Welcome back, fellow wanderers, to Day 2 of “Unlocking the AI Toolbox – A Skill-Wanderer’s Journey“! It’s insightful how AI explorations intersect with daily work. Recently, a colleague asked if she could share my NoteBookLM Plus account, having heard it’s great for quickly extracting info from documents. She was drowning in reports!

That request highlighted NotebookLM’s value not just for tech enthusiasts, but for anyone needing to learn, research, or make sense of large texts efficiently—perhaps even at higher volumes or needing advanced features. So, for Day 2, we’re diving into what I consider a must-have for all learners and information-workers: Google’s NoteBookLM.

My goal isn’t just listing features. As a Skill-Wanderer, I want to explore how to wield this tool, its strengths, and how it aligns with my AI Compass: augmenting abilities with human oversight. Let’s explore NoteBookLM!

I. Why NoteBookLM is a Game-Changer (And Next Up in My Toolbox)

My colleague’s interest encapsulates NotebookLM’s promise: a personal AI research assistant, expert in your own information. Its defining “source-grounding” means its knowledge is strictly limited to your uploaded documents , becoming an “instant expert” on your materials .

I experienced this firsthand, as I mentioned in Day 1, when it helped me make sense of that massive Orange Pi manual. But it was more than just general sense-making. I was specifically trying to figure out how to install Ubuntu on its eMMC (embedded MultiMediaCard). The seller had told me they only knew how to install it on an SD card, which was less ideal for performance. I’d even bought an SD card based on that, which is now, amusingly, left for nothing!

Frustrated but hopeful, I fed the lengthy manual into NotebookLM and asked directly: “What are the methods to install Ubuntu on this Orange Pi model?” To my delight, NotebookLM pointed me exactly to the section detailing eMMC installation. It was a breeze to follow the instructions once I knew where they were. Without asking NotebookLM that specific question and having it search the document for me, I’m sure I would have missed that capability, relying only on the seller’s limited knowledge and wasting a lot more time. That discovery alone saved me significant setup hassle and showed me the power of having a tool that can deeply query your specific sources.

That experience, now reinforced by my colleague’s interest in the Plus version (perhaps due to its higher usage limits or collaborative features ), is why NoteBookLM is front and center for Day 2. It directly addresses a common, critical challenge: the sheer volume of information we often face and the difficulty of extracting specific knowledge, aiming to be a “thinking partner” . Today, I’ll demonstrate its broader capabilities.

II. Getting My Bearings: Setting Up and Feeding NoteBookLM

For my main exploration this time, I decided to tackle a real beast: the “Workday Adaptive Planning Documentation.” This isn’t your average manual; we’re talking a colossal 2721-page PDF (Workday-Adaptive-Planning-Documentation.pdf) which you can find here: https://doc.workday.com/content/dam/fmdita-outputs/pdfs/adaptive-planning/en-us/Workday-Adaptive-Planning-Documentation.pdf and see the sample below. My specific goal was to quickly get up to speed on how “model sheets” are handled within this ecosystem as it related to my BA (Business Analyst) role.

Uploading even such a large PDF was handled smoothly. NotebookLM supports various formats: Google Docs/Slides, PDFs, web URLs, copied text, and YouTube URLs . It can even suggest web sources via “Discover Sources” . Remember, uploads like Google Docs are “snapshots” ; changes to the original require re-syncing . As my AI Compass states: quality in, quality out. With the Workday document, its comprehensiveness was key.

“Tackling Dense Docs” – Putting NoteBookLM to the Test

With the 2721-page Workday document loaded, I put NotebookLM through its paces.

• Summarization Power – Conquering the Colossus: NotebookLM automatically generates an initial summary . For the massive Workday document, I asked for detailed summaries of sections related to “model sheets.” It quickly provided coherent overviews and key takeaways, making the dense material immediately more digestible. This wasn’t just a list of sentences; it was a genuine distillation of complex information. It also suggests related questions to dive deeper .
• Question-Based Interaction – Pinpointing “Model Sheets”: This is a core strength. You ask natural language questions, and the AI answers only from your documents . For the Workday manual, I queried: “What are the primary differences between cube sheets and modeled sheets?” and “Explain formulas in model sheets based on this documentation.” Critically, NotebookLM provides inline citations , linking answers to exact passages in your source . This is vital for trust and verification, allowing rapid location of relevant sections for your own critical review . Sifting through 2772 pages for these details manually would have taken days; NotebookLM did it in moments.
• Multi-Document Analysis & Visualizing “Model Sheets” with Mind Maps: While my Workday exploration focused on one huge file, NotebookLM can synthesize across multiple sources. But even with a single large document, its visualization tools are powerful. For my “model sheets” query, NotebookLM generated an interactive mind map . This visually connected “model sheets” to concepts like data import, versions, and reporting within the Workday documentation . Being able to see these complex relationships laid out, click on nodes for further summaries, and navigate the information visually made understanding the architecture an absolute breeze. It truly transformed a daunting research task into an efficient and insightful exploration. It can also analyze images in Google Slides/Docs

