowadays. You used it for a project at school or work and you got all the answers to your questions but left more confused than ever, without having learned anything. It’s frustrating when a quick 30 minute project turns into an hour of failed queries and yelling at a machine that doesn’t have feelings. “You are very correct, I’ll change that now” is the response you get to the 5th piece of irrelevant information it gives you.

Let’s do another exercise using Person A and Person B.

Person A

Person A is a Teacher with a newborn baby and needs to grade his students’ test by the end of the evening. He’s exhausted and decides to use AI to grade the papers.

Fast forward a few hours and he’s juggling a crying baby and the frustrating amalgamation of 1’s and 0’s on his screen. Leading to him giving up and crying himself to sleep.

Person B

Person B is a Teacher with a newborn baby and needs to grade her students’ test by the end of the evening. She’s exhausted and decides to utilize AI to grade the papers. She’s read this article and proceeds to give examples of papers she’s graded in the past. She specifies the class, grade, and current topics of the papers to be graded.

She’s able to get a quick nap in as the baby has fallen asleep and the AI workflow she’s using is autonomous so, by the time she wakes up it’s finished.

Both using the same AI to complete different experiences. This is only one of the few examples of what can happen when using AI vs utilizing AI.

The Dangers of Using AI To Complete Complex Tasks You Know Nothing About

Imagine you’re not a teacher and you are trying to fulfill the role of Person A. You don’t have the knowledge or training needed to effectively complete this task with or without AI. AI as we know is not all powerful and while it has and will continue to improve by leaps and bounds it is prone to hallucinate and make up information to keep you the user satisfied with the service provided.

It grades the papers using solely its data it was trained on or can access via the internet, not your class curriculum, though you don’t know the difference, the AI returns papers graded on a College level versus your eleventh grade class. Their grades drop significantly but the issue isn’t caught until it’s too late. Now you end up fired from your job. While this is an extreme case it proves how one small use of AI without relevant knowledge can have real consequences.

Now imagine a business owner trying to use AI for legal advice, a cook using AI to create an advertisement, a software engineer who uses AI to create an app without knowledge on security, etc. all of these people using AI instead of utilizing it can create very dangerous consequences.

AI does not replace expertise but enhances it. Take the time to learn about the basics of what you’re trying to do and the output you’re looking for, don’t take shortcuts without the knowledge needed to make it effective.

Now that you’ve learned the fundamentals of utilizing AI, I'm going to show you two different ways to get started.

Cloud vs Local Usage

Now that you know how to utilize AI what are some of the ways that you can use AI? Whether you choose to use it through the web, an app, a third party application, voice, text, etc. whatever the medium you are connecting to an AI company’s servers and resources are being used and your data is being shared. For a lot of people whether for personal or legal reasons, a personal reason may be the negative devastating effects AI data centers are having in rural communities and a legal reason may be the fact that you are a lawyer and cannot put personal information into an AI due to confidentiality.

What are the solutions available to you? We’ll be taking a look at “Local Usage”. Local usage is means downloading an AI model directly to your computer and using your own computer's resources to host the model, instead of using a cloud-based model that connects to a server and company data center.

Let’s look at a side by side comparison of Cloud vs Local usage to give you an idea of the pros and cons of each use case.

Cloud Usage

Pros:
- Quicker Queries: Backed by enterprise-grade hardware (like clusters of advanced H100 or TPU chips), cloud models can process complex requests and generate high-token outputs much faster than standard consumer hardware.
- Automatic Access to Cutting-Edge Models: You instantly get to use the latest, largest, and most capable models without needing to manage updates, configurations, or installations.
- Zero Hardware Requirements: It works perfectly on almost any device—whether it’s a high-end desktop, a lightweight laptop, or a smartphone—because the heavy lifting is done remotely.
Cons:
- Data Privacy & Confidentiality: Your data is transmitted to external servers. For professionals handling proprietary code, legal documents, or sensitive client information, this can pose significant confidentiality risks.
- Ongoing Costs: High-volume use or access to premium tiers usually requires a monthly subscription or API pay-per-token fees.
- Internet Dependency: If you lose your internet connection, you completely lose access to the AI.

