Some Personal Reflections Going into My 38th Birthday
📅 When we begin our New Years Resolutions, most are lucky if we get through the first couple of days. My personal journey has been a different story. As I’m writing this (August, 15, 2025), I can say that I’ve stuck to – and even exceeded my goals.
📱 I deleted a vast amount of my attention depleting social media (LinkedIn doesn’t count, nor does Bluesky for news feeds), and replaced it with reading. As I’m composing this, I’m approaching my 30th book completed.
🏋♂️ I set out to lose weight with the goal of 15-lbs. As of last week, I’ve hit the 50-lbs mark and have gone into weight maintenance mode.
📚 I listen to and read more content than I ever have, completing a lot of online learning; feeling confident about my base of knowledge for life. I’ve written more blog and content posts than I have ever have.
🏆 Has any of this helped me with my career aspects? No, not at all. Has it helped me become a better person? Yes. That is the most important part of it all – doing it for yourself.
🎂 As I approach my 38th birthday in a few weeks, I look back at all I’ve accomplished, and what a rocky road it’s been–yet I realize how much further I still have yet to go.
This was originally posted on my Linkedin.
The Great Commodification of LLMs
The price wars in LLMs have begun. This will lead to margin collapse in the industry, while consumers benefit. Alternatively, one of the only movements in the domestic technology industry holding the United States economy above water is the GenAI boom.
OpenAI has priced ChatGPT-5 competitively with Anthropic’s Claude models. Data center buildouts continue to expand with Microsoft announcing during quarterly earnings that they will spend $120b additional per year (up from $80b or so this previous fiscal year).
This TechCrunch article also states Meta plans $72b spend, and Alphabet with a $88b CapEx spend. Additional buildout is still needed and planned, however, with margin compressions, especially in tech come second looks on whether these data center buildouts will net a long-term return on investment.
We must also consider the localization models (SLMs, etc.) in this equation. Giants like Perplexity and OpenAI will gladly train their models on what the user inputs into the LLMs, thus it has become a privacy concern for many. The more efficient and prevalent open-weight models become, the more the end-user will be comfortable utilizing them on their local GPUs and/or NPUs. Consequently, most of these models run neck and neck as far as performance and returns are concerned. Those customers who pay for multiple models will become comfortable paying for just one or two.
Commodification is the sign of a mature market in the technology space. Consider the early 2000s when a massive amount of fiber optic cables were deployed. The infrastructure companies such as Lucent, went out of business, but the end result was a higher reach of broadband penetration by the mid 2000s. LLMs may reach the same end point, but this by no means contributes to the idea that GenAI is over. It just means that LLMs have almost reached a diminishing return.
Data centers will continue to be built at the pace they are so more powerful types of GenAI down the road can be marketed and productized to consumers, businesses, and academia. GenAI is more than just LLMs. Multi-modal models, agentics, and real-time machine models have a bright future ahead, and are only just getting started.
Finished reading: Zillow Talk by Spencer Rascoff 📚
Finished reading: Feel-Good Productivity by Ali Abdaal 📚
MIT Technology Review Comment: Don't Ban ChatGPT in Education
Like all new technologies in education, the initial response is to ban them. Consider Wikipedia, for example. Over a decade ago, the resource was chastised for the chance that a student may plagiarize an entry. On the contrary, it’s become a well sourced tool for initial research on a subject, with well sourced citations.
Fast forward to today – LLMs are tools to be used in critiquing arguments, the creation of ideas, and a second opinion to inform the writer or reader of salient points that may have been missed. Never truly trust a technology on face value, but proper use cases must be taught (by faculty and parents) or students will fall behind.
My comment was originally posted for LinkedIn.
Finished reading: Building a Second Brain: A Proven Method to Organize Yo… by Tiago Forte 📚
Knowing Your Workflow for Note Taking
🔖 In my quest to become more knowledgeable in topics and subjects that are either relevant to me or my career, I like to use tools such as Obsidian and Google Keep to just down notes and reflections.
📗 In conjunction with my routine of ingesting insightful blogs and journalism, I’ll peruse Reddit (for example) to brainstorm blog ideas or retaining useful facts to whatever project I’m considering. When following the appropriate subreddits, consider the comments as a way to consider others' opinions to challenge your own. Lastly, I will synthesize it for later use.
🧠 Personal Knowledge Management, or PKM, is only becoming more important in the age of GenAI (filtering LLM considerations from original sourcing). I have a lot of work to do in this area, but it takes practice to grasp a workflow that works for you and your needs.
