Stack Traces

How Your Phone and Laptop Can Hear Your Voice but Ignore the Music Playing Beside You

George Wanjohi — Mon, 26 Oct 2026 06:27:21 GMT

Understanding noise cancellation, voice isolation, and the clever technology that makes modern calls sound clearer.

Have you ever been on a phone call while music is playing in the background, only for the person on the other end to tell you they cannot hear it?

I recently noticed something interesting while talking to my girlfriend on my Samsung Galaxy A26. When noise cancellation was enabled, she could barely hear the music playing from another device in the room. When I switched it off, the music became audible to her.

I noticed something similar when making calls on my Windows laptop. I could play music through the laptop's speakers while talking, yet the person on the other end could not hear the music.

At first, this might seem strange. After all, the microphone is in the same room as the music. If the microphone can hear the sound, why doesn't the other person hear it?

The answer lies in how modern devices process audio.

1. Your microphone doesn't simply transmit everything it hears

A basic microphone captures sound waves. It doesn't inherently know whether a sound comes from your voice, a television, music, a passing car, or someone talking nearby.

If you connect a simple microphone to a recording device without any additional processing, it will generally capture a mixture of all these sounds.

Modern phones and computers do something more sophisticated.

Before sending your voice to the other person, they can process the microphone's input to reduce unwanted sounds and make speech easier to understand.

Depending on the device and calling application, this processing can include:

Noise suppression: Reduces unwanted background sounds such as fans, traffic, and air conditioning.
Voice isolation: Attempts to preserve your voice while suppressing other sounds, potentially including music and other people speaking.
Acoustic echo cancellation: Removes audio that the microphone picks up from the device's own speakers, particularly sound from the other person on the call.

These techniques are related, but they are not identical. Microsoft documents both conventional and machine-learning-based approaches to audio processing for communication applications. Source: Microsoft Learn.

2. What happened on my Samsung Galaxy A26?

When I was speaking to my girlfriend, I had music playing from another device.

With the noise cancellation feature enabled, she could not hear the music. When I turned the feature off, she could hear it.

One likely explanation is that the phone's microphone processing was suppressing sounds that were not recognised as my voice.

Imagine that the microphone picks up this mixture:

My voice: "How was your day?"
Background music: a song playing from a laptop.
Environmental noise: fans, traffic, or other sounds.

The phone processes that mixture before transmitting the audio.

With voice-focused processing enabled, it tries to preserve the speech while reducing the other sounds. The music may therefore be heavily reduced or removed from what the other person receives.

With less aggressive processing, more of the original sound mixture may pass through.

The exact behaviour depends on the phone's microphone mode, the calling application, and the available audio-processing features. I cannot confirm that every Galaxy A26 noise-cancellation setting works in precisely the same way, but the result I observed is consistent with voice-focused audio processing.

3. Why does the same thing happen on a Windows laptop?

This is where the story becomes even more interesting.

When I make a call on my Windows laptop, I can play music through the laptop's speakers without the other person hearing much of it.

There is a particularly relevant technology called acoustic echo cancellation.

Consider what happens during a video call:

The other person speaks.
Their voice comes through your laptop speakers.
Your laptop microphone picks up some of that sound.
Without processing, the other person might hear their own voice echoed back to them.

Echo cancellation helps prevent this by comparing the microphone input with a reference of the sound being played through the speakers. It estimates the component that came from playback and attempts to remove it from the microphone signal. Source: Microsoft Learn.

However, there is an important distinction: echo cancellation is primarily designed to remove speaker playback that would otherwise return to the caller, not to remove all music in the room.

If the laptop is suppressing music that you started playing yourself, other processing may also be involved, such as noise suppression, voice isolation, or audio processing built into the calling application or microphone driver.

For example, Microsoft Teams offers voice isolation that uses AI to distinguish the user's voice from other speakers and sounds. Other communication applications may use different methods. Source: Microsoft Support.

The key point is that the laptop is not necessarily sending everything the microphone captures. The audio can be filtered before it reaches the other person.

4. Is artificial intelligence involved?

Sometimes, yes.

Traditional audio processing can use mathematical techniques to identify and reduce predictable noise. Newer approaches can use machine-learning models to distinguish speech from other sounds.

A machine-learning model can learn patterns associated with human speech and patterns associated with competing sounds. During a call, it can use those patterns to estimate which parts of the audio should be preserved.

This does not mean every phone or laptop uses a generative AI model, nor does it mean that every noise-cancellation feature understands language.

