Published

All published posts

2457 posts latest post 2026-04-19
Publishing rhythm
Apr 2026 | 40 posts

A really interesting long form interview with @simonwillison.net. If you follow him closely most of it is probably not new, but I found some interesting nuggets.

Simon is writing most of his code from his phone these days using anthropic hosted platform. He mentioned that a lot of security risks go away when you don’t put secrets on the platform and you let them take the risk of running ai written code with ai chosen supply chain.

He talked about the Pelican Riding a Bike benchmark for quite awhile. He was surprised at how well of a proxy it is for how capable a model is at just about everything. He also said that when he runs the benchmark he also runs half a dozen others that he’s never talked about so that He could see if they were to train a model specific to his benchmark he could catch them, but it seems they had caught on and if they were they seem that they would already be doing it on all of his others anyways.

TDD is incredibly boring for humans, it strips so much creativity and joy from the process. Who cares if agents are bored they do better when doing TDD.

Absolutely sick texture app from cnc kitchen. Like him I’ve spent a bunch of time attempting and failing to learn blender, I’m so glad someone else vibe coded out such a good app that can just add texture to stls with basic masks and is the very basics of what you would want to add to 3d prints to make them interesting, I’m excited to use this for some real projects.

Wonka Letters
Wonka letters all cut out ready to get some stiffeners and go off for paint.

What is this job anymore

The job of writing code is dying, models are getting better, the average person will have their average features implemented in average ways with no effort by agents, the writing is on the wall. We are still trying to review most of the critical code, this is slowing us down, is it really stopping any bugs or giving us any more familiarity with the product, marginally. The time is now to grease up your UAT, testing, deployment pipelines. Dont let agents delete entire regions. Review your backup and restore strategy, you do have a DR plan right? Things are changing fast, the best of us are still better than the clankers. Most of us have more context than the clankers. Most of us have more intuition of what and where to implement fixes. Context windows and memory will be solved problems. Your DR plan, UAT, testinng and QA environments will not come for free, you need to make them, and deeply integrate them into your processes.
View
Hair Whittling Sharp
Hair whittling sharp, Do I get my redneck nerd card yet?
Llama In Pi Thinks Its Claude
I just launched ollama picked pi as it asked what harness I wanted to run, and it responded telling me it was claude.
Ty 0.0.26
ty 0.0.26 was released on 3/26/26, nice work planning.
What a banger of a tui, fantastic job cloning monkeytype. Looks so good. The toast messages are a tell tale built with textual.
Sparklines On The Feeds Header
View of the new markata-go feeds header with the banger of a sparkline.
Getting Excited For This New Feeds Page
This sparklines on this new feeds page are chefs kiss.

The year of the supply chain attacks

I think I'm starting to understand my role as a platform developer in 2026. * least priveleged access * default deny + explicit allow * understand your blast radius * **GREASED** creds rotate process * PIN EVERYTHING * keep packages up to date * but not too up to date, use dependency cooldowns
View

The final nail for Windows?

Easy anticheat for linux is out. !!! tip look at the date If this were real what would you play first? For me it's `skate .` is really the only thing I care about and I'm fine without it.
View

I’ve been thinking about this for awhile and Daniel makes some great arguments here. Interestingly keeping inference cheap removes the incentives to make our tools better, help us choose the right model, lean on local models, open weight models. The frontier models are so affordable through subsidized subscription models why would you deal with anything less intelligent at this point. The tooling we use is not optimized for it, and why should it be.