Finished reading: No Rules Rules by Reed Hastings 📚
Making Siri Great Again
If true, this Bloomberg report would be one of those rare instances that Apple would admit defeat – at least for the time being. Partnering with OpenAI or Anthropic for Siri may buy them some time. Then again, it could be analogous to Apple ceding the ad market, which is why they claim to be ‘privacy first’.
In recent weeks, it has been circulated that many firms have been interested in Anthropic such as Meta and Apple itself. If Meta were to be successful, they would gain valuable real estate on the MacOS and iOS platforms, likely ending Google’s multi billion dollar a year contract for search.
This would also confirm that Apple was unable to purchase the company, thus, moving to the next best thing – a partnership that would vastly boost Anthropic’s already sky-high evaluation.
In the past few weeks, OpenAI has not expressed its feelings with Microsoft as positive, so a decoupling from the software giant would result in new revenues from Apple. Nobody ever went broke from diversification, which includes the customer base.
Finished reading: Desi Arnaz by Todd S Purdum 📚
I Have LLMs Critique My Writings and Posts
Continuing on the theme of LLMs enhancing your own work rather than replacing it. I had several platforms sift through and critique my writings on my posts and blogs. Some of the results were interesting, and unexpected to say the least. For this test I prompted Gemini 2.5 and Claude Sonnet 4 the following: “Critique the writings of michaelmartinez.co.”
This is yet another tool in my workflow. I could take or leave the suggestions, but as always, it’s an option. After reading through the “experiment” I want to know what you – the reader would like to see more of from me in future posts and writings.
Gemini 2.5 talked of the following strengths:
Breadth of Knowledge: Martinez exhibits a wide array of interests, seamlessly transitioning between discussions on cybersecurity, the nuances of digital advertising antitrust cases, and historical analyses of xenophobia. This breadth enriches his perspectives and allows for interdisciplinary connections.
Topical Relevance: His chosen subjects are highly relevant to current global discourse, addressing pressing concerns in the tech industry (LLMs, AI, antitrust) and societal issues (xenophobia, foreign policy).
Clear and Concise Communication: The articles are structured well, with clear headings and a direct writing style, making complex topics accessible to a broad audience. He effectively summarizes key points and arguments.
Thought-Provoking Insights: Martinez often delves beyond surface-level observations, offering insightful critiques and raising pertinent questions, such as the effectiveness of breaking up large tech companies and the future impact of AI on advertising. His willingness to express concern about negative rhetoric and connect it to historical patterns of xenophobia adds a moral and ethical dimension to his foreign policy discussions.
Personal and Reflective Tone: Elements like his “Reading Updates” and experiences with Micro.blog add a personal touch, making his content more relatable and authentic. His emphasis on continuous learning and reading widely underscores a commitment to self-improvement and intellectual growth.
Use of External References: He frequently references external sources, such as TechCrunch’s “Cyber Glossary,” Jesen Huang’s views on AI, Mark Gurman’s criticisms of Apple, and Clayton Christensen’s “The Innovator’s Dilemma.” This demonstrates research and supports his arguments with credible insights.
The second part of the output were the “Areas for Potential Enhancement”.
Deeper Dive into Specific Arguments: While he touches on various aspects of each topic, providing more in-depth analysis or expanding on specific arguments with additional evidence or case studies could further strengthen his positions.
Original Research or Unique Perspectives: While he synthesizes information effectively, incorporating more original research, unique data points, or highly distinctive viewpoints could elevate his critiques to a more authoritative level.
Engagement with Counterarguments: Presenting and directly addressing potential counterarguments to his positions could lead to more robust and nuanced discussions.
Claude provided me with almost the same strengths and weaknesses as the Gemini prompt but went further in two ways. First, is a “Content Assessment”.
Technology Commentary: His AI/LLM analysis is generally sound but not groundbreaking. He correctly identifies key limitations but doesn’t offer particularly novel solutions or frameworks for thinking about these issues.
Political Commentary: His piece on American xenophobia attempts to draw historical parallels between treatment of Japanese Americans, Muslims post-9/11, Russians during the Ukraine conflict, and current Latino deportations. While the historical connections are valid, the analysis lacks nuance and depth that such complex topics deserve.
Educational Insights: His observations about AI in education are practical and well-reasoned, particularly his emphasis on teaching critical thinking alongside AI literacy.
