I'm surprised I haven't seen an FPS frontline drone sim game yet

Kinda cool for a marketing website component?

The modern AI stack has three layers. Models. OpenAI, Anthropic, Google. Billions in capex, rapidly commoditising. Indie devs can't compete here. Harnesses. Claude Code, Cursor, Copilot, custom chat UIs. The runtime that calls models, manages conversation, executes tools. Commoditising fast. Indie devs probably can't compete here either. Context. Domain knowledge, structured data, business rules, accumulated expertise. The stuff that turns a generic AI into a domain AI expert. This is where indie devs build. Context is the new SaaS layer. Models get cheaper every quarter, harnesses get replaced every year, but a well-built context layer compounds. Store briefs, pricing playbooks, supplier notes, audit history. Each entry makes the next agent smarter, and none of it is locked to a specific runtime. So pick a vertical (retail, legal, healthcare, music production) and build the context infrastructure that makes any harness an expert in that domain. Files the user owns. Indexed for search. Versioned with git. Exposed through MCP, so it works with Claude Code today and whatever wins tomorrow. Don't build the AI. Build what makes the AI worth using.

Holy shit, this is BRILLIANT: This dude breaks down why everyone thinks the whole WHCD event was fake... everything about trump is FAKE. Best video you'll see today.

Context Surfaces are the new SaaS

as the youngest millennials turn 30 it feels like PS1 aesthetic is slowly overtaking 8bit pixel art

🦝 Animal #985 🐀 I figured it out in 8 guesses! 🟥🟥🟥🟥🟥🟥🟧🟩 🔥 4 | Avg. Guesses: 10.8 metazooa.com

Some admin dashboard design

Can we ship in Australian journalists? Because of course this was one of them. Their entire country is better at our national media than we are.

Trump is sitting down a lot. So low energy.

Reddit sucks now for real-time news, Twitter is full of AI fakes, wtf has the internet turned into?

I much prefer using Cursor, but I run out of tokens way too fast vs Claude Code and the max sub.

I hope skinny leg jeans comes back

There's a moment when you're using AI and it shifts from "impressive party trick" to "this actually changes how I run my business." For me, it was watching Claude read the entire Bloody Good Tests product catalogue and start working with it like a colleague who'd been on the team for years. Not summarising. Not chatting about it. Actually pulling products, checking fields, cross-referencing data, and writing structured reports. All from a terminal. Claude Code supports the MCP (Model Context Protocol). It's a standard way for AI to talk to external services. Instead of copying data into a prompt, the AI connects directly to your Shopify store via the Admin API (and a custom app). That means Claude can: - Pull your full product catalogue with pagination - Read every field: descriptions, SEO metadata, tags, variants, pricing, inventory, images, metafields - Update products, manage variants, create redirects - Do all of this in the middle of a conversation, on demand It's not a one-way data dump. It's a live, two-way connection. Claude can read from Shopify and write back to it. Once Claude has live access to your store, the question stops being "can AI help with my Shopify store?" and becomes "what should I point it at first?" Some things I've built on top of this connection: A 10-module product audit that checks every product for content quality, SEO health, regulatory compliance, pricing sanity, image completeness, variant data integrity, cross-sell opportunities, and more. Each module runs as a separate AI agent, all in parallel. Results go straight to Notion as colour-coded to-do checklists. A competitor price comparison pipeline that scrapes six competitors' catalogues, matches their products to mine using name matching and fuzzy aliases, and classifies every product as overpriced, underpriced, or competitive relative to market median. None of this requires leaving the terminal. No app switching, no admin panel clicking, no CSV exports and imports. The product audit is where this gets interesting architecturally. Each audit module is defined in a markdown file (a SKILL.md) that tells an AI agent what to check and how to score findings. The orchestrator reads all 10 modules and dispatches them as separate agents simultaneously. Each agent works independently, reads the product data from Shopify, and returns structured JSON findings. The orchestrator merges everything, groups by product, and writes it to Notion for review. This is the direction ecommerce operations is heading. Not AI that writes a product description when you ask nicely, but AI that has persistent access to your systems and can do real operational work across them. The gap between "AI can help with ecommerce" and "AI is plugged into my store and doing operational work" is smaller than most people think. The tools are here. The hard part isn't the technology — it's deciding what to point it at first.

The next project will be connecting the Shopify Admin MCP with a custom Google Ads MCP, Google Search Console MCP, and Ahrefs to build an automated marketing engine for each product. 🙌

I built an AI system that audits all my Shopify products across 10 quality dimensions simultaneously using parallel Claude Code agents, a custom Shopify MCP server, and Notion as the review/approval interface

This is a very bad change. This will make the developer experience terrible. @liam_at_shopify, can the team reconsider this change?