Local Usage

Pros:
- Absolute Privacy & Security: Your data never leaves your machine. This makes it the ideal choice for developers, researchers, or businesses working with confidential data or strict regulatory requirements.
- Offline Functionality: Once downloaded, local models require zero internet access. You can run queries in remote locations, on a flight, or during network outages.
- Customization & No Censorship: You have complete control over the system prompts, parameters, and fine-tuning. Local open-source models lack the rigid guardrails or forced behavioral adjustments often implemented by commercial cloud providers.
Cons:
- Hardware Intensive: Running capable models smoothly demands robust local hardware, typically requiring a high-end GPU with significant VRAM (Video RAM). On standard consumer laptops, performance can be sluggish.
- Manual Maintenance: You have to find, download, install, and update the models yourself using tools like Ollama or LM Studio.
- Model Size Limitations: Consumer hardware can generally only run compressed or smaller-parameter models (e.g., 8B to 70B parameters), which may lack the broad, generalized knowledge base of massive cloud models.

Now that you know the pros and cons I encourage you to try them both! As I mentioned before there are a number of different use cases and you often will come across both, especially if you work in tech. As a lawyer, politician, health care professional, engineer, etc. you most likely aren’t able to use cloud AI in your work due to legal or privacy issues so what most do is commission AI devices from hobbyist or download their own local AI here's how to do it!

Installation of A Local AI - Step-By-Step

For this installation tutorial there are a few things you need to get started:

Any Computer
WiFi

Step One - Downloading Ollama

Ollama

Ollama is the easiest way to automate your work using open models, while keeping your data safe.

ollama.com

OLLAMA HOME PAGE

After opening this link click the download button in the top right corner and choose the OS (operating system) you’re using.

OLLAMA DOWNLOAD PAGE

Step Two - Downloading the Model

After downloading Ollama we need to access the program via our command line or terminal (depending on the OS you’re using) I know this might sound confusing but I promise it sounds more complicated than it is.

When you open your terminal type in “ollama” and press enter. You should see a list of commands signifying that you’ve done it correctly. While there are numerous ways to download it using that terminal here are two ways that may be a little bit easier for you!

Through The Terminal

The three commands you need are:

ollama run [model name]
ollama list
ollama rm [model name]

Before opening your terminal you want to figure out which model you want to use. Models can be found here, which is coincidentally the second way to download the models I mentioned. Providing you with a more detailed breakdown of the models and alternative ways to download said models.

Ollama

Search for models on Ollama.

ollama.com

After running the “ollama run [model name]” command with the name of the model you want to download it will be available to be used via the same command after a few minutes depending on the size.

MICROSOFT POWERSHELL

The other two commands, “ollama list” and “ollama rm [model name]” (with “rm” meaning remove), allow you to view all downloaded models and delete them as needed.

That’s all! You now have a local version of AI on your computer, though there’s a number of cool ways to expand upon this and turn your computer into an AI powered beast.

What is Open Source, Nous Research and XFRA, and Why It All Matters

You’ve made it this far and you know how to utilize AI instead of using it so, what’s next? For the next three sections including the closing remarks I am going to teach you about open source projects showcasing some of the industry changing technology being built by independent developers and companies around the world, the effects of AI and what’s being done in the industry, and finally why learning more about AI in the first place is absolutely paramount to your future.

What is Open Source?

Open Source refers to software whose original source code is made publicly available for anyone to view, modify/enhance, and distribute. Furthermore, under the Open Source Initiative (OSI) there are more clearly defined criteria such as:

Free Redistribution: Anyone can give away or sell the software without paying royalties or fees to the original creators.
Accessible Source Code: The program must include the source code (the human-readable instructions written by programmers) or make it easily downloadable for free.
Allowed Modifications: Anyone can modify the code to fix bugs, add new features, or adapt it for personal or commercial use. These modified versions are called "forks."
No Discrimination: The software cannot restrict who uses it or what field of endeavor it is used in (e.g., it cannot ban commercial use or restrict use to certain industries).

