danmurrelljr

Software Engineer | Writer | Technologist

Former Engineering Manager, Mobile Apps @BigCommerce

Austin, TX | Mississippi State University Alumnus

He/Him/His

linkedin.com/in/danmurrelljr

danmurrelljrdanmurrelljr
2025-06-10

@fcloth @marcoarment it's beta 1, dude. Maybe even beta 0, because what I installed on my iPad yesterday doesn't look as good as what was shown in the keynote or in other videos I've watched so far.

The first few betas are always rough – then they get tons of feedback and dial it in by the fall.

danmurrelljrdanmurrelljr
2025-06-09

“Back up my iPad..” oh, please lol

danmurrelljrdanmurrelljr
2025-06-09

Wait, is the iPad getting the most "holy shit"and "finally" updates this year?

danmurrelljrdanmurrelljr
2025-06-09

These new features are really cool, but isn't this a developer's conference? Not a whole lot of the "And for developers, ..." content.

(But maybe it's always been this way, and we're more sensitive this year for some reason 😊)

danmurrelljrdanmurrelljr
2025-06-09

This Liquid Glass design is a lot more interesting than I was expecting. Good stuff.

danmurrelljrdanmurrelljr
2025-06-06

Today's episode of Dithering fooled me with the title (Markdown and AI Dev Tools). They were two separate, unrelated topics.

I thought @gruber and Ben were going to talk about how Markdown is the perfect portable format for AI development. Because it kinda is.

Think about it.

(Also I've already been working on a generative AI project that uses Markdown as its portable format. .md hasn't let me down yet lol)

danmurrelljrdanmurrelljr
2025-06-05

We need a sharper line between augmented engineering and vibe coding.

A vibe coder is a creator, but not an engineer.

An engineer studied and honed the skills to understand what is doing.

Both provide taste, the critical subjective human element necessary to create something good using AI.

danmurrelljrdanmurrelljr
2025-05-25

Trying to set up a project on to test models, whether on Azure or other open source models. I thought 's dashboard was byzantine, frustrating, and aggravating. Now there's a new leading candidate for user-hostile interfaces.

Who designs and builds this? Do they use it?

Why does it let me deploy a model, but then I can't use it because inexplicably it either deploys under Azure AI Foundry instead of Azure OpenAI, or in swedencentral instead of useast? I'm in neither

danmurrelljrdanmurrelljr
2025-05-19

@manton If you only have Plus like me, you can try it out with the CLI version. It's pretty interesting as well (you do have to give it an API key though, so will be charged, but it's o4-mini rates)

danmurrelljr boosted:
kottke.orgkottke
2025-05-14

Three Fascism Experts on Why They’re Leaving the US. “I want to do my work without the fear that I will be punished for my words.” And: “The lesson of 1933 is you get out sooner rather than later.” kottke.org/25/05/three-fascism

danmurrelljrdanmurrelljr
2025-05-13

@benbrown not really, just learned to live with it and tune it out :/ It's Texas, so maybe having ceiling fans going 24/7/365 helps too

danmurrelljrdanmurrelljr
2025-05-08

@manton I wonder if @gruber is still surprised at Markdown's legs. It keeps being perfectly suited for every new watershed tech moment that comes along

danmurrelljrdanmurrelljr
2025-05-07

@manton ah! I did read that the other day.

In my package github.com/AuroraToolkit/Auror I did pretrain a few CoreML models for the examples, but if you're wanting to take a blog's tags and train a model to classify or suggest tags for a post in realtime, I think you can do it all in code.

I'm working on being able to mix cloud/local LLM and ML tasks, with a fallback to a heavy hitter LLM as last resort. Best of all worlds :)

danmurrelljrdanmurrelljr
2025-05-07

@manton I originally built my AI swift package to mainly use LLM for things like categorization, but currently am adding CoreML & NaturalLanguage counterparts for some of these tasks (like categorization). You might be able to do this with CoreML instead of an LLM. Those models are tiny and wickedly fast (you might still want to call out to an LLM to get training data though for your categories).

Looking for your post now :)

danmurrelljrdanmurrelljr
2025-05-04

@fichek right but the users are going to ultimately choose which payment method they'll go with, if any at all, and IAP is almost frictionless. We can offer that link out, but user behavior is crystallized by this point.

danmurrelljrdanmurrelljr
2025-05-04

@lexfri that Mrs. Roper bar crawl is a real thing. I haven't seen one in real life yet, but the idea is hilarous

danmurrelljrdanmurrelljr
2025-05-04

@eliaschao @fichek @davedelong they still have a legitimate argument that the benefits of being on the App Store in front of a billion potential users is worth the 30% or 15% for small devs.

But for businesses that have to compete on prices, that 30% made you choose between eating the cost, or adjusting the price and becoming uncompetitive or even unprofitable.

When companies like Amazon and Netflix won't play your game, you know it's a bad idea.

danmurrelljrdanmurrelljr
2025-05-04

@davedelong I'm really curious how it will play out. I have a feeling indie devs like me will still see IAP continue to be the user’s payment of choice, because the experience Apple built is so good.

But for businesses, it'll be huge and immediate. My mobile team at BigCommerce were forced to remove a link to our website because an app reviewer found a way to get to our company's subscription page. The old rules unfairly went after our business model, unrelated to mobile.

danmurrelljrdanmurrelljr
2025-05-03

Domain classification strategies for routing to multiple models: danmurrelljr.micro.blog/2025/0

Client Info

Server: https://mastodon.social
Version: 2025.04
Repository: https://github.com/cyevgeniy/lmst