Transforming Information: NoteBookLM as a Creative Partner

NotebookLM also helps create new things from your sources.

• Generating New Formats: From the Workday document, I asked it to “Create a study guide for the key concepts related to ‘model sheets’.” It produced key terms, definitions, and discussion questions. It also generates FAQs, tables of contents, timelines, and briefing documents . I prompted, “Create an outline for an internal training session on ‘model sheets,'” and got a solid starting point, great for overcoming “blank page syndrome” .
• Diving into Web Sources, YouTube, and the Audio Overview Surprise: One of the areas I was keen to test was NotebookLM’s ability to process web URLs directly. You might remember from my Day 1 post, I mentioned my very latest exploration was digging into something called an “MCP server” (Model Context Protocol server). To understand more, I fed NotebookLM the URL for the https://github.com/github/github-mcp-server repository. NotebookLM ingested the content, allowing me to query it to understand what github-mcp-server was all about. Then, “for fun,” I generated an Audio Overview from this source. As the document you shared described , it created an informative and entertaining podcast-style conversation between two AI voices (male and female) discussing github-mcp-server. The surprise was how human-like they sounded. My wife, hearing it, thought the female AI voice was a familiar (human) podcast host and mistook the male voice for human too! It shows how far this tech has come. NotebookLM can also process public YouTube video URLs , using their transcripts to provide summaries, answer questions, or even generate those audio overviews. This sounds incredibly useful for learning from the vast amount of educational content on YouTube. However, I must admit I haven’t had much opportunity to try the YouTube feature extensively. The reality for me, and likely for many of you, is that a significant portion of my learning material comes from paid e-learning platforms. I’m often immersed in courses on Coursera, Pluralsight, LinkedIn Learning, Udemy, DataCamp, ACloudGuru, and other fantastic (but subscription-based) learning sites. As a result, because NotebookLM needs direct access to the content URL, it’s currently unable to process materials that sit behind a login wall. This is a practical limitation for those of us who rely heavily on these structured, paid courses. If any readers have found clever workarounds or know of ways to bridge this gap with NotebookLM (while respecting content rights, of course!), I would be genuinely thrilled to hear about it and would gladly update this post with your insights!
• Multilingual Outputs: A valuable feature for those working across languages is the output language selector. You can choose your preferred language for generated text outputs like study guides or chat responses, making it easier to share work internationally .

V. NoteBookLM Through the Skill-Wanderer’s Compass: Reflections

Using NoteBookLM extensively brought several of my AI Compass principles into sharp focus:

• Augmenting Abilities: NotebookLM handled sifting and summarizing, freeing me for analysis and critical thinking.
• Human Oversight & Verification: Citations are paramount. Google warns it can be inaccurate , so always verify.
• Quality & Purpose: Output quality reflected input quality and focus.
• AI Literacy in Action: Effective prompting is key .
• An “AI General” in my “Specialized Army”? Yes, a specialized intelligence officer for my document “battlefields.”
• Data Privacy: Google states Workspace content isn’t used for general model training or reviewed without permission . Personal accounts reportedly also have data privacy .

Key Takeaways & What’s in My NoteBookLM Toolkit Now

1. Information Retrieval Perfected: A game-changer for large texts (like a 2772-page manual!).
2. Summarization Superpower: Distills dense documents effectively.
3. Content Creation Catalyst: Great for brainstorming and outlining.
4. Learning Accelerator: Study guides, Q&A, mind maps, and audio overviews enhance learning.
5. Source Grounding is Key: Answers based only on your sources (with citations) builds trust and avoids “hallucinations” .

Limitations (and the document confirmed):

• Text-primary , somewhat image analyze .
• Accuracy isn’t perfect; critical verification needed . Can struggle with complex reasoning or specific formats .
• Uploads are “snapshots”; refresh updated documents .

