What clicked for us this week

Another week of our unfiltered thoughts - what we're actually doing with AI, what's working, what surprised us, and what you might want to try yourself.

Let’s get into it.

This past week we plugged our AI engineering workflow into a task management program called Linear, and it has completely changed our development process. Linear is a product management tool, like Jira or Notion, designed for teams to coordinate work together. Linear was actually purpose built for engineers, with lots of keyboard shortcuts so that an engineer could whip through their tasks with command keys, like a financial analyst in excel. 

In fact, Linear is considered to have one of the most user-friendly experiences for products on the market.

But when software development switched from engineers writing code to developers prompting AI models, Linear doubled down on their AI strategy. They decided to optimize their product for AI agents, and having just experienced what this means for the past week, it is truly a game changer.

Now, every single timeI do any development work, I have my agent create a ticket in Linear and track the work. I don’t have to do any writing or organizing – the agent just does it for me. And it creates a system of record that compounds knowledge for our projects over time, so that we have complete contextual memory through each task we accomplished. This is hands-down one of the most useful additions that we’ve made to our AI tech stack this year. 

A client of ours runs their entire ecommerce operation through Square. It's pretty standard; most ecommerce businesses are on platforms like Square or Shopify because they're industry leaders. But being an industry leader doesn't mean you have the best UI. Sometimes it just means you have the most market share. So there was a lot of optimization work to do: inventory settings, product configurations, site cleanup. The kind of stuff that matters but is painfully tedious to do through their interface.

Then I discovered Square has an MCP connection (a miracle!). For anyone unfamiliar, MCP (Model Context Protocol) is basically a way for AI tools like Claude to plug directly into the software you're already using – read data, make changes, update settings – without you ever touching the UI. So instead of clicking through menus for hours, I connected it to Claude Code and started making changes conversationally: “Update this. Change that.” 

Here's what clicked for me: we spend so much time thinking about replacing our tools when the real problem isn't always the tool itself. Sometimes, it's the interface. MCP basically lets you skip the interface entirely and talk directly to the system underneath. The platform stays the same, the data stays the same, you just stop fighting the UI. Check whether the tools you're already paying for have MCP connections. A lot of them do now and most people have no idea. You don't always need to switch platforms to work faster, you might just need a better way to talk to the one you're already on.

Stay curious,

Julia & Russell