Change is in the Air
Often times I start out these daily tech journal entries with a topic that evolves during the course of the day, and I end up renaming it to a better topic. Today, I simply name it “change”. Maybe it’s a reminder to me to change the topic before publishing, but also it seems like an apt simple topic.
Change is in the air, from the new old president that just took office, to the scuttlebutt about the DeepSeek r1 (v3) models, reducing the uniqueness of OpenAI’s offerings and dumping much of it into free and open source. Nvidia is releasing their DGX Mini Petaflops computer that shows off Jenson’s Law: It takes 100 months to get equal AI performance for 1/25th of the cost. Also, Google just made scraping their search results much more difficult and costly by requiring the execution of JavaScript.
The Data Training Debate
Elon Musk is echoing Ilya Sutskever’s assertion that we’ve run out of data to train the models on, but that’s disingenuous as hell given most of the data on the Web is on the invisible web, requiring the execution of JavaScript, which almost nobody’s large-scale crawlers but Google’s does, giving Google a huge advantage on access to new data for training models. And that’s not even taking into account that they’re not supposed to be letting their models train on copyrighted data in the first place. The words “we’ve achieved peak data and there will be no more” is such a weak position. I don’t think Google feels that way.
Operation Stargate and Infrastructure Investments
Ahh, Google… who was left out of the just announced Operation Stargate, a 500 billion dollar project announced by Trump, Oracle’s Larry Ellison, Softbank’s Masayoshi Son and OpenAI’s Sam Altman. An AI infrastructure company, they call it. Google already has their AI infrastructure, and Elon Musk just built out his. So those who didn’t yet make the big datacenter investments band together and jump on this bandwagon. Sure, we’re in an AI hype cycle. But it’s not a hype cycle that’s going to piddle out. It’s more akin to the hype cycle that was near the beginning of computers themselves.
OpenAI needs new hardware. Currently Microsoft’s Azure cloud is the exclusive infrastructure host to ChatGPT and all OpenAI’s products. But Microsoft hired Mustafa Suleyman and absorbed most of Inflection AI’s talent with a soft-acquisition. And so Sam Altman goes to the soft bank… Uh, I mean Softbank for the money and to another oracle of datacenter hardware… uh, I mean Oracle. And together you have hardware, software and money to open a star-gate… uh, I mean Stargate. It’s all so very poetic and transparent. Those big ol’ hardware datacenter moats of Amazon, Google and Microsoft (and most recently, Musk) are gradually evaporating away, courtesy of Trump. Fascinating!
The Human Element in AI Innovation
Just because the tools are becoming amazing doesn’t infuse us with the ability to use them in brilliant new ways. Quite the opposite, it shines the light on the fact that we have a follow-the-leader mentality and a general lack of creativity. We are going to do the same old things we always did, just now AI-assisted. Stamping templated things out is a lot easier, because that’s exactly the kind of continuation of old patterns AIs do well. But go off the beaten track and try to do something novel and creative, and you suddenly go to war with your AI coding assistant who tries to herd you back to how it’s flocking done.
The Value of Human Context
If there’s competitive advantage to be had off the beaten track, help isn’t coming from an AI. It’s coming from the same novel, unexpected place deep inside yourself as this kind of stuff has always come from. You are scratching an itch that maybe others feel, but only you have taken the initiative to articulate and reach. It’s all about the very context that the AIs are without. They just wouldn’t be able to imagine doing something in a creative novel new way if someone weren’t already out there training them on how to think exactly that way, and apply that sort of thinking to exactly your situation — at least in the current state of affairs.
So, the humans add the unique, novel new context. The very things the AIs lack is the thing that escalates in value. Most humans will by virtue of human nature follow the more traveled path, take the easy rode, and enjoy the AI templated road, making them feel smart and productive stamping out new instances of the same old apps. Very few humans, the outliers, will realize how much opportunity this flock mentality is opening up for new thinking. But if you bring that novel new context to the picture, add the new thinking and sensible ways of tackling problems, new value is created.
The Challenge of Agentic Frameworks
All the talk is about agents and agentic frameworks. Agentic implies agency, and that implies you have the LLMs doing stuff on their own in a sort of self-prompting way. But there’s so much here that jumps over what I think is the essential point: doing one thing well! Doing one thing well is the precursor to everything else, because if you can’t at least do one thing well according to as mostly a rote procedure as possible, how are you going to have an agentic swarm of things doing lots of things well simultaneously and unattended? I know the YouTube videos tell stories like that’s already yesterday’s news, but I don’t buy it. It’s a formula for exploding complexity and unintended consequences.
The Path Forward
So, how do you rein it in? How do you do one thing well? Well, that’s the workflow pipeline stuff I’ve been going on about. I have to really show it soon. Without producing videos on the finished result, it’s just meaningless blog posts to everyone but the LLMs, mostly indecipherable. Making it decipherable is part of the point. That’s exactly the vector embedding and similarity search capability that’s fueling so much of the AI revolution. So I’m making the fodder by documenting the process of making a system, which in itself should be able to make sense of the fodder that was produced in making the system. Very meta.
