Nebulab

I took a much-needed LinkedIn hiatus during December, so… Here’s my take on Shopify’s Winter ‘26 Edition, in anticipation of a longer and deeper post on the Nebulab blog. SIDEKICK Love all the progress and how it’s getting integrated into more and more surfaces, but I’d like to see more reliability. Only a few weeks ago, I asked it how much revenue Klaviyo was driving for one of our clients, and it kept telling me zero. Most of the time, it’s been a hit-or-miss experience for me. That said, it’s great to see Shopify embrace the concept of Skills, and I’m excited that app developers can finally integrate with Sidekick, which was initially advertised in Toronto last year. Can’t wait to see what the ecosystem builds with this. As for the ability for Sidekick to generate admin apps, I predicted this would happen a couple of months ago right here on LinkedIn—I HAVE PROOF. It’s clear that this will eventually be expanded to storefront-facing functionality. Horrible news for anyone (barely) making a living by selling $9.99/mo apps, and a net positive for the app ecosystem, which is in dire need of a shakeup. AGENTIC STOREFRONTS This stuff is the reason people go with Shopify. The average Shopify brand can now click a couple of buttons and start selling on ChatGPT. Meanwhile, everyone else out there will have to either ignore agentic shopping until it’s big enough to warrant a structured investment or jump on the train early at the expense of other channels. A/B TESTING I’m not sure how I feel about this, and I won’t know until I see it in action. If I know anything about Shopify, this will be a good-enough entry point for small brands but not usable by larger players with more sophisticated needs (e.g., multiple treatments, multiple or custom metrics, audience exclusions, etc.). What I would have loved instead of a first-party solution is a native API that enables third parties to A/B test with fewer gimmicks. The bottom line is that A/B testing is a large and complex field, and even those who do it for a living often get it wrong. I feel like Shopify would do better to sell shovels than dig gold here, but who am I to say. PRODUCT NETWORK I feel like this has the seeds of something good, but it’s addressing the wrong problem. Rather than launching dropshipping 2.0, Shopify should use the tons of data it has to play matchmaker across its merchant network and facilitate proper brand collabs. Also, isn’t this the same thing as Shopify Collective? What am I missing here? SPECIAL MENTION: SIMGYM This was truly baffling. My understanding is that synthetic audiences are close to useless, especially when used for qualitative research. In my (admittedly limited) experience, qualitative synthetic data has been unreliable and inconsistent. This is confirmed by whatever little research has been done on the matter. I’m not sure why Shopify thinks this is a good thing to pour resources into. #shopify #ai #ecommerce #dtc

Most e‑commerce teams think their biggest risk is moving too slowly. In reality, it’s moving fast in the wrong direction. In our latest post, Andrea Vassallo breaks down a problem we see a little bit too often: brands hiring agencies that say yes to every request—new features, redesigns, unrealistic timelines—without asking the hard question: Does this actually serve your long-term digital strategy? Inside the article, Andrea digs into why a quick yes from your agency can lead to technical debt, a bloated CX, and declining performance, and why incremental, data-led changes often outperform big-bang projects. If you’ve ever felt like you’re managing your agency more than they’re managing your roadmap, this one’s worth a read. 👉 Read the full piece: https://lnkd.in/di-FPCpD #ecommerce #dtc #digitalstrategy #cro #agencies

Shopify Functions let you customize checkout logic, create dynamic discounts, and modify delivery options in ways that weren't possible before. But most brands are still figuring out how to use them effectively. In our latest learning session, Giacomo Baggio walks you through building custom Shopify Functions from scratch. You'll learn: → What Shopify Functions are and how they work → How to build discount and delivery customization functions → Best practices for monitoring, debugging, and performance 👉 Watch the whole session and see Giacomo build working functions in a live Shopify store: https://lnkd.in/dg4bRXim #shopify #shopifyplus #ecommerce #dtc #webdevelopment #shopifyfunctions

We're hiring a Shopify Engineer in Italy—and here's what you'd actually be working on: A 280-year-old textile brand managing four storefronts. A sustainable laptop company pushing the boundaries of repairability. A fresh dog food subscription rethinking pet nutrition. That's a typical month at Nebulab. You'd work on platform migrations, custom theme development, and the kind of complex catalog and subscription challenges that don't have obvious solutions. A few things that make us different: → Open-source is part of our DNA—several team members joined through community contributions, and we actively maintain our own Shopify tooling → We work with mid-market and enterprise brands who care about craft, not just speed → Remote-first, but we get together regularly If you're a developer who wants to work on interesting problems with a team that takes the work seriously but not themselves—we'd love to hear from you. Want to apply? Link in comments 👇 Know someone who'd be a great fit? We'd appreciate the share. #shopify #hiring #ecommerce #remotework