Lastly, Claude provided an “Overall Assessment”.
Martinez’s blog represents competent but not exceptional commentary on technology and society. His strength lies in providing accessible, practical perspectives on complex topics, particularly around AI adoption and digital literacy. However, the writing lacks the depth of analysis or original insight that would distinguish it in a crowded field of tech commentary.
The blog serves its apparent purpose as a personal platform for sharing thoughts and maintaining an online presence, but it doesn’t rise to the level of must-read commentary on the topics it covers. For readers seeking practical, moderate perspectives on AI and technology adoption, it offers value. For those looking for cutting-edge analysis or original research, they would likely need to look elsewhere.>The author’s commitment to continuous learning and reading is admirable and likely keeps his commentary reasonably well-informed but translating that input into more distinctive analytical output would strengthen the overall impact of his writing.
I will keep retooling and refining my methods in the future based on this feedback. What do you, a human, think of the assessments from two of the biggest GenAI tools out there?
GenAI is Still Not Replacing You
Back in 2023, when LLMs and GenAI was still in its infancy, I argued that GenAI will be a tool for those in the job market and change the workflow of the way we spend our careers. In the time since I wrote that piece, not much has changed and I still stand behind the rationale.
It is essential to clarify that LLMs, including GPT-4, are not true AI. Despite their impressive capabilities, they lack true understanding, consciousness, and self-awareness. LLMs rely on pattern recognition and statistical processing rather than genuine cognitive reasoning. They do not possess subjective experiences or emotions. They are tools designed to process and generate text based on patterns learned from vast amounts of data. Therefore, LLMs cannot fully replicate the complexities of human intelligence, nor replace the multifaceted skills that humans bring to the workforce.
The core to my argument is the lack of reasoning and thinking. To this day, I do not “like” those terms and believe we should choose better words to describe the tokenization aspects of it all.
Recently, The Economist published a piece entitled, “Why AI hasn’t taken your job”. It takes the argument that AI has changed the nature of some careers such as those in translation (See Duolingo) and learning, but postulates that upskilled careers such as interpretation of language learning has increased. Upskilling and upshifting of productivity are still key to future successes in sectors such as this.
Klarna is also given as an example where a choice is given between GenAI based customer service or a human:
“There will always be a human if you want,” Sebastian Siemiatkowski, its boss, has recently said.
More importantly (nothing is definite with new technologies), given the massive technology layoffs with claims that AI is replacing work by CEO’s, the data tells a different story.
Across the board, American unemployment remains low, at 4.2%. Wage growth is still reasonably strong, which is difficult to square with the notion that AI is causing demand for labour to fall. Trends outside America point in a similar direction. Earnings growth in much of the rich world, including Britain, the euro area and Japan, is strong. In 2024 the employment rate of the OECD club of rich countries, describing the share of working-age people who are actually in a job, hit an all-time high.
The conclusions are the same – GenAI replaces redundancy and not people. As the technology matures, it may change if new unforeseen breakthroughs were to surface, but as of now it’s still best to teach yourself how to use these tools so that you aren’t replaced by an employee who is already familiarized with it.
Finished reading: Nuclear War by Annie Jacobsen 📚
Finished reading: Quantum Supremacy by Michio Kaku 📚
AI and Learning: There's a Major Disconnect Occurring in Education
Just like with all industries and applications, the use of LLMs is literally rewriting all of the best practices with education currently being the largest social sector being disrupted at the moment. Rather than using LLMs to ‘cheat’, educators and public administrators must teach students how to use and coexist with them rather than relying solely on them.
A lot of students have abandoned critical thinking all together and resulted in outsourcing all of their knowledge to these LLMs just to make answers and save time without learning or even conducting reasoning of their own to consider what the prompts are or what they’re telling them. As a result, teachers, administrators, and parents are beyond frustrated and returning to the good old blue books as cited by this Gizmodo article.
Students need to learn how to properly research while understanding proper time management. I won’t say laziness is at work on the part of all parties involved, but the quickest way to get from A to B is often using tools at your disposal – this is true in education or eventually the workplace, however, a lot is lost in this process including how to critically think about what output LLMs are displaying or what it might mean for the overall context of the subject.
Education has always been a lagging indicator of technological trends, and this is no different. LLMs and other types of GenAI are tools, not the end all be all solution in the classroom. Using it as a partner in research yet taking a critical view of what it’s telling you is paramount to making research easier for all, without compromising the time-honored tradition of writing research papers and a child’s knowledge retention.
