1441
Some thoughts on LLMs and software development
a short post
New
The latest blog posts to take off on Hacker News. 14,537 posts.
1442
Building your own CLI coding agent with Pydantic-AI
How to build a CLI coding agent
1443
Making Minecraft Spherical
Discover the unique design challenges of creating a spherical planet out of Minecraft-like blocks.
1444
Uncertain<T>
GPS coordinates aren’t exact. Sensor readings have noise. User behavior is probabilistic. Yet we write code that pretends uncertainty doesn’t exist, forcing messy real-world data through clean Boolean logic.
1445
That boolean should probably be something else
One of the first types we learn about is the boolean. It's pretty natural to use, because boolean logic underpins much of modern computing. And yet, it's one of the types we should probably be using a lot less of. In almost every single instance when you use a boolean, it should be something else.
1446
Unexpected productivity boost of Rust
Rust is known for having a steep learning curve, but in this post we will be looking at another curve, the one showing developer productivity in relation to the project size.
1447
Implementing Forth in Go and C
I first ran into Forth about 20 years ago when reading a book about designing embedded hardware . The reason I got the book back then was to actually learn more about the HW aspects, so having skimmed the Forth chapter I just registered an "oh, this is neat" mental note and moved on with my life. Over the last two…
1448
How to slow down a program and why it can be useful
Making programs slower can be useful to find...
1449
Delete tests
We’ve had decades of thought leadership around testing, especially coming from holistic development philosophies like Agile, TDD, and BDD. After all that time and several supposedly superseding movements, the developers I talk to seem to have developed a folk wisdom around tests. That consensus seems to boil down to…
1450
Malleable Software
In the AI era, the winners won’t be the tools you adapt to — they’ll be the tools that adapt to you.
1451
Thrashing
I’ve invested in a new odometer to help my car go faster. Accountability is important, and as a driver I believe the most important thing I can do is set measurable, achievable and inspiring goals for the people pushing my car down the road.
1452
Why I'm declining your AI generated MR
Sometimes a merge request (MR) doesn't merit a code review (CR) because AI was used in a bad way that harms the team or the project. For example:
1453
"Special register groups" invaded computer dictionaries for decades (2019)
Half a century ago, the puzzling phrase "special