Shopify development doesn't have to be painful. We've been building on Shopify for years, and we've learned that the right tools make all the difference between frustration and flow. That's why we've open-sourced our internal toolkit—the same tools our team uses every day to build better Shopify experiences, faster. Inside this carousel: → ShopifyToolkit for metaobject/metafield management → ShopifyThemeToolkit for a production-ready theme starter → ShopifyThemeBuilder for component-based theme development → Minitest Shopify for unit-testing your themes All free. All on GitHub. All battle-tested on real client projects. If you're a Shopify developer, give them a try. We'd love to hear what you think! #shopify #shopifydevelopment #opensource #developerexperience #ecommerce

Brands don’t pick Shopify because the product is good. It is, but that’s only a small part of the rationale. Brands pick Shopify because: → They want to have first dibs on any innovations, trends, and integrations, and they know Shopify will always be at the bleeding edge. → They don’t want to deal with the operational overhead of maintaining something else, whether it’s a custom build or even a less-known SaaS player. → They want to feel like they’re well inside the Overton window when it comes to their choice of e-commerce platform, especially if they have a board to report to. How much merit these arguments have is a moot point. No one has ever sold anything by being rational about it. What matters is that anyone who wants to have a serious shot at competing with Shopify (which, and I think everyone agrees, could only be a good thing for all parties involved) must do much more than build a better mousetrap. So yes, the product matters. But so does the ecosystem, and so do the vibes. Shopify has nailed all three. #ecommerce #shopify #dtc

What if your AI assistant could browse your product catalog, understand customer preferences, and make recommendations—all without you hardcoding every scenario? Model Context Protocol (MCP) is Anthropic's standard for connecting AI agents to external systems, and it opens up a whole new world of possibilities. While most brands are still experimenting with basic chatbots, forward-thinking teams are building AI agents that can access your catalog, generate personalized recommendations, and augment your customer experience in ways traditional UIs cannot. In our latest learning session, Nicolò Rebughini walks you through a real-world MCP implementation for e-commerce. You'll learn: → What MCP is and why it matters for DTC brands → How to build your first MCP server in under 30 lines of code → How to integrate your MCP server in your LLM of choice 👉 Watch the full session and see Nicolò build a working AI agent for one of our clients: https://lnkd.in/dHgFGSJR #ecommerce #ai #mcp #dtc #shopify

Generative AI is NOT going to fix the internet. There’s a repeatable pattern in how the Internet evolves. It goes something like this: 1. We create a new technology that lowers the cost of entry. 2. Said technology creates an overabundance of choice. 3. We create another technology to help us navigate said overabundance. First, it was the search engines, which promised to help us filter through all the websites, the blog posts, and the product breakdowns and surface only those most relevant to our keyword. Then, it was the For You feed, which promised to help us filter through all the cat GIFs, vacation reels, and political rants and surface only those most relevant to our interests. Now, it’s generative AI, which is promising to help us with… everything, I guess? Ostensibly, OpenAI wants to displace search engines (ChatGPT), but they also want to displace social media (Sora), marketplaces (Instant Checkout), and browsing more broadly (Atlas). The promise is that AI operates based on stated intent, not assumed intent. Rather than praying to the algorithm gods and hoping that whatever you’re looking for will find its way to you, you can just say what you need and the LLM will move heaven and earth to bring it to you in its purest, most innocent, least adulterated form—like operating a giant, invisible hand that can reach straight into the belly of the internet. And right now, that promise mostly holds up to scrutiny, if you don’t mind the occasional hallucination. Except. Search engines got taken over by ads in a million different shapes and forms. The For You feed got taken over by content meant to appeal to our worst selves in a largely successful attempt to foster engagement and—you guessed it—show us more ads. Given the above, and given the state of OpenAI’s finances, on what basis exactly are we convincing ourselves that the same thing will not happen with generative AI? My hunch is that the only reason this hasn’t happened yet is that OpenAI (and everyone else) has much bigger fish to fry. They know we’re all willing to suspend disbelief as they solve for technical feasibility and distribution before they focus on fixing their P&L. But once that time comes—and I don’t think we’re too far away now—things are going to get interesting. To be clear, this is not to say that OpenAI is evil or that generative AI is useless. I’m still extremely bullish that it is changing and will continue to change our lives for the better. It’s just that the sirens’ call of advertising is inescapable, especially in an age where everyone wants everything and they want it for free and they want it right now, thank you very much. To look at a company that aims to quite literally become your one and only interface to the web and hope that they will somehow resist the urge to monetize every last bit of it, just like everyone who came before them, is to put it mildly, optimistic. #ai #advertising #openai