A full education, as always, should concentrate on a child’s soft-skills – learning how to critically think, put together proper research, how to write for life, and home in on communication skills to make them successes in their careers and lives. Tools can do that, pouring prompts into an unchecked LLM cannot.
Google's Gemini will Contain Ads -- It's Only a Matter of Time
It’s common knowledge that Google is an advertising company first and foremost. All of the products weave together this ad network that a lot of governments and organizations consider to be anti-competitive. If you query a standard Google search, you’re starting with sponsored posts, followed by more advertising – which can be hard to navigate when you need accurate and actionable information fast.
Gemini, Google’s AI and LLM platform is just that. It’s a tool for consumers and businesses alike, which can be marketer’s dream. Recently, it was announced that marketers can use the service to insert more strategic ad placement into YouTube. The use of agentics, that is, using GenAI on your behalf to make decisions, automate decision making based on the user’s parameters, and adaptability – is now becoming commonplace among browsers, LLMs, and other AI tools and infrastructure. Marketing and advertising firms will have to adapt to serve these minimally invasive asks with minimal human intervention. The trick is serving ads through these means.
In the first paragraph, I’ll reiterate that Google is an advertising company, thus this is just one more avenue they must figure out to use its closed ad-network to bring in more revenues. They must walk a tight rope though due to antitrust concerns currently before the US federal government and EU agencies. While the early 2000s was dominated by SEO; GEO (Generative Engine Optimization) is firstly built to be utilized properly on AI platforms.
Anthropic recently announced that Reed Hastings, founder of Netflix has joined their Board of Directors. This may seem strange on the surface but makes perfect sense when we realize that Netflix has a robust advertising platform as the world’s top streamer. Hastings expertise in this arena will fit in perfectly to what Anthropic wants to pursue – becoming its own advertising platform to rival that of Google, which is no surprise given if the courts rule that Google has to divest Chrome; that Anthropic was a party interesting in purchasing it.
Google’s strategy has always been that of the tech industry as a whole; gaining as many users as possible with a new product, then turn on the spicket of advertising. We can only hope that the sheer number of ads inserted won’t damage the Gemini product as it has Google search. We might not need a government to dismantle the firm, the vast competition within the LLM space and all tech wanting a piece of the pie may force them to retool to strike a balance between users and AdSense. Thus, Google wielding its vast ad network to other products might become a coherent strategy – much as Microsoft focused on cloud rather than Windows throughout the 2010’s.
WSJ Study on Career Skill Certificates
Job skill certificates from the likes of Coursera or edX are exactly what you make of them. They are a jumping off point for learning new skills and techniques that can be put forward in the workforce. A recent WSJ article cast doubt on their validity and career advancements that they claim to skill-up for.
While I disagree that they should be the end all be all requirement for an employer to make decisions on a candidate – they can provide must more value that one might think. Intrigue and curiosity are the main drivers why a prospective employee may try to take one of these credentialed courses or professional certifications. It’s meant as a stepping off point in order to build upon that knowledge. It’s not what they’re doing in the course, it’s what comes after.
That is the issue with his study – what’s in demand will always change. What was data analytics and coding yesterday, will be something completely different tomorrow.
Let’s take the example of the role of a Business Analyst, for example. Coursera has a wonderful program called “Microsoft Business Analyst Professional Certificate”. The idea is to take someone from the fundamentals to a working project at the end to prove their knowledge retention and introduce knowledge workers to the skills that the software company has to offer. If you’re new to Power BI, Power Apps, and the Power Platform in general, than this is a great introduction that weaves these concepts into a curriculum.
Like all skills and experience, nobody can be an expert from one pass through of this course, but what it does is lay the foundation to garner an introduction to the company’s products which are widely used in many industries. It’s not limiting to a single platform, but the concepts can be taught to many competitors products (i.e. if you understand how to use Power BI, then Salesforce’s Tableau will be understood just as easily).
Upon competition of the course, Microsoft offers a voucher for half-off the price of the entry level Power Platform certification. While the merits of the study from The Burning Glass Institute and the American Enterprise Insitute need to be delved in further, this conclusion is not a one-size fits all study. Many of us pivot careers every 2 to 5 years in today’s information economy. It’s likely that you need to reskill and retool yourself before then. Skill certificates have been always a great introduction to a new topic and interests, and they will continue to provide that crucial role in the future.