What happens when traditional online selling platforms meet the power of specialized LLMs wrapped in entrepreneur-friendly tools?



Well, predictably, it helps the average small business owner to build better, and increase his or her revenue, in a time when conventional jobs, by most accounts, are disappearing. That’s not to say that all of them are disappearing, but I think it would be fair to say that some of them are, especially given mass layoffs (Meta 16,000, UPS 48,000, etc.) and fears of job displacement from automation.



Anyway, some experts looking at the fallout from AI in the coming years suggest these may be great times for America’s small businesses, which can increase their productivity with new tools.



For example, the Shopify platform now has something called Tinker, which is supposed to enable the scores of tinkers on Shopify to tinker with the ecosystem in better ways. So what does it do?



Using Shopify Tinker



Tinker is something that at least one insider has called a “Iterative experimental playful canvas,” according to coverage from my favorite podcaster, Nathaniel Whittemore, who talked on a recent episode of his show, AI Daily Brief, about how all of this works, and how Tinker might help to boost revenues “where a ton of small entrepreneurship lives.”



Essentially, Tinker gives users access to a whole bunch of models, so that they can brainstorm what their business paths will look like. The application can also help them to generate resources like:



· product ideas



· marketing images and videos



· product descriptions or ads



By allowing sellers to experiment with different AI tools in one place, the thinking goes, Tinker will open up creative avenues to shop enhancement.



A More Technical Tinker



In a frustrating piece of theatre that’s unfortunately common in the tech world, hapless readers have to distinguish Shopify’s Tinker from another one. This other Tinker is made by Thinking Machines Lab, an AI startup founded in 2025 by Mira Murati, formerly of OpenAI, and is headquartered in San Francisco. It builds tools that help people create and customize AI models.





Specifically though, this technical Tinker makes use of something called LoRA, or “low rank adaptations.” If you’ll permit the tangent, I’d like to go down this rabbit hole to explain a little bit about this technique, since it also pertains, at its core, to AI development.



One way to describe what TML’s Tinker does is this: it freezes the initial pre-trained model weights, and then adds supplementary “low-rank matrices” into the overall equation. When I asked GPT to help me with this, it described a visual example where, instead of rewriting an entire database, an engineer can just attach two little thin rows of text or numbers, thereby “amending” the project for a more refined result.



Here’s another way to explain it: Tinker (with LoRA) is a way to quickly customize an AI model for your own tasks, like shop running or general tinkering. It lets you take a big, general model and “tune” it with small adjustments instead of retraining everything. You give it examples (like writing style, domain knowledge, or tone), and it learns those patterns fast.



Because it uses LoRA, it’s cheaper, faster, and needs less computing power. You can create a specialized version of a model—like for legal writing, customer support, or coding—without rebuilding the whole system.



The Merchant as Tinker



As for that Shopify version, redditors critiquing the project note that it sometimes helps a shop owner with mental block to “get over the hump” and manage the malaise that comes with being overwhelmed by the nooks and crannies of modern commerce. Proponents project an increase in small business entrepreneurship. One, quoted by Whittemore, axiomates the concept this way: “If you want more artists, lower the cost of paint.”



That’s the Tale of Two Tinkers, in a world where each new model application gets its own strange name, and dozens of them pop onto our radar frequently. Stay tuned for more.