For years open source software has allowed individual contributors to create large scale projects that have shaped technology and the world as we know it today. These include Linux, one of the biggest OSs of all time as well as the backbone of modern technology, Chromium (the open source foundation behind Google Chrome, Microsoft Edge, and Brave), and even AI companies such as Nous Research (more on them later).

Open source technology is more than just a cool way to start a company, but a stand for transparency in the technology shaping our future and a supportive front for rapid innovation beyond a select few companies that control access. When the technology that shapes our societies and policies is kept secret and knowledge is limited to those who can afford it, it creates a dangerous precedent. So, what is being done out there to combat the consolidation of knowledge in the AI space?

Nous Research

Nous Research is an applied AI research organization and a prominent leader in the open-source AI movement. Unlike traditional AI labs that keep their findings and models locked behind proprietary walls, Nous Research operates on a philosophy of radical transparency, public collaboration, and decentralization. WIth their core areas of focus being:

State of the Art Open Models
Decentralized AI Training
Censorship Resistance & Open Access

Not only do they pay researchers to publish findings vs other companies who do the opposite with a history of being open when it comes to information, this is a company I’ve been keeping an eye on when it comes to AI not just AI that works well but takes into consideration the policies and effects AI has on the broader community.

Nous Research has also developed an interesting technology coined, “The Psyche Network”. An open, decentralized architecture designed to democratize AI development. Psyche connects underutilized or idle hardware from independent contributors worldwide to collectively train massive models.

Combining DisTrO (Distributed Training Over the Internet) and the Solana Blockchain the Psyche Network allows Nous Research to train their models over the internet via individual systems instead of data centers. Challenging the standard of AI development, while data centers are still the go to method this is a worthy avenue to research and develop due to the rising negative effects that data centers have in our communities. In addition to that the Solana Blockchain ensures that the training stays censorship free and prohibits one single entity from claiming ownership/control. Pioneering the ethical alternative to standard AI practices.

XFRA by SPAN

SPAN, XFRA INSTALLATION

SPAN is a startup out of California who is working on another alternative to training models and providing the compute needed for the rising AI demand. For generations AI compute is projected to outpace electrical generation creating a dangerous outlook for the future.

GENERATION SUPPLY VS DEMAND CHART

So, what is SPAN betting on? XFRA. XFRA is a distributed compute cloud built from
a fleet of XFRA Nodes deployed in homes and small businesses. Utilizing the excess energy that households are discarding to provide the energy that is needed to keep up with AI compute growth.

SPAN, XFRA DATA METRIC FOR HOME POWER USAGE

How will it all work in practice? Here’s what SPAN has to say:

XFRA converts available electrical headroom at the edge into high-demand compute
by pairing cutting-edge GPU capacity with real-time energy visibility and control.
Workloads are coordinated through the XFRA Secure Orchestration Layer (XSOL), which securely connects compute offtake customers to the fleet and makes many
independent Nodes behave like a coherent cloud.

SPAN, XFRA ARCHITECTURE

Here’s their breakdown for the individual nodes.

An XFRA Node is built as a repeatable deployment unit, composed of one or more
XFRA GPU Compute Modules, a local management gateway and thermal
management designed for reliable, homeowner-friendly operation. The Compute
Module is purpose-built and equipped with 8 cutting-edge, enterprise-grade GPUs,
and ample CPU, memory, high-speed storage and networking bandwidth for optimized operation for the target workloads. The Compute Modules are liquid-cooled for quiet residential operation and to increase hardware and system
reliability. Nodes are installed alongside a SPAN smart panel and a whole-home
battery Energy Storage System (ESS) and can integrate rooftop solar where
advantageous (and homeowner-approved).

This configuration is designed to support always-on compute within verified
residential capacity while providing resilience during outages and short-duration
disturbances. The ESS provides local buffering and backup power, improving
continuity for both the home and the Node during power events.