Despite these, NotebookLM is a prominent tool in my AI Toolbox.

What are your experiences with NoteBookLM or similar tools? Share in the comments! Let’s learn together.

Welcome! I’m really excited to finally kick off this new blog series, something I’m calling “Unlocking the AI Toolbox – A Skill-Wanderer’s Journey“ Thanks for joining me here on Day 1.

As I promised when I temporarily paused the “Chronicles of a Home Data Center“ series a little while back, my focus for now is shifting to delve into the world of Artificial Intelligence first. It feels like the right time, and honestly, it’s where my curiosity has been pulling me strongly lately! This AI exploration feels like a natural next step in my Skill-Wanderer journey.

As the name of this blog suggests and as a Skill-Wanderer, I’m constantly finding myself drawn to new areas, picking up different skills, and figuring out how things connect – maybe you feel the same way? Lately, my wandering has led me deep into this AI landscape. It feels like AI tools are popping up everywhere, and it’s both exciting and a bit overwhelming.

I realized pretty quickly that before I could really start making sense of specific tools like GitHub Copilot and what they can do, I needed to get my own mindset right. It felt like needing to find my bearings before setting off into new territory. So, that’s what I want to share with you on Day 1: first, a recap of my Skill-Wanderer’s Compass for AI based on my previous reflections, and second, what I’ve actually been experimenting with lately. And, related to sharing knowledge, I’ll give you a quick update on a personal learning platform project I’ve just gotten up and running.

Calibrating the Compass

As I explored in much more detail in my previous post, “Before We Continue ‘Chronicles of a Home Data Center’: Let’s Talk AI Skills“, setting my “Skill-Wanderer’s Compass” for AI involves navigating some critical ideas. It starts with understanding that AI, powerful as it is, primarily augments our abilities and absolutely requires human oversight, context, and verification – it’s not autonomous, and we can’t blindly follow its output without understanding the bigger picture (as my coworker’s WordPress story illustrated).

My compass also points towards prioritizing quality and purpose in how we use AI, avoiding the trap of generating hollow, valueless content and remembering that meaningful results come from human-AI partnership, not just automation (those terrible AI sales calls and my bank support experience were stark reminders!).

Furthermore, I firmly believe AI doesn’t make fundamental skills obsolete but significantly raises the bar, demanding both strong core knowledge and AI proficiency for continued productivity and relevance – lifelong learning is key.

Finally, acknowledging the sheer unpredictability of AI’s future path underscores the vital importance of cultivating AI literacy now, so we can adapt and hopefully shape its evolution responsibly.

My personal hunch is that this literacy will increasingly involve learning how to effectively lead and orchestrate AI – essentially, I believe everyone will eventually become a general, commanding their own specialized army of AI tools to achieve their goals in the future.With these core principles forming my compass, I feel better equipped to start the practical exploration.

Putting the Compass to Use: Early AI Experiments

But theory needs practice. So, where have my wanderings taken me so far in actually using these AI tools? My background is primarily as a developer, but I often wear BA, PM, and test automation hats, so my experiments tend to reflect that blend, mostly focusing on software development and related tasks, but sometimes wandering further. Here’s a snapshot of my initial forays:

• Tackling Dense Docs with NoteBookLM: One of my first really practical uses was feeding the massive, hundreds-of-pages user guide for my Orange Pi into NoteBookLM. Being able to ask specific questions and get relevant info pulled out instantly, instead of scrolling endlessly, was a game-changer for getting that hardware set up.
• “Vibe Mockups” (Getting Ideas Visual): I’ve been playing with what I call “Vibe Mockups” – trying to go from a rough idea in my head to a visual quickly. Tools like Loveable.dev, sometimes prompted with help from GitHub Copilot, have been interesting for generating initial UI/UX ideas almost intuitively.
• “Vibe Prototyping” (Quick Code Scaffolding): Taking it a step further, I’ve experimented with “Vibe Prototyping.” Using tools such as Fine.dev, again often paired with GitHub Copilot, I’ve tried generating simple functional code snippets or scaffolding basic app structures from high-level descriptions. It’s amazing how fast you can get something tangible, even if it needs heavy refinement. This feels very relevant for my dev/BA side.
• Generating Images: Stepping outside the direct development workflow a bit, I’ve experimented with image generation using Gemini, ChatGPT, and Claude. Mostly for fun or creating visuals for blog posts like this one, but it’s another facet of the current AI landscape.
• “Vibe Install & Maintenance” for Kubernetes: Connecting back to my home lab, I’ve started using GitHub Copilot for what I think of as “Vibe Install” and “Vibe Maintenance” on my k8s cluster. Instead of digging through kubectl cheatsheets or Helm docs, I’ll ask Copilot to generate the command for a specific task or help troubleshoot a configuration issue. It doesn’t always get it right, but it often gets me closer, faster.
• “Vibe Documentation” (Getting Thoughts Down): I’ve started experimenting with drafting documentation, like Readmes or explanations of code sections, using a combination of Gemini (for initial structure or prose) and GitHub Copilot (for code-specific details or comments). It helps overcome the ‘blank page’ problem when documenting my work.
• “Vibe Diagram” (Visualizing Concepts): More recently, I’ve been trying to generate diagrams – like flowcharts or simple architecture sketches – using text prompts with tools like Claude, and exploring if GitHub Copilot can assist in generating code or markup (like Mermaid.js) for diagrams directly in my editor.
• “Vibe Automation Test” (Generating Test Cases): Given my background includes test automation, I’ve naturally explored using GitHub Copilot to help generate boilerplate code for test scripts (using frameworks like Selenium or Playwright) or even suggest potential test cases based on existing application code or requirements. It’s proven useful for speeding up the initial setup phase of writing automated tests.
• “Vibe CI/CD Setup” (Pipeline Configuration): Setting up Continuous Integration/Continuous Deployment (CI/CD) pipelines often involves wrestling with YAML syntax or complex scripting. I’ve experimented with using GitHub Copilot to generate configurations for platforms like GitHub Actions or Jenkins, asking it to create build, test, or deployment steps based on my descriptions. It often provides a solid starting point that I then need to tailor and refine.

You might notice GitHub Copilot pops up quite a bit in these experiments. While it’s known primarily as a code completion tool, as a developer, I’m actively exploring how I can stretch its capabilities and use it more like a general-purpose AI assistant across various tasks in my workflow – from infrastructure and testing to documentation and prototyping.

My very latest exploration is digging into something called an “MCP server” (Model Context Protocol server). The potential, as I understand it, is to enhance tools like GitHub Copilot, possibly by giving it more local context or allowing more control over the models used. I’m still very much in the learning phase here, figuring out what it is and if it’s feasible for my setup.

These are just my initial forays, scratching the surface of integrating these AI tools into my workflow across development, analysis, documentation, testing, deployment, and even system administration tasks. Each experiment teaches me more about the capabilities and limitations.

My Open Learning Project – The Moodle Platform

True to the Skill-Wanderer spirit, I believe that sharing the journey is as important as the journey itself. That led me to a recent project milestone: I’ve successfully set up my own personal instance of Moodle LMS!

If you haven’t used it, Moodle is a free, open-source Learning Management System – basically, a platform for hosting online courses. My reason for setting this up is actually quite mission-driven. I aim to use it as a platform to teach what I’ve learned along my own journey. There are two core motivations driving this: firstly, I strongly believe that the act of teaching is one of the best ways for me to deepen my own knowledge and solidify my understanding (‘learning by teaching’). Secondly, and just as importantly, I want to give back to the wider community. My goal is to make the knowledge I share as accessible as possible to everyone.

Therefore, my firm intention is for all the course content I eventually create and host here to be completely free to access. Think of it less as my ‘private lab’ and more as a future ‘open classroom’ where I can share what I figure out.

I’m happy to report the basic platform is up and running! And for those who followed my “Chronicles of a Home Data Center“ series, you might remember my goal of leveraging free-tier and self-hosted solutions. True to that spirit, this Moodle instance is actually running on my home Kubernetes (k8s) cluster, built largely on resources I already had or could access freely. My philosophy here is simple: keep the operational costs as close to zero as possible. This isn’t just about the technical challenge; it directly supports the mission. By minimizing costs, I can genuinely commit to making the learning content accessible to everyone, without potential financial barriers down the line.

