Hrishi Mittal

“We’ve had a quote from another developer who’ll do it for less than £1,000, so we’ll go with that.” I’ve heard this line more times than I can count. And fair enough — budgets are real, margins are tight. But let’s talk about what that actually means in eCommerce. A #Shopify build isn’t just a digital brochure. It’s your storefront, your sales engine, your first impression. And when a developer quotes under £1,000 for a full build or complex customisation, it usually means: – No proper QA – No documentation – No post-launch support – A Frankenstein of cut-and-paste code from GitHub or old projects – Or worse — pirated themes, dodgy apps, and shortcuts that’ll come back to bite What starts as “cheap” often ends in costly rebuilds, lost sales, and frantic emergency calls. I’ve been brought in too many times to clean up that mess — and by then, the savings are long gone. If your Shopify store is a key revenue channel, treat it like one. Pay for craftsmanship, not just code. Cheap is expensive when it’s your brand on the line. #shopifywebsite #shopifydevelopment #shopifyplus #shopifyexperts #shopifystore #shopifytheme #shopifyapps #shopifyappdevelopment #themes #integrations #solutions #software #architecture #development #brands #shopifybrands #shopifyextensions #ecommerce #Shopify #webdevelopment #merchantadvice #longtermthinking #investwisely

94

38 Comments

Going from business ID or company name → verified website and socials is trickier than it sounds. This is a super common use-case if you're targeting local companies. You might find a registry of governmental-type data like business ID, company name, perhaps their address. But these records seldomly contains the company's website – and virtually never their LinkedIn. Almost all enrichments in Clay depend on a website or company LinkedIn. If you have them, everything's easy. If you don't, nothing works. That's why I built this Clay template that does this research automatically. How it works: 1. Input whatever you have (business ID, company name, address – anything helps) 2. AI researches to find their website, LinkedIn, Twitter, Facebook, Instagram, TikTok, and YouTube 3. Returns each result with a confidence score and reasoning 4. Create formula columns to filter by your desired confidence threshold 5. Recommended: Use high-confidence results immediately, manually verify low-confidence matches What's important for you to decide:do you want to minimize false positives (only high confidence findings) or false negatives (accept some findings with uncertainty)? The best answer depends on your use case. But either way – this template lets you control it. Hope you find it useful! Free template link: https://lnkd.in/dAXcvzEk

43

17 Comments

Architecture components should not be optional. Architecture is all about consistency and predictability. It's meant to free our minds to focus on our unique problem without having to worry about gluing it together. If you feel that a component in your system should be optional, ask yourself this: can remove it entirely without sacrificing the architecture principles? If the answer is yes, remove it everywhere. Otherwise, keep it everywhere. Consider the alternative: you remove a component from some of the places but not from others. Now, when you approach some code - you don't know what to expect. Is the component going to be there? If you change the code - should you add it? Should you remove it? Did you account for the extra work when you planned the change? Is it worth it? In short, consistency and predictability win the day. Every time.

29

4 Comments

WTF changed!? Agentic coding... finally works. Over 50% of my programming is now done by claude. In a year, it will be over 90%. I am already 3x as productive. Mind blown. Still doesn't work very well out-of-the-box, but now 100x easier to get something insanely productive: * Planning first! If you ask for claude to generate a step-by-step plan _in an editable file_, you will get 10x better results. Let it ask you questions before it dives in. * Invest in building custom commands. That kicks off the iterative improvement loop. Don't like what the AI did? Tweak the command. Rinse and repeat. Ad hoc prompting won't get you very far. * Context compaction is the devil. Code quality drops 90% after. Invest in building commands to rebuild the relevant context. * Code quality problems = Style guide problems. Give AI plenty of examples about the "right" and "wrong" way to do things, it writes better code than 90% of programmers I've worked with. * Embed relevant third-party documentation directly in your codebase. AI sucks at searching for it on the fly... yes, even with Context7. I hope this changes, but it's worth it. * 99% of MCPs aren't worth the hype... yet. Don't waste your time... yet. Getting an AI to intelligently use an MCP tool without constant, explicit reminding is an exercise in futility. Something needs to be fixed here; the issue feels structural and we probably need a model that is more "tool-oriented."

370

240 Comments

I see a lot of developers being skeptical of AI code generation. Some stating that it´s only useful for inline completions, some stating that it´s not useful at all. I myself want to see where the boundaries actually are. This is an ongoing proof of concept to build a turn based hex-map strategy game, with 100% AI generated code. Zero lines of manual code so far. So far we have: * 3D hex map * Pan and zoom features * Pathfinding * Minimap overview and quick navigation * Unit movement More to come. Can AI generate all this from a single prompt? No it can not, you have to have a continuous dialog with the AI to tune each feature. But it can indeed build this faster than I could do it myself, even with my 30+ years in the industry. What about the code quality? It´s pretty decent, but you do have to guide it to re-use calculations and to create components for things to avoid unintended changes to unrelated parts of the code. Give this a few years and software development will be completely changed forever

66

55 Comments

One of the hardest truths in software is that users are only one bug away from losing trust in your brand. At Spur, that truth drives everything we build. We want founders and product teams to see QA not as a checkbox, but as a smart, strategic investment that protects their business and reputation as they scale. Spur’s agentic QA behaves like a real user. It explores your product, surfaces friction points, and catches issues traditional scripts can’t reach. Every test reflects a real user journey, not just a technical path. We take a holistic approach to QA. Focusing only on code paths or brittle test suites misses what matters most: user experience and trust. Our goal is to make QA an invisible, always-on part of your workflow, helping you fix issues before your customers ever find them. And because we’re always listening, something new is on the way. Soon, the bugs caught by Spur won’t just be surfaced. They’ll be fixable instantly, right where you work. Stay tuned. Big things are coming for teams who want QA to be fast, actionable, and seamlessly connected to development.

67

5 Comments