The Race for Intelligence
I have to get to that part of the project sooner rather than later. The world is in this new race, figuring out how to use all this intelligence. There’s one game being played to make more big scalable blobs of this intelligence and bringing them to bear on classes of problems like protein folding and miracle materials. That’s above my pay-grade. But there’s this other race as all these various models trickle down on us like DeepSeek r1 did recently, and that race is a sheer creativity-to-productivity context.
The Bottleneck of Ideas
Distillation of ideas is the bottleneck for me. Gathering up, sorting, organizing and clarifying my ideas is the bottleneck. And it’s so related to the field of SEO that I’m in and the needs of my clients. It’s the same recurring story. In the case of clients for SEO, it often goes like:
The Disintermediation Challenge
We have something to sell, and we would like for people to buy it from us first — usually, online. And often in a battle against the original manufacturers who are now realizing they can sell directly just as easily, especially with services like Shopify, and cut out the middleman. The term is disintermediation, and it has been in full-swing for awhile. It hasn’t been more popular or discussed because of the existence of the opposite, mega-intermediators like Amazon and Temu. It’s niche sellers who are not on the Amazon or Temu platforms who are at the mercy of the Shopify infrastructure providers of the world who are in trouble.
That’s today’s landscape. And I have to figure out how to make the situation of those types of companies in less trouble.
Survival Between Giants
Is it possible to survive squeezed between Amazon and the manufacturers trying to sell direct. Quite how dire is that situation? Is the writing on the wall for these types of companies? If no, then why? Added value of some sort? Intangibles that can’t just be fulfilled through a hard-goods distribution infrastructure? Experiences? Emotional stuff? Yeah, for sure. Added value.
But what value? Well, it certainly somehow starts with an organization job far beyond anything Amazon itself would be able to carry out within a niche. It will have to be something that requires the domain expertise of someone in that space to help with the organization.
The Customer Experience Factor
Why do I feel this way? Well, what do people want when they’re looking to buy something? Confidence? Expertise? Experiences of others who have been in similar situations? The whole reviews and customer experience thing. Amazon does this well. Google rewards this well. This is the aggregate reviews and star rating system that shows up in SERPs. Oh, there’s also the Google Merchant Center ruling the landscape right now. How deeply do I go into that here? The explicit and often proprietary data-feeds versus the structured data embedded into the non-proprietary websites? What are those formats again?
Perplexity: Structured data for product reviews and availability can be implemented in multiple formats, not just JSON-LD. While JSON-LD is the recommended and most commonly used format, there are other options available[1][4][7]:
-
JSON-LD (JavaScript Object Notation for Linked Data): This is Google’s recommended format due to its ease of implementation and maintenance[4]. It’s flexible and can handle complex, nested structured data[3].
-
Microdata: An older format that embeds structured data directly within HTML elements[1][4].
-
RDFa (Resource Description Framework in Attributes): Another format that can be used to embed structured data in HTML[4].
While all three formats are accepted for schema markup, JSON-LD has several advantages:
- It’s easier to implement and maintain, especially for complex data structures[3].
- It’s not restricted by the content and structure of the HTML, offering greater flexibility[3].
- It’s endorsed by Google and other major search engines for its readability by both machines and humans[3].
However, JSON-LD does have some drawbacks:
- It can have a steeper learning curve, especially for those unfamiliar with JSON or linked data concepts[3].
- Implementing JSON-LD may require more technical expertise compared to Microdata or RDFa[3].
Regardless of the format chosen, the structured data helps search engines better understand and display the content of a website, potentially improving visibility in search results[1][7].
Citations:
[1] https://perfist.com/what-is-structured-data-structured-data-types-and-benefits/
[2] https://developers.google.com/search/docs/appearance/structured-data/review-snippet
[3] https://www.schemaapp.com/schema-markup/what-is-the-recommended-format-for-schema-markup/
[4] https://www.collaborada.com/blog/structured-data
[5] https://ralfvanveen.com/en/technical-seo/my-introduction-to-review-structured-data/
[6] https://webmasters.stackexchange.com/questions/116459/do-i-need-json-ld-on-all-pages-or-just-the-home-page
[7] https://www.seoworks.co.uk/structured-data-types/
[8] https://www.blog.nimitkapoor.com/whats-new/pros-and-cons-structured-data/
HTML vs JavaScript Implementation
And those formats should be in the actual view-source HTML of the page and not added with JavaScript. I mean, they technically can be added with JavaScript, but then Google or whatever other crawler/bot/search engine needs to be able to execute the JavaScript which is a lot more resources. It makes it more expensive to do quick data lookups, which should be trivial.
Technical Solutions and Tradeoffs
And yet doing this sort of stuff in JavaScript is the main lever to pull for fast-action on most sites, given the way they’re published today. There’s technical solutions for this, like pre-rendering the JavaScript to HTML for the bots (and only bots) so that they don’t have to render it themselves. But this often takes a lot of setup, configuration, and isn’t the cheapest solution in the world. Baking it into the publishing system itself is often best, but most Shopify customers (for example) are on the standard solution. What’s that language they use for it again?