Some systems use conventional digital signal processing. Others use machine learning, and some combine both.

The result can feel almost magical: you speak normally, music plays nearby, and the person on the other end hears a much cleaner version of your voice.

5. Why doesn't noise cancellation remove every background sound?

Because sound is complicated.

Music and speech can overlap in frequency, timing, and loudness. A song with vocals may contain sounds that resemble human speech. A loud television may be harder to suppress than a quiet fan.

The system must make a trade-off between removing unwanted sounds and preserving the speaker's voice.

If the filtering is too aggressive, it can make your voice sound unnatural, cut off words, or suppress quiet speech. If it is too weak, background sounds remain audible.

This is why a good call-quality system tries to balance clarity with natural-sounding speech rather than simply muting everything except the loudest sound.

6. How can you test this yourself?

You can conduct a simple experiment with your phone or laptop.

What you need:

A phone or laptop with a microphone.
A second device playing music.
Someone willing to listen on the other end of a call.

The experiment:

Start a call and speak normally.
Play music at a moderate volume from the second device.
Ask the other person whether they can hear the music.
Turn the available noise suppression or voice-isolation feature off, if the app allows it.
Repeat the test with the music playing at the same volume.
Compare how much of the music reaches the other person.

For your Samsung phone, try comparing the available microphone modes during an active call. For Windows, check the calling application's audio settings and the microphone's audio-processing options.

Keep the microphone position, speaker volume, and distance as consistent as possible. That way, you can better identify the effect of changing the processing setting.

7. Does this mean I can play music during every meeting without anyone noticing?

Not necessarily!

The behaviour varies depending on the device, the calling application, the microphone, and the audio-processing settings. Some music may get through, particularly if it is loud or contains vocals.

Also, suppressing music from the microphone is different from sharing your computer's system audio. If you deliberately share system audio in a meeting, the music can be transmitted directly to participants regardless of what the microphone is doing.

So, if you want to share a song during a presentation, use the application's audio-sharing feature rather than relying on your microphone to capture the music.

And if you want the other person to hear music playing in your room during an ordinary call, you may need to reduce or disable voice-focused filtering where that option is available.

Final thoughts

What I found fascinating about these two experiences is that the phone and laptop are not merely recording sound. They are processing it to make communication more effective.

My Samsung Galaxy A26 can suppress background music during a call, while my Windows laptop can keep speaker playback from becoming part of the transmitted conversation.

Although the underlying techniques differ, the broader idea is the same: the sound you hear in a room is not necessarily the sound another person receives over a call.

Modern communication devices can filter, suppress, and reshape microphone audio before transmitting it.

The next time someone tells you that they cannot hear the music playing right beside you, remember that your microphone may be hearing much more than your caller ever will.

Further reading:

Why Feedback Matters More Than Confidence

George Wanjohi — Thu, 22 Oct 2026 10:30:38 GMT

Confidence alone does not guarantee success; actual results and progress depend on calibrating actions against real-world feedback and data. [1]

Why Feedback Matters More Than Confidence

Confidence vs. Competence: Studies on decision-making show that high self-assurance or confidence does not reliably match actual skill or correctness. Overconfidence often leads to inflexibility and blind spots. [1, 2]

Feedback as Data: Treating feedback as neutral information—rather than an attack on self-worth—allows you to see what is working and what needs to change. [1, 2]

Confidence is a Byproduct: Real confidence is not a prerequisite or a magic ingredient you "fake"; it is a side effect that develops after taking action, measuring outcomes, and learning from mistakes. [1, 2, 3, 4]

How to Use Feedback Effectively

Unhook from Emotion: Separate the message from your ego. Ask what the data or the user/client/colleague's response tells you about the actual output.

Track Surprises: Notice when confident decisions succeed or fail unexpectedly, and adjust your assumptions accordingly.

Iterate in Small Steps: Test your work early, collect real-world responses, and adjust your approach one step at a time. [1, 2, 3, 4, 5]

A Beginner’s Guide to Git and AI for Electrical Engineers: Protect Your Drawings, Work Smarter, and Design Better

George Wanjohi — Mon, 19 Oct 2026 06:01:49 GMT

Author: George Wanjohi
Audience: Electrical and Electronics Engineers who use Autodesk Revit and AutoCAD
Reading time: 8–10 minutes

Introduction

As an electrical and electronics engineer, you spend a lot of time creating drawings, placing electrical components, routing cables, preparing schedules, and making sure your designs are correct.

You probably already know how important it is to save your work regularly. But what happens when a client requests changes, a drawing becomes corrupted, or you accidentally remove something that was working perfectly in the previous version?

