Jason’s north star for coding decisions: simplicity

Jason, a software engineer with over 15 years of experience who works at Meta, argues that every choice made by a developer or coding agent shapes an app. Blindly accepting an agent’s fixes can leave a system fragile as it scales. Simplicity is not choosing the easiest immediate fix; it is doing the harder work of finding the cleanest solution to the underlying problem.

A water-tower illustration contrasts repeated patches with stepping back to repair a leak from inside the tank [inferred from described visual]. The coding equivalent is a shopping cart whose separate cart count variable needs another update rule whenever an item is added, removed or restored on page reload. Deriving the count from the cart array removes that coordination problem. Jason applies the same reasoning to DRY (don't repeat yourself): use design tokens rather than setting every button’s CSS color separately; extract reusable code into shared libraries; and, in a large codebase, use configs and generators to produce compliant schemas and required validations instead of asking each developer to remember them.

For Altitude, Jason weighed two designs for a tutor that helps users understand code produced by AI. An all-in-one app would bundle the AI, editor and terminal, but require rebuilding a coding tool and paying API token prices. A plug-in for an existing tool would let users keep their workflow and subscription; more importantly, they could continue using the same tool after learning. He chose the plug-in because it better serves users who want to build real projects, use AI well and get jobs.

A second Altitude decision concerned its mock-data project prototypes. Adding detailed rules, changing models and testing effort levels did not reliably fix generic-looking results. Jason instead supplied the generator with users’ goals and intent, and replaced prescriptions with guidance to make the prototype clear, intuitive and useful. He reports better results in side-by-side tests and a simpler skill file [inferred from described comparisons]. His two takeaways are to let simplicity guide decisions—step back to the higher-level goal when fixes accumulate—and to slow down and care about the user’s experience. For him, that care distinguishes thoughtful AI-assisted work from ‘AI slop.’