Perplexity: ### Shopify Headless vs. Non-Headless: Overview and Terminology
Headless Shopify refers to a decoupled architecture where the front-end (user interface) is separated from the back-end (e-commerce functionality). The front-end communicates with Shopify’s backend via APIs, such as the Storefront API, allowing businesses to build fully customized user experiences using modern technologies like React, GatsbyJS, or NextJS. Shopify provides tools like Hydrogen (a React-based framework) and Oxygen (a hosting solution) to support headless development[1][2][4].
Non-Headless Shopify, often called “traditional Shopify,” uses a monolithic structure where the front-end and back-end are tightly integrated. Merchants rely on pre-designed themes and Shopify’s built-in theme editor for customization. This approach is simpler and more user-friendly, with features like drag-and-drop editing and app integrations readily available[1][5][9].
Key Terminology
- Headless Commerce: Decoupling the front-end from the back-end for flexibility in design and functionality.
- Hydrogen and Oxygen: Shopify’s headless development stack for building and hosting custom storefronts.
- Storefront API: The API layer connecting the custom front-end to Shopify’s back-end services.
Advantages and Challenges of Each Approach
Headless Shopify
Advantages:
- Customization: Full creative control over the user interface, enabling unique designs and advanced features[1][8].
- Performance: Faster page load times since heavy backend processes are decoupled from the front-end[4][8].
- Omnichannel Experience: Supports multiple front-ends (e.g., websites, apps, kiosks) connected to a single backend[4][7].
- SEO Optimization: Enhanced control over SEO elements like navigation, rendering, and rich snippets[4].
Challenges:
- Complexity: Requires significant development expertise to build and maintain[3][8].
- Higher Costs: Initial setup is expensive due to custom coding and ongoing developer involvement[3][8].
- Loss of Built-In Features: Lacks access to Shopify’s theme editor, plugins, and some customer account functionalities[3][8].
- Integration Effort: Requires more effort to integrate third-party apps compared to traditional Shopify setups[8].
Non-Headless Shopify
Advantages:
- Ease of Use: No coding required; merchants can use pre-designed themes with drag-and-drop customization[1][9].
- Cost-Effective: Lower initial setup costs as it doesn’t require custom development[9].
- Built-In Features: Access to Shopify’s extensive app marketplace and built-in tools for managing customer accounts, payments, etc.[3][9].
Challenges:
- Limited Customization: Restricted by Shopify’s themes and templates; harder to create unique user experiences[4][9].
- Performance Issues: Heavy reliance on third-party apps can lead to slower loading times compared to headless setups[4].
Conclusion
The choice between headless and non-headless Shopify depends on your business needs:
- Opt for headless Shopify if you prioritize customization, performance, and omnichannel capabilities.
- Stick with non-headless Shopify if simplicity, cost savings, and ease of use are more important.
Citations:
[1] https://www.shopify.com/enterprise/blog/headless-commerce-vs-traditional-commerce
[2] https://lunatemplates.co/blogs/shopify-blog/what-is-headless-shopify-and-when-to-use-it
[3] https://www.plytix.com/blog/headless-commerce-with-shopify
[4] https://buttercms.com/blog/headless-shopify/
[5] https://www.shopify.com/ca/enterprise/blog/headless-commerce-vs-traditional-commerce
[6] https://community.shopify.com/c/hydrogen-headless-and-storefront/headless-domain-vs-shopify-checkout-domain/td-p/1275904
[7] https://www.shopify.com/enterprise/blog/headless-commerce
[8] https://naturaily.com/blog/headless-shopify
[9] https://swankyagency.com/shopify-themes-vs-headless/
Getting Started
Okay yeah. Here we go. I’m getting into the right mindset for the work ahead of me.
Finding the Path Forward
I don’t know the precise path yet. But I have some broad strokes. And it’s a very similar picture for a lot of clients these days, being squeezed between Amazon and the manufacturers, with a uniquely Shopify-flavored twist in the middle. Yup.
The Shopify Opportunity
Becoming a Shopify expert could go a long way to helping these types of clients.
And I need to continue that type of thinking.
Personal Goals
I should start measuring myself again on how much thought-work like this I get done, ere even the workday begins!
Gemini 1.5 Pro’s Thoughts
Title and Headline Ideas:
- The Squeeze: How to Survive in the E-Commerce Age
- Disintermediation and the Shopify Challenge
- Adding Value in a Crowded Marketplace: The Key to Success
- The Future of E-Commerce: Will Niche Sellers Survive?
My Opinion:
The article provides a thought-provoking analysis of the current state of e-commerce and the challenges facing niche sellers. The author’s discussion of disintermediation and the role of Shopify is particularly insightful. While the article focuses on the challenges, it also offers some solutions, such as adding value to products and services and leveraging technology. Overall, I found the article to be informative and helpful.