While the courses themselves are still just ideas swirling in my head, you can check out the live platform (though it’s pretty empty right now!) at: Skill-Wanderer Dojo

Now, I know I might have mentioned plans for specific AI courses here in previous posts “Before We Continue ‘Chronicles of a Home Data Center’: Let’s Talk AI Skills“. However, planning course content in the AI space right now feels particularly challenging. The tide of AI is changing so incredibly fast that any course detailing specific tools or step-by-step processes runs a serious risk of being outdated the moment it’s published. Given my goal is to provide lasting value and accessibility, this rapid pace has given me pause. As a result, I’m putting some serious thought into what the first course should actually be. Maybe focusing on more durable foundational concepts, adaptable workflows, prompt engineering principles, or even the meta-skill of how to learn and evaluate AI tools might be more beneficial long-term than a deep dive into a tool that could change dramatically next month.

So, figuring out the best starting point for sharing this knowledge effectively is the next step in this particular side quest, and it’s proving to be an interesting challenge in itself!

Where I’m Heading Next on This Journey

With my compass roughly calibrated, my early experiments logged, and my open learning platform taking shape, where am I heading next in this series?

Starting from Day 2, I plan to begin unpacking the AI Toolbox itself in more detail, sharing what I find as I go. I want to explore beyond just using AI for basic code generation. I’m curious about how tools like GitHub Copilot (and maybe others I discover) can help with practical, everyday tasks – things relevant whether you code, manage projects, or analyze business needs.

Specifically, I want to investigate things like:

• Using AI for terminal commands (because remembering arcane flags is not my favorite thing).
• Seeing how it helps with prototyping ideas quickly.
• Exploring its use in drafting documentation.
• Testing its suggestions for debugging.
• And whatever else I stumble upon!

I’ll be sharing my experiences, successes, and probably some frustrations as I explore these capabilities step-by-step, always trying to keep that Skill-Wanderer’s Compass handy.

Conclusion

So, Day 1 of my journey into “Unlocking the AI Toolbox” is complete! For me, it really had to start with trying to calibrate that Skill-Wanderer’s Compass – getting my head straight about how I want to approach these powerful new tools based on my previous reflections, and then diving into actual experiments.

My Moodle project, running lean on my home k8s cluster, reflects a core part of this journey for me – the desire to learn deeply and share openly and accessibly. The real adventure lies ahead as I start opening that AI toolbox, sharing details about these experiments, and discovering how these tools might enhance the way I (and maybe you) work.

What are your thoughts on developing an AI mindset – what’s on your compass? What AI experiments have you tried recently? I’d genuinely love to hear about your experiences in the comments below! Let’s share the journey.

Hey everyone,

If you’ve been following along with my ‘Chronicles of a Home Data Center‘ series – charting the journey of building the very infrastructure hosting this blog – you might be wondering where the next technical deep-dive post is. Well, I’ve decided to hit the pause button on the ‘Chronicles of a Home Data Center‘ series, just for a little while.

This wasn’t an easy decision. I’m incredibly excited about self-hosting, Kubernetes, and sharing that journey through the ‘Chronicles of a Home Data Center‘. However, as I went through the process of setting everything up – configuring the cluster, tackling networking, deploying persistent storage, and getting this WordPress site running smoothly – I had a crucial realization.

My Secret Weapon: AI Assistants

The truth is, I didn’t do it alone. Far from it. Throughout the setup, troubleshooting, and optimization phases documented (or soon-to-be documented!) in the ‘Chronicles of a Home Data Center‘, I relied heavily on my trusty AI companions – tools like Google’s Gemini, Anthropic’s Claude, and others.

• Stuck on a cryptic kubectl error? AI helped decipher it.
• Needed a baseline YAML configuration for a service? AI provided a starting point.
• Trying to understand a complex networking concept within k8s? AI explained it in different ways until it clicked.
• Debugging why a pod wasn’t starting? AI offered potential causes and solutions.

These tools were instrumental. They accelerated the process, helped me overcome hurdles I would have spent hours (or days!) wrestling with, and ultimately enabled the success of the project featured in the ‘Chronicles of a Home Data Center‘ so far.

And, if you don’t already notice, even crafting this very blog post explaining the pause involved collaboration with my AI friend, Gemini. While the core idea, the desired style, and the final check on all content are firmly mine, Gemini helped handle some of the nitty-gritty details of phrasing and structuring the text – a perfect illustration of how integrated these tools can become, even beyond purely technical tasks.

The Dilemma: Setting You Up for Success

And that’s where the pause comes in. It struck me that continuing to post detailed technical walkthroughs for the ‘Chronicles of a Home Data Center‘ without acknowledging the significant role AI played in my process, and more importantly, without ensuring you feel comfortable leveraging these same tools, would be a disservice.