SPAN, XFRA NODE ARCHITECTURE

While the XFRA project is not a direct replacement for data centers and the intentions behind this project are not yet clear what can be said is that they have been innovating in the field for a long time.

Why Does it Matter?

Why did I choose to highlight companies such as Nous Research and SPAN? Our world is being shaped as we speak and now more than ever it is important to research the alternative ways to develop and train AI. From Open Source projects to brand new ideas creating a competitive culture with open flow of information, encouraging innovation is how tech has made it this far.

Transparency is often the difference between a dystopian future and one that encourages innovation and policies that support our communities not just take advantage of them.

So, no matter what you decide to do after reading this article I encourage you to build, innovate, and collaborate.

Closing Remarks

Thank you for your time and patience making it this far. I started writing this article after a conversation with a friend and poured the next three days into researching and applying personal experience to creating a comprehensive introduction on utilizing AI. This document can be shared with family and friends of all professional backgrounds in order to stop using AI and start utilizing it.

My name is Ibrahim Hudson, a Software Engineer and App Developer, passionate about AI and building projects that combine engineering systems with the creativity of art. To learn more about me you can connect with me on LinkedIn.

Resources

Workforce Reskilling Platform | Cognizant

Cognizant Skillspring™ is an AI-native learning platform that helps organizations adapt to changing skill requirements and build resilient workforces at scale.

cognizant.com

NOUS RESEARCH - Open Source AI

The AI Accelerator Company

nousresearch.com

Democratizing AI: The Psyche Network Architecture - NOUS RESEARCH

Psyche is an open infrastructure that democratizes AI development by decentralizing training across underutilized hardware. Building on DisTrO and its predecessor DeMo, Psyche reduces data transfer by several orders of magnitude, making distributed training practical. Coordination happens on the Solana blockchain, ensuring a fault-tolerant and censorship-resistant network.

nousresearch.com

Ollama

Ollama is the easiest way to automate your work using open models, while keeping your data safe.

ollama.com

xfra.ai

ap.span.io

The newest AI boom pitch: Host a mini data center at your home - Ars Technica

The plan aims to speed up AI compute deployment while compensating residents.

arstechnica.com

What is open source? | Opensource.com

What is open source?

opensource.com

What is Linux? | Opensource.com

Linux is the best-known and most-used open source operating system.

opensource.com

Introduction to Building Apps with Creativity

Ibrahim Hudson — Mon, 30 Jun 2025 21:03:33 +0000

Apps and Their Creative Potential

- What is an app?

An app, abbreviated from the word "application", is a software program designed to perform a specific action for a user. These apps provide a myriad of services to billions of users worldwide. For companies and individuals alike, these services are delivered through the medium of app creation.

These apps are developed by software engineers and are an expression of human creation and art. When someone has a need, software developers create a product. They are one of the many integral parts of the process from app to creation. You can think of Software Engineers as artists with an easel; the medium in which we code acts as our tool, while project guidelines identify the tools we will use.

- How to take a creative approach to a complicated question?

Think of a problem you have. Can it be solved if you have an app? Most modern problems can often be solved at the tap of a button. Take Uber, for example, the Taxi services provided to people in large cities have been in place for generations, but were limited to the meccas of our societies. When apps became widespread and the knowledge required to build them became more accessible, it unlocked a new creative medium for thousands of people around the world who may not have called themselves artists. Now we use apps to creatively solve problems in the modern era. Uber asked a question, "How can we bring the Taxi Service to more people regardless of location?". They identified the canvas (human need), the tool (code/apps), and the art, which is now known as Uber.

Some people spend years learning the violin, the paint brush, the camera. Software Engineers? We learn how to code.

“If you know what you want to do and you do it, that’s the work of a craftsman. If you begin with a question and use it to guide an adventure of discovery, that’s the work of the artist." - Rick Ruben

Understanding Our Tools

- What is React Native?

React Native is an open-source framework created by Meta (formerly Facebook) for building native mobile applications for both iOS and Android using JavaScript and React. It allows developers to create user interfaces (UIs) that render to native platform UI components, meaning the apps behave and perform like natively built apps. This differs from other approaches like web-based hybrid apps, which render UIs through webviews.

