Andrew Gremlich

@gremlich.dev

Senior Software Engineer and Transformative Mediation

I'll be presenting on conflict management in the workplace for the first time at UtahJS SLC meetup! This is something I think is much needed. I look forward to presenting this and refining the presentation! Come see the presentation! www.meetup.com/utahjs/event...

UtahJS SLC Meetup - Generators & Healthier Workspaces, Tue, Aug 18, 2026, 7:00 PM | Meetup

Join us in person or [on Zoom](https://us06web.zoom.us/j/86837702144?pwd=vkE9HYJqIKKbnUSl7xcq5hqsEU5PwS.1)! **7:00pm** \- Dinner and Welcome **7:05pm** \- "Generators in

meetup.com

I've felt to clarify my stance on LLM technology, just to be sure. It's great tech and it will definitely influence many aspects of life — for good and bad. I want use the tech for good, meaning healthy usage that promotes a better overall well being. That can take many shapes!

One pet peeve of LLMs has been when any bash commands run because the LLM just shows one huge blob and they're hard to read. Then I had an idea to put in the system prompt to make those bash commands more readable. Turns out that was the trick!

I think that Vite + Web Components is an unknown elite combo! Using Vite's `?inline` and `?raw`, `.css` and `.html` files can be imported into a `.ts` file and attached to the web component. No framework, extensions, or library. I did this as well to get the syntax highlighting of the editor.

Code editor screenshot showing TypeScript web component with all lifecycle callbacks: sheet const defined at module level (line 4-5, fixing earlier undefined-variable bug), CustomWebComponent class extends HTMLElement, private #shadow field, constructor attaches shadow root and sets adopted stylesheets plus innerHTML from template, followed by observedAttributes getter and single instances of connectedCallback, disconnectedCallback, connectedMoveCallback, adoptedCallback, attributeChangedCallback — duplicate attributeChangedCallback from earlier removed.

I'm working through how I am justifying some of my behavior which leads to a self-betrayal. It would seem that I have an expectation to receive something once I have given feedback. I turn things transactional. So if I don't receive some benefit from what I give, then I justify and withdraw.

I’m partly inclined to view using LLMs similar to gambling. I feel it implicitly discourages diligent work, and yet there’s a potential for a big payoff. This may be problematic for using LLMs. Several times I’ve corrected someone using LLM info with actual documentation.

I'm redoing my portfolio website, and I love just taking a pause to try and make my portfolio website a bit more creative! I'm making a hero component for it and I am animating it to reveal the background image. Check out my progress! www.gremlich.dev

I was seeing how I could make a starfield in CSS, and I stumbled upon the random() CSS function! In my use, it takes two parameters that are of the same unit, and it can generate any value in between those two units. Unfortunately it's only supported in Safari.

Bild

I got to try this out in my web graphics workshop where I learn about CSS! I'm trying to make a digital art landscape using only CSS and the random function helps in one aspect. I am making a night sky and I use the random function in placement of the stars.

CSS-Tricks*@css-tricks.com · 4mo ago

We're finally getting new random() and random-item() functions! I say *finally* because the road to get there has been a long and winding one. @alvaromontoro.com paves that road nicely. css-tricks.com/the-importan...

Another study saying be careful when having AI-enabled workflows. I know AI technology is great to automate, jump start a project, and figure out obscure bugs. But if it isn't used carefully, then it might cause more harm than originally intended. hbr.org/2026/02/ai-d...

AI Doesn’t Reduce Work—It Intensifies It

One of the promises of AI is that it can reduce workloads so employees can focus more on higher-value and more engaging tasks. But according to new research, AI tools don’t reduce work, they consisten...

hbr.org

Anthropic came out with a paper that seems to say skills development in an AI enabled work place needs to be done cautiously. Namely, don't let AI do all the work, instead, interact with AI to make sure a conceptual understanding is gained. www.anthropic.com/research/AI-...

How AI assistance impacts the formation of coding skills

Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

anthropic.com