It would be like showing you how to assemble complex furniture but neglecting to mention I used power tools while you only have a manual screwdriver. The end result might look achievable, but the process would be vastly different and potentially frustrating if you tried to replicate it directly without the same assistance or the skills to use it effectively.

My goal with the ‘Chronicles of a Home Data Center‘ isn’t just to show what I built, but to empower you to build similar things. If a core part of my process involves effectively interacting with AI, then simply showing the technical steps isn’t enough. It feels incomplete and potentially sets you up for unnecessary hurdles. Addressing the AI skills first feels crucial for genuine empowerment.

My personal dilemma reflects a larger context we’re all navigating in this rapidly evolving technological landscape. To effectively build the AI skills we actually need, it helps to first grapple with the reality of AI beyond the headlines and the hype. So, before we discuss how to build AI literacy later on, I want to generally share some stories and thoughts about AI based on my experiences. My hope is that these perspectives can help us all develop a more grounded and realistic mindset for collaborating with these powerful tools as we move into the future.

AI Hype vs. Reality: My Thoughts on Collaboration and Quality

1. The Illusion of Autonomy: Why Human Oversight is Non-Negotiable

There’s a lot of talk these days about AI replacing humans. While AI is transforming industries, my experience suggests it’s less about replacement and more about augmentation – AI as an incredibly powerful tool that still requires human guidance and understanding. Let me share a brief story to illustrate. A coworker of mine, a brilliant marketing and PR specialist but without deep technical web knowledge, needed to manage and update a WordPress website. She turned to an AI assistant for instructions on making a specific change. She followed the AI’s advice meticulously, step-by-step.

The result? She successfully achieved the exact outcome she described to the AI. The AI fulfilled the request based precisely on the prompt. However, because she lacked the broader technical context of how WordPress themes, plugins, and core files interact, she didn’t foresee (and the AI didn’t warn her about, as it wasn’t asked to check for conflicts) that the change would clash with another part of the site. So, while the intended task was completed, another feature unexpectedly broke. This isn’t really a failure of the AI – it did what it was explicitly asked.

It’s a stark reminder that human understanding and oversight remain crucial. AI, in its current form, often lacks the holistic view, the intuition born from experience, and the ability to anticipate unintended consequences outside its specific instructions unless prompted very carefully (which itself requires knowledge!). We need to be the architects and supervisors, verifying the plans and checking the work, not just blindly following blueprints generated on request. Even highly intelligent professionals in other fields need that foundational understanding when applying AI to technical domains.

2. Quantity vs. Quality: The Trap of Hollow AI Content

This ties into another trend I see: the rise of courses advertising fully automated AI solutions, especially in marketing – promising systems that post to social media without any human input. While the course creators might profit, I seriously doubt the long-term value for the students or their audiences. Why? Because it’s incredibly easy nowadays to generate purely AI-written content, but it’s often incredibly hollow. Frankly, I find interacting directly with an AI much more useful and engaging than reading floods of text generated by one without purpose. Some of my friends have already started complaining about how much of this generic, soulless AI content is overflowing the internet.

My friends aren’t alone; I’ve certainly had my own jarring experiences. For instance, I’ve started receiving AI-powered cold sales calls. If getting an unsolicited call from a stranger wasn’t already off-putting enough, hearing a cold, synthetic AI voice trying to sell me something is genuinely freaky. I hang up immediately whenever I detect that unmistakable AI sound.

Even worse was when I called my bank about a serious system problem needing urgent attention. Instead of a human, I got an AI support agent. Her voice was choppy, clipping words in each sentence, and she just kept asking me again and again to restate my problem, clearly unable to grasp the context or complexity (“her content awareness seem to be problematic” indeed!). My mood shifted rapidly from ‘I need help logging a critical issue’ to ‘Miss AI, please just tell me how to close my account with this bank!’ And perhaps luckily for the bank, though frustratingly for me at the time, she couldn’t even guide me on how to do that properly.

These kinds of interactions exemplify that hollow, unhelpful side of AI automation when implemented poorly or without adequate human backup or understanding. This blog post itself serves as a counterpoint. Yes, Gemini helped write it. But look at the process we’ve gone through (even in our interaction here!): it required significant human direction – me telling it what to write, how to phrase things, defining the core message, providing the stories, requesting specific word changes – to create something that hopefully offers genuine value and reflects my perspective, rather than just being AI-generated “trash” content. Meaningful output requires partnership.