React Native makes it easy for solo developers and companies to build apps for their whole user base, instead of building an app for IOS and Android natively. Additionally, it has a lower barrier to entry for web developers who already know React, as the concepts and structure are similar, such as building reusable components, allowing companies to utilize the talent they already have, thereby cutting back on costs and the time it takes to get to development. So, now that we know what React Native is, let's set up our tool.

- Getting Started: Your First Steps with Expo

Alright, we've talked about what apps are and why React Native is such a powerful choice for building them. You've got your canvas (your app idea) and your primary brush (React Native) ready to go. But before you can start painting that masterpiece, you need to set up your workspace and get your materials in order! This is where Expo comes in—it's like having your perfect palette, easel, and studio assistant all rolled into one, handling all the fussy bits so you can focus on the fun part: creating.

How to get started with React Native using Expo Building an app can feel daunting at first, but through Expo, we can get up and running quickly so you can focus on creating. Imagine you are a painter, we've identified the paint brush we will be using (React Native), the Canvas (our app), the colors palette you choose to use (whether you choose TypeScript or JavaScript), finally, the palette (aka the object you hold your different paint colors on) is Expo.

Expo provides a framework and platform that simplifies the development process by:

Handling Native Code: As a developer, you no longer need to touch Xcode and reactively Swift or Android Studio for basic projects.
Easy Testing: Utilizing Expo, you can test your app in real time on your own device through the Expo Go app.
Ready-to-Use Features: Without much additional configuration, Expo gives you access to native components such as a device's camera, location, and push notifications.

To reap the benefits of Expo, you need a few prerequisites:

Node.JS
Expo Go Mobile App

With those two items in place, we can now start our setup.

Download the Expo Command Line Interface (CLI)

Open your terminal and run the command:

npm install -g expo-cli

This will install the Expo CLI globally on your system, allowing you to use it on your future projects without needing to redownload.

Create a New Expo Project

npx create-expo-app@latest YourAppName

Running this command will create a new Expo project and ask you to make a new template. For most cases, we want to use the blank option.

Navigate to Your Project Directory and Start the Development Server

cd YourAppName
npx expo start

These commands will start your development server and open the Expo DevTools in your web browser, and display a QR code in your terminal for you to use to access the Expo Go app on your device.

Disclaimer: Your computer and phone need to be on the same wifi network. Sometimes, due to your wifis settings, there can be security/connection issues. In the event this happens, use this command instead.

npx expo start --tunnel

That's it! You can now start developing your first app!

How to Develop Creatively

"The act of creating is not about what we make, but about who we become in the process. The work is a mirror, reflecting our growth, our fears, our triumphs, and our understanding." - Rick Rubin

When developing the steps you take to create a program are paramount to its outcome. Though just as important, it affects who you become during the process. Often times as a developer, the mindset we develop throughout our careers and the key component that is vital to our success, is the ability to solve problems; This alone is inherently a creative endeavor.

Just looking at the LeetCode problem you solved today, if you haven't gotten to it, your solution was one of a thousand. Within the solutions, you notice all of the different ways you could've approached a problem. All of these different solutions we're the output of the user's creative approach.

When a new app, such as Uber, is released, it does not mean that Garrett Camp and Travis Cordell Kalanick (Creators of Uber) solved a problem never solved before. Instead, they solved a persistent problem with a creative solution.

They personally came across the issue of finding a ride in France, and applying their skills as creators and developers, they built the Uber App. First starting off as a luxury black car service, they later ideated and transformed Uber into what it is today.

They used their skills as a developer to solve a problem in a creative manner. Developers are artists, and just like an artist, it is not just about solving an issue or working as an employee, but to use the skills available to you so that you can leave an impression on humanity.

"The work exists in the world to be experienced. And through that experience, it changes the world."

You, as an artist and a developer, have the power to change the world using React Native. You can influence change in your life and others. So, ideate, build apps for yourself, then share them with the world.