What if you could ask an AI assistant to help you create a drawing, automate repetitive tasks, or identify potential mistakes before submitting your work?

There are two technologies worth exploring: Git and Artificial Intelligence (AI).

Git helps you manage changes to your project files. AI can help you automate design tasks, learn new techniques, and review your work.

You don't need to be a software developer to benefit from either technology.

Let's explore how they work and how you can start using them.

Part 1: Understanding Git without learning to code

1. What is Git?

Imagine you are designing the electrical installation for a commercial building.

You finish the initial design and send it to your client. The client requests changes to the lighting layout, and later asks you to restore the original arrangement in one section.

Without a proper version history, you might end up with folders like these:

Office_Design.dwg
Office_Design_Final.dwg
Office_Design_Final_Updated.dwg
Office_Design_Final_Updated2.dwg
Office_Design_Really_Final.dwg

We've all seen variations of this problem.

Git is a version-control system. It records snapshots of your files at different points in time, allowing you to review changes and return to an earlier version when necessary.

Think of it as a carefully maintained history book for your engineering projects.

For example, your project history might look like this:

Version	Description
Version 1	Initial electrical layout
Version 2	Added lighting fixtures
Version 3	Revised socket-outlet positions
Version 4	Updated distribution board schedule
Version 5	Corrected cable routing

Each version can have a description explaining what changed and why.

If a later change causes a problem, you can refer to an earlier version rather than relying on memory or searching through dozens of backup files.

Git is widely used in software development, but the underlying idea of tracking changes can be useful in other fields too.

2. Does Git upload your drawings to the internet?

No. Git does not require cloud storage.

Git can maintain a complete repository on your own computer. In this setup, your project history remains on your local drive unless you deliberately copy or transmit it elsewhere.

For this guide, we will use local-only Git.

We will not use GitHub, GitLab, or any other online repository.

This matters when your drawings contain copyrighted designs, client information, proprietary layouts, or commercially sensitive engineering work.

One important distinction: Git being local does not automatically guarantee that your files never leave your computer. Cloud-synchronised folders, AI applications, backup software, and other connected services may transmit data independently. We will address AI privacy later.

3. What do you need to install?

You only need two things to get started.

Git is the software that records your project history.

Download it from the official website:

https://git-scm.com/downloads

GitHub Desktop is an optional application that gives you a visual interface for working with Git. You can see changes and create saved versions without memorising commands.

Download it from:

https://desktop.github.com/

You can use GitHub Desktop without publishing your project to GitHub. When creating a repository, choose the option to create it on your local computer, and do not publish it.

For a beginner, I recommend using the visual interface first.

4. Create your first local repository

Let's imagine you have an AutoCAD project called Commercial_Building_Electrical.dwg.

First, create a folder on your computer:

D:\Engineering\Commercial_Building_Electrical

Copy your project files into this folder. Keep the original project somewhere safe until you have verified your new setup.

Open GitHub Desktop and select:

File → New repository

Enter the following information:

Name: Commercial Building Electrical
Local path: D:\Engineering
Repository location: The application will create a project folder under the selected path.

Choose the option to create the repository locally. Do not publish it.

Your project is now being managed by Git.

In Git terminology, this folder is called a repository. It contains your working files and a hidden .git directory that stores the version history.

You can continue opening and editing your drawings in AutoCAD or Revit as usual.

Git works alongside your existing engineering software. It does not replace it.

5. Save a version whenever you reach an important milestone

In Git, saving a version is called making a commit.

Imagine that you have just completed the first lighting layout.

Open GitHub Desktop. It will display the files that Git can see have changed.

Enter a summary such as:

Completed initial lighting layout

Then click Commit to main.

You have now created your first recorded version.

Later, after updating the distribution board schedule, create another commit:

Updated distribution board schedule

After correcting the cable routing:

Revised cable routing on Level 2

You can repeat this process throughout the project.

A commit is not the same as pressing Save in AutoCAD or Revit. Save your drawing in the engineering application first, then create a Git commit when you want to record a milestone.

A practical workflow

Open your drawing in Revit or AutoCAD.
Make your changes.
Save your drawing.
Open GitHub Desktop.
Review the changed files.
Write a short description of the changes.
Commit the changes.

You now have a history of your project.

6. What happens when you make a mistake?

Suppose you spend two hours changing the lighting layout, only to discover that the changes introduced a problem.

With Git, you can review the previous versions and recover an earlier copy of the affected file.

For binary engineering files such as DWG and RVT, Git generally cannot show you the individual drawing elements that changed in the way it can show changes in a text document. It can, however, preserve earlier file versions.

For example, you could maintain this history:

Initial electrical layout
Added emergency lighting
Revised emergency lighting
Corrected emergency lighting

If the latest version is faulty, you can retrieve the earlier file and open it in your engineering software.

Important: Restoring an older version can overwrite newer work. Recover it into a separate folder or filename first, then inspect it before replacing anything.

7. An important limitation: Revit and AutoCAD files are not ordinary documents

This is where we need to be careful.

A Word document, a text file, and an AutoCAD drawing are different types of files.

Git is particularly good at tracking text files. AutoCAD DWG files and Revit RVT files are largely binary files, which means Git generally stores a new version of the file rather than showing you a meaningful, element-by-element drawing comparison.

Large engineering files can also consume considerable disk space when many versions are saved.

There are additional considerations for Revit projects involving linked models, families, worksharing, and central models.

For these reasons, I recommend the following approach:

Use Git to track project milestones and preserve selected versions.
Keep all related project files together, including relevant families, links, and supporting documents.
Avoid committing temporary files, caches, lock files, and unnecessary generated files.
Keep a separate backup of important projects.
Do not use Git as a replacement for Revit's native worksharing or backup mechanisms.

For workshared Revit projects, follow Revit's central-model and synchronisation workflow. Do not treat a live central model as an ordinary file that can be freely replaced with an older Git version.

For AutoCAD, the built-in DWG Compare command can also help identify drawing differences between revisions.

Git is most useful here as an additional version-management tool, not as a replacement for the engineering software's native project-management features.

Part 2: Using AI to improve your engineering workflow

Git helps you protect your work. AI offers a different opportunity: reducing the time you spend on repetitive tasks and helping you inspect your designs.

But we need to distinguish between two types of AI tools.

General-purpose AI assistants answer questions, analyse information you provide, and help develop automation.

CAD-integrated AI tools can work with supported information inside engineering software and, depending on the available features, perform actions on the actual model or drawing.

The second category is particularly interesting for electrical engineers.

8. Start with AI as your engineering assistant

You can begin with a general-purpose AI assistant to help you understand engineering workflows and solve problems.

For example, you could ask:

"Explain how to create an electrical distribution board schedule in Revit. Assume I am a beginner. Give me step-by-step instructions and explain the purpose of each step."

Or:

"I have an AutoCAD electrical drawing with several layers. Explain how to organise lighting, power, emergency lighting, and cable-routing information into separate layers without changing the geometry of the drawing."

You can also use AI to help you understand error messages, learn commands, develop checking procedures, and plan repetitive work.

The key is to give the AI enough context.

Instead of asking, "How do I improve my drawing?", try:

"I'm working on the electrical layout for a two-storey commercial building. I want to improve the organisation of lighting circuits, socket outlets, distribution boards, and emergency lighting. Suggest a practical drawing-review checklist. Do not assume that you know my project specifications."

Specific questions usually produce more useful answers.

9. Can AI actually create Revit and AutoCAD drawings?

Yes, in some workflows, but the capabilities vary significantly.

There is a difference between asking AI to draw an illustrative image and asking it to create an editable, technically meaningful DWG or RVT model.

A generated picture of an electrical distribution board is not the same as a Revit model containing correctly configured electrical equipment, connectors, circuits, parameters, and schedules.

Here are three practical levels of AI-assisted design.

Level 1: AI guides you through the drawing process

This is the easiest starting point.

You describe what you want, and AI explains the steps to achieve it in Revit or AutoCAD.

For example:

"Guide me through creating a lighting layout for a small office. Explain how to place lighting fixtures, connect them to the appropriate circuits, and prepare a schedule. Ask me for any project requirements you need rather than inventing them."

You still perform the operations, but you have an assistant available to guide you.

Level 2: AI helps automate actual drawing operations

This is where things become more interesting.

AI can help create or modify automation scripts that interact with CAD software. These scripts can potentially place symbols, create layers, label circuits, generate schedules, or perform repetitive checks.

In AutoCAD, this could involve AutoLISP or other supported automation mechanisms.

In Revit, automation could use the Revit API or visual programming tools such as Dynamo.

You do not have to learn programming before benefiting from these approaches. AI can help develop the automation, while you explain the engineering rules and verify the results.

For example, you might request:

"Help me design an automation workflow for AutoCAD that checks whether electrical symbols are on the correct layers. First, explain the approach in plain English. Then propose a solution that reports possible errors without modifying my original drawing."

The safest first automation tools are those that report problems rather than immediately changing the drawing.

Level 3: AI works directly with your CAD model

This is the longer-term goal.

Autodesk is introducing AI capabilities into its design applications. Autodesk Assistant, for example, offers product guidance and, where supported, model-aware queries and task automation.

In AutoCAD, supported capabilities include AI-assisted workflows such as Smart Blocks and drawing-related assistance. In Revit, Autodesk Assistant capabilities are evolving, with supported workflows depending on the release and availability.

These features are worth investigating because they can bring AI closer to the actual design environment rather than requiring you to describe every detail manually.

Read the official documentation:

Revit AI features: https://www.autodesk.com/support/technical/article/caas/sfdcarticles/sfdcarticles/What-AI-features-are-available-in-Revit.html
AutoCAD AI features: https://www.autodesk.com/support/technical/article/caas/sfdcarticles/sfdcarticles/What-are-the-AI-driven-features-in-AutoCAD.html
Autodesk Assistant: https://www.autodesk.com/solutions/autodesk-ai/autodesk-assistant

Check the features available in your specific software version and licence. Some capabilities are previews, may require an Autodesk account, and may process information through Autodesk's services.

AI integration is promising, but it should not be assumed that every feature can independently generate a complete electrical installation design from a single prompt.

10. Can AI find and correct mistakes in electrical drawings?

Potentially, yes. However, identifying a suspicious condition and proving that it is an engineering error are two different things.

AI-assisted checks can be useful for finding inconsistencies, missing information, and deviations from explicitly defined rules.

Consider these examples:

Potential issue	What an AI-assisted workflow could do
Missing circuit labels	Identify unlabelled circuit elements
Inconsistent equipment naming	Flag naming differences
Duplicate annotations	Identify possible duplicate labels
Missing drawing information	Compare the drawing against a defined checklist
Incorrect layer usage	Identify objects on unexpected layers
Inconsistent schedules	Compare model data with schedule requirements
Changes between revisions	Highlight additions, deletions, and modifications
Potential electrical design issues	Flag conditions that need engineering verification

Some of these checks are better handled by conventional CAD rules, Revit schedules, model validation, or deterministic calculations than by AI alone.

For example, if a project requires a specific circuit naming convention, a rule-based check can verify it consistently. AI can help create the checking workflow and explain the findings.

For an electrical design review, AI could also help you investigate questions such as:

Are the circuit descriptions consistent with the project requirements?
Are the equipment ratings consistent with the approved design information?
Are the relevant drawing notes and schedules complete?
Are there apparent inconsistencies between the drawings and the specification?
Which items require further review by the responsible engineer?

The important limitation is that AI cannot reliably determine every electrical design error from a drawing image alone. It may miss information, misunderstand symbols, or make incorrect assumptions about design intent.

A technically correct review may require access to the actual model data, project specifications, calculations, applicable standards, and manufacturer information.

Never accept an AI-generated correction to a safety-critical electrical design without appropriate verification.

The engineer remains responsible for the design and its compliance.

Part 3: Protecting copyrighted engineering work while using AI

11. Local Git and local AI are two separate things

Keeping your Git repository on your computer does not automatically make your AI workflow private.

Git can store files locally. An AI application may still send prompts, images, drawings, or extracted model data to a remote service.

If client confidentiality and copyright are important, you should decide which AI workflow is acceptable before sharing any project information.

There are two broad options.

Option A: Use a cloud-based AI assistant

Cloud AI services can be useful for learning Revit, understanding AutoCAD commands, writing automation scripts, and discussing general engineering workflows.

However, you should not upload confidential drawings or client files unless you have confirmed that the service and its terms are appropriate for that information.

You can still ask general questions without sharing actual project files.

For example:

"I'm working in Revit on an electrical layout for a commercial building. Explain how to check circuit assignments and prepare a distribution board schedule. Do not use or require any client drawings."

This lets you learn without disclosing the project itself.

Option B: Run AI locally on your own computer

A local AI model runs on your computer instead of relying on a remote AI service for its model inference.

Tools such as Ollama and LM Studio can run supported language models locally. These can help explain tasks, develop scripts, and support workflows that you build around your CAD applications.

Web search, online integrations, software updates, and other connected features may still communicate externally, so local model inference alone does not guarantee that an entire application is offline.

A local model also needs sufficient hardware resources, and its performance will depend on the model and the complexity of the task.

Most importantly, running a chatbot locally does not automatically give it the ability to edit Revit or AutoCAD drawings.

To make that possible, you would need a compatible integration or automation workflow that lets the model interact with the engineering software. You would then need to test that integration carefully.

This is a worthwhile longer-term direction if you want to combine privacy with AI-assisted drawing creation and checking.

Part 4: A practical plan for getting started

You do not need to learn everything at once.

I would start with small, low-risk experiments and gradually build towards more advanced automation.

Week 1: Learn local version control

Install Git and GitHub Desktop.

Choose a small, non-confidential practice project. Create a local repository and record a few meaningful milestones.

Practise finding previous versions and recovering an older file into a separate folder.

The objective is simple: become comfortable maintaining a history of your work without uploading it.

Week 2: Use AI to learn your existing tools

Use AI to answer questions about Revit and AutoCAD.

Ask for beginner-friendly instructions, drawing-review checklists, and explanations of unfamiliar features.

Where possible, test the instructions on a copy of a practice drawing.

The objective is to make AI a useful learning assistant without giving it control over your live project.

Week 3: Experiment with drawing checks

Choose one repetitive task, such as checking naming conventions, identifying missing annotations, or reviewing drawing layers.

First, define the rule clearly.

Then investigate whether a built-in Revit or AutoCAD feature can perform the check. If not, ask AI to help design an automation workflow.

Start with a report that identifies potential problems. Do not begin with a tool that automatically modifies the original drawing.

Week 4: Explore AI-assisted drawing creation

Investigate the AI capabilities available in your current Autodesk software.

If a suitable workflow exists, test it on a simple practice project. Otherwise, explore a small automation using Revit's API, Dynamo, or AutoCAD automation tools.

Compare the result with a manually prepared drawing.

Evaluate whether the automation saves time, produces editable engineering objects, preserves project standards, and introduces any errors.

At the end of the month, you should have a basic local version history, a useful AI-assisted workflow, and a better understanding of which tasks can safely be automated.

Final thoughts

Git and AI solve different problems, but together they offer a useful way to modernise an engineering workflow.

Git gives you a history of your project files. AI can help you learn faster, automate repetitive operations, and investigate potential drawing errors.

The most exciting possibility is not simply asking AI to explain how to draw something. It is building a workflow in which AI helps create or inspect actual engineering models, while version control and backups give you a safer way to manage the results.

Start small. Protect the original files. Verify the engineering. Expand the automation only when you understand what it changes.

You do not have to become a software developer to benefit from these tools. Your engineering knowledge is the foundation; Git and AI are additional tools that can help you apply it more efficiently.

The goal is not to replace the engineer. It is to give the engineer better tools.

Building a Reading Culture.

George Wanjohi — Sun, 13 Sep 2026 07:07:08 GMT

This being the internet, the above post is no longer available. Did I not read the terms and conditions?

The research findings - as of when first I penned down this draft - showed, that women are more likely to read a product/gadget manual. Lucky for men, we coexist with women at work, at home, and in many other social settings.

In the technology space, your efficiency is pegged on knowing your problem very well in order to match your solutions with the best tools available. Luckily, there is no deficiency of manuals out there for anything you might be looking for. Your solution is always a question away. If this is not the case, it is always an opportunity to write.

The post above looks at a parallel problem - getting directions. I will look at one thought -

"Most people prefer to learn by doing, not being told what to do."

My grandmother told stories of traveling long distances during her youth - without maps, cars, or marked roads. This scares the hell out of modern-day we. Luckily, during those times, without technology, people really had to depend on, and trust their brains and ask for directions of course.

With the increase in the number of things, and their interactions growing ever more complex, there is a need to read manuals more than ever. Modern professions are riding on the availability of information and the ease with which such information can be easily transferred.

As a budding software architect - I am more aware of the fact that there is a need to read more and write about my approaches. The worry of being wrong should not be a stopper. I am more appreciative of the fact that being wrong is a step in the right direction. It takes time to know the right questions to ask. In the meantime, there is a need to keep on asking.

To quote Confucius, “True wisdom is knowing what you don't know.”

When working on Software, reading manuals could be present as

Understanding a problem domain.
Understanding your toolset.
Appreciating work politics.
Getting feedback from software users, among others.

References:

Photo by Ben White on Unsplash

Happy Coding!

Why the Soviet Computer Failed

George Wanjohi — Sun, 13 Sep 2026 07:05:11 GMT

"In 1986, the Soviet Union had slightly more than 10,000 computers. The Americans had 1.3 million.

At the time of Stalin's death, the Soviet Union was the world's third most proficient computing power. But by the 1960s, the US-Soviet computing gap was already years long. Twenty years later, the gap was undeniable and basically permanent. Why did this happen? The Soviet state believed in science and industrial modernization. Support for research & development and the hard sciences were plentiful. They had the country’s finest minds. Goodness gracious, they launched Sputnik! They landed on Venus! How did it come to this?"

.... check the link below to see the full story on YouTube

Notice!

This content has been provided with kind by Asianometry on youtube. Watch the full video Why Soviet Computers Failed. Next time you want to extract subtitles or closed captions from YouTube videos, consider using youtube-dl.

"youtube-dl is a command-line program to download videos from YouTube.com and a few more sites. It requires the Python interpreter, version 2.6, 2.7, or 3.2+, and it is not platform-specific. It should work on your Unix box, on Windows, or on macOS. It is released to the public domain, which means you can modify it, redistribute it or use it however you like."

$ youtube-dl --write-sub  --skip-download <youtube-address-goes-here>

What we get is a *.VTT file. See some additional processing done below.

Happy Coding

Cover photo by Md Riduwan Molla on Unsplash

Memory Management at Runtime

George Wanjohi — Sun, 13 Sep 2026 07:03:26 GMT

In computing, memory is a device or system that is used to store information for immediate use in a computer or related computer hardware and digital electronic devices. The term memory is often synonymous with the term primary storage or main memory. An archaic synonym for memory is store. Wikipedia

Classes of Memory

Generally, there are 3 classes of Memory

a. Static memory.

Allocated for the entire lifetime of the program.

b. Automatic memory.

Allocated and freed automatically when functions or methods are executed.

c. Dynamic memory.

Explicitly allocated by the program and freed explicitly—or by an automatic program called the GARBAGE COLLECTOR.

Rules

Top Level variables and static fields belong to Static Memory.
FunctionArgs and local Variables belong to automatic memory.
Memory is explicitly allocated using the NEW operator, belonging to dynamic memory.

The most important element of a programming language runtime


                                                    is memory

Building Software that People ♥️

Execution Stacks and when they Overflow

George Wanjohi — Sun, 13 Sep 2026 07:01:47 GMT

In computer science, a thread of execution is the smallest sequence of programmed instructions that can be managed independently by a scheduler, which is typically a part of the operating system. Normally, when you start your program, the compiler already breaks your program into independent threads which are passed to the runtime independently.

Each execution thread has a stack where local variables and arguments are allocated when a function or a method is invoked. A stack is used because it naturally follows the execution flow of method and function calls. The topmost record contains the data about the currently executing function; below that is the record of the caller of the function, which sits on top of another record of its caller, and so on.

In the example above, we have a function that calculates the factorial of a number n. Since this function is recursive, it does not compute the final result until the end condition n <= 1 is met. All preliminary result sets are held in memory and are contained in a stack that is bound to the currently executing thread.

e.g. the factorial of 5

In memory, our stack contains 5 variables, this is all good. A design-based question arises, how much information can this stack hold? We need to know this threshold to avoid the next error, the Stack Overflow exception.

Trying to calculate the factorial of n = 50000000 results in a Stack Overflow Exception - easily.

Accounting: The great divide.

George Wanjohi — Sun, 13 Sep 2026 06:58:59 GMT

Among other routines, managers are inherently accountable for most of their departmental/business finances. Provided with a budget, and timelines, we need a trail of accounts that shed light on business products, projects, and operational undertakings therein.

Where there are proper and accurate records of transactions, it is easier to drive business processes to profitability.

Devoid of an accurate record of accounts, Risky undertakings are highly frowned upon. More so, quantifying success is nearly impossible. Where success is not measurable, conflicts will always arise. You will find reward schemes bogged with injustices. Political affiliations are rewarded, and dissent—coupled with financial implications, is highly discouraged.

It is records of accounts that inform budget planning, expenditure tracking, etc. Additionally, at a glance, you get a detailed view of records of liabilities, expenses, incomes, the value of assets, etcetera.

There is plenty of software out there that can aid in keeping and extracting helpful information from journals. My personal favorites are -:

GnuCash is personal and small-business financial accounting software, freely licensed under the GNU GPL and available for GNU/Linux, BSD, Solaris, Mac OS X, and Microsoft Windows.

Strings Revisited

George Wanjohi — Sun, 13 Sep 2026 06:48:47 GMT

“Ten, nine, Ignition Sequence Start, six, five, four, three, two, one, zero. We have a light off! Lift off…”,

Strings (of Text) are everywhere. Used widely in writing. To tell stories, share ideas, and much more. In programming, a text string, also known as a string or simply as text, is a group of characters that are used as data.

History Scholars generally agree that the earliest form of writing appeared almost 5,500 years agoin Mesopotamia (present-day Iraq). Strings have been with us for the better part of our recent history—think of it as it pertains to the build-up in the ubiquity of computer hardware and software today.

Textual Data is used in many places. Consider secure communication over the webusing SSL Certificates.

For a browser on the web, the process looks like this:

A browser attempts to connect to a website (i.e., a web server) secured with SSL.
The browser requests that the web server identifies itself.
The web server sends the browser or server a copy of its SSL certificate in the response.
The browser checks to see whether it trusts the SSL certificate. If it does, it signals this to the web server.
The web server then returns a digitally signed acknowledgment to start an SSL-encrypted session.
Encrypted data is shared between the browser and the web server.

Broadly, an SSL certificate would contain information like,

issuer_name
  country= kenya
  Organization = acme
Validity
  Not Before = Fri, 25 Nov 2022 00:00:00 GMT
  ...
    ...

Strings are useful in interfaces where machines take input from humans or where machines need to communicate information to end users. Strings can present as emails, visual queues to information held in computers, navigation elements that project the meaning of underlying data. This data might be other forms of data such as numbers or even sounds, and other visual artifacts such as images and videos.

In systems design, it is imperative that communication projected to the user of a program or other programs be concise and adhere to a set of pre-agreed-upon requirements. It is also important that the language of the problem domain be used to communicate meaning clearly.

“Strings are very useful when communicating information from the program to the userof the program. They are less useful when storing information for the computer to use”.

Since all textual data is a sequence of characters, there's a need to define what makes texts different. Lightly put, there is a difference between an email address and a person's name—both texts with distinct requirements.

Remember when information is transmitted through various media, often, given factors such as entropy: meaning, and context might be lost. Imagine getting a bad reading in court.

Some links:

Wikipedia Python ms-sql matlab java javascript

Stack Traces

How Your Phone and Laptop Can Hear Your Voice but Ignore the Music Playing Beside You

1. Your microphone doesn't simply transmit everything it hears

2. What happened on my Samsung Galaxy A26?

3. Why does the same thing happen on a Windows laptop?

4. Is artificial intelligence involved?

5. Why doesn't noise cancellation remove every background sound?

6. How can you test this yourself?

7. Does this mean I can play music during every meeting without anyone noticing?

Final thoughts

Why Feedback Matters More Than Confidence

A Beginner’s Guide to Git and AI for Electrical Engineers: Protect Your Drawings, Work Smarter, and Design Better

Introduction

Part 1: Understanding Git without learning to code

1. What is Git?

2. Does Git upload your drawings to the internet?

3. What do you need to install?

4. Create your first local repository

5. Save a version whenever you reach an important milestone

A practical workflow

6. What happens when you make a mistake?

7. An important limitation: Revit and AutoCAD files are not ordinary documents

Part 2: Using AI to improve your engineering workflow

8. Start with AI as your engineering assistant

9. Can AI actually create Revit and AutoCAD drawings?

Level 1: AI guides you through the drawing process

Level 2: AI helps automate actual drawing operations

Level 3: AI works directly with your CAD model

10. Can AI find and correct mistakes in electrical drawings?

Part 3: Protecting copyrighted engineering work while using AI

11. Local Git and local AI are two separate things

Option A: Use a cloud-based AI assistant

Option B: Run AI locally on your own computer

Part 4: A practical plan for getting started

Week 1: Learn local version control

Week 2: Use AI to learn your existing tools

Week 3: Experiment with drawing checks

Week 4: Explore AI-assisted drawing creation

Final thoughts

Further reading

Understanding AsTracking vs AsNoTracking in EF Core

Understanding AsTracking vs AsNoTracking in EF Core

Why Tracking Behavior Matters

The Default Behavior

When AsNoTracking() Is the Better Choice

When to Use AsTracking() Explicitly

A Practical Rule of Thumb

Takeaway

Building a Reading Culture.

Why the Soviet Computer Failed

Memory Management at Runtime

Classes of Memory

Rules

Execution Stacks and when they Overflow

Accounting: The great divide.

Strings Revisited

Understanding `AsTracking` vs `AsNoTracking` in EF Core

When `AsNoTracking()` Is the Better Choice

When to Use `AsTracking()` Explicitly