Scott Tolinski

Scott Tolinski co-created Level Up Tutorials and co-hosts Syntax, the web development podcast he started with fellow educator Wes Bos in 2017. He began by filming the Drupal steps other tutorials assumed viewers already knew. His work on AI coding asks a related question about what gets skipped. When an agent writes the code, how much work does the developer still have to do?

From Drupal tutorials to Syntax

Scott began recording YouTube tutorials in 2012 while building client sites at a small agency. A Drupal fix might be buried in a forum thread, and a tutorial could skip a step as basic as installing the software or adding a post. With a background in audio engineering and equipment already on hand, he filmed the steps he wished someone had shown him. He and Ben Schaaf started Level Up Tutorials together in March 2012, and Scott carried it on after Ben stepped away. Scott kept teaching while working as a developer, including at the University of Michigan’s internal web agency, before making Level Up Tutorials his full-time work. In his Svelte for Beginners course, learners start with components and inputs, then use those pieces to build a working quiz.

Scott and Wes met after releasing React courses with the same name. Wes sold his course. Scott put his on YouTube for free. They began talking regularly and decided to make a podcast together. When Syntax joined Sentry in 2023, Scott moved to working on the show full time, and the Level Up courses became free through Syntax.

What to build and what to hand off

Scott used The Break, an app for fellow breakdancers, to learn Svelte 5 and PocketBase. Dancers can keep a notebook of moves and combinations. In battle mode, they can mark moves as used after each round so they know what they still have available for the finals.

In his writing, Scott looks for work the browser already knows how to do. His repositioning dropdown uses an HTML popover to open and dismiss the menu and CSS anchor positioning to keep it next to its button. Where those features are supported, the menu can flip upward when it would run off the screen without JavaScript handling the flip. Knowing what the browser can do matters, he argues, when a coding agent can add a dependency faster than a developer can assess it.

With Prompt Boy, Scott replaced his own brittle agent runtime with Pi Durable, which saves progress in SQLite so another process can resume an interrupted session. His example shows why saving progress is only part of the job. An agent can usually repeat a search. If it was interrupted while sending an email, it should check whether the message went out before trying again.

The work an agent leaves behind

In the Syntax video The True Cost of AI Coding, Scott uses survey responses and expert interviews to explore why AI coding can energize some developers and wear others down. Responses continued to arrive after the video came out. By the time of AI Engineer Code 2026, nearly 4,000 developers had responded.

Scott makes the cost concrete in a Syntax discussion of AI fatigue. Answering an agent’s questions can improve the code it writes. The developer still has to make those decisions, read the result, and send it back when it misses the mark. An agent can take over the typing and leave the developer with a day full of decisions and review.