3. Skills Evolve, They Don’t Disappear: AI Raises the Bar

This brings me to a third point regarding AI replacing human skills, particularly the idea that senior technical roles are becoming ‘obsolete’. The word ‘obsolete’ implies our skills become useless, which I find fundamentally incorrect. None of my core technical skills feel useless – not the understanding of how to write a loop, design a database, apply algorithms, architect a full solution like this blog, or any other fundamentals. These remain the essential building blocks.

I’ve trained countless interns, freshers, and juniors. Giving them tools like GitHub Copilot can speed things up, but when the AI fails or introduces bugs (relating back to my coworker’s story), they’re often lost without solid foundational knowledge. It’s why I sometimes implement temporary ‘AI bans’ (months for interns, weeks for juniors) to ensure they grasp the concepts before using AI assistants.

However, the other side of the coin is crucial: failing to learn and leverage AI does impact productivity. To keep up with today’s technological progress, embracing AI and committing to lifelong learning is essential. An experienced senior developer who doesn’t learn to use AI effectively will likely see their productivity lag, and in today’s environment, companies notice this.

I saw this starkly when a junior struggled with a bug for a day; using GitHub Copilot and its agent/chat mode, I diagnosed and generated the fix in about 5 minutes (plus 10 minutes for deployment). The difference, enabled by combining experience with AI, was immense. So, AI isn’t making skills obsolete; it’s raising the bar. New tech means entry-level roles require broader skills and understanding plus AI proficiency. For everyone, staying relevant means mastering fundamentals and mastering the tools that amplify them.

4. The Unpredictable Horizon: Embracing Change Through Literacy

Finally, it’s crucial to acknowledge the sheer unpredictability of where AI is headed. It reminds me somewhat of the early days of nuclear research. At the outset, no one could fully grasp the dual potential – that the same fundamental discoveries would lead to the terrifying power of the nuclear bomb, but also to nuclear energy, a significant power source for humanity.

AI feels similar. It’s a powerful, rapidly evolving technology with two sides of a coin, capable of bringing both the ‘ugly’ and the ‘good’. We can speculate, but we genuinely don’t know its ultimate trajectory. Perhaps my opinions and observations shared here today will be completely deprecated or seem naive in a year or two – the pace of change is that fast.

However, one thing feels certain: AI will fundamentally change how we work, learn, and live. We can’t predict exactly how, but we know transformation is coming. And that very unpredictability is perhaps the strongest argument for focusing on AI literacy right now. Being literate doesn’t mean predicting the future, but it equips us to understand, adapt, and hopefully shape that future responsibly as it unfolds, navigating both the challenges and opportunities AI presents.

Shifting Gears: Focusing on AI Literacy (Temporarily!)

So, based on my own experience and these broader observations, for the next little while, I’m going to shift focus. Before we dive into Docker and other application deployments within our home data center chronicle, I want to dedicate some posts to AI literacy.

For those of you interested in learning more about AI literacy, please know that I’m actively thinking about the best way to achieve this and deliver the content effectively. I have some initial ideas brewing. For example (and as a little teaser!), one avenue I’m seriously considering – tying directly back into the ‘Chronicles of a Home Data Center’ theme – is setting up and hosting a dedicated Moodle LMS (Learning Management System) site right here on my Kubernetes cluster. This could potentially serve as a free, non-profit platform for interactive AI literacy learning. It’s just one idea at this stage, and I’ll share more concrete plans on how we’ll tackle the AI literacy content with you all soon.

I believe building this foundation will make the rest of the ‘Chronicles of a Home Data Center‘ journey (and many other tech projects you undertake) much smoother and more successful for everyone.

What Do You Think?

This is a bit of a detour for the ‘Chronicles of a Home Data Center’, but I genuinely think it’s the right move. I’d love to hear your thoughts!

• Do you use AI tools for your technical projects?
• What are your biggest challenges or questions when using AI for coding, configuration, or troubleshooting?
• What specific AI skills would you find most helpful?
• Have you encountered situations like my coworker’s story where AI assistance led to unexpected issues?
• What’s your take on the quality of AI-generated content you see online?
• How do you see AI impacting technical skills and career progression in your field?

Let me know in the comments below! Your feedback will help shape this new mini-series before we resume our main chronicle.

Thanks for your understanding. Rest assured, the ‘Chronicles of a Home Data Center’ series isn’t abandoned! It’s just waiting patiently while we sharpen our AI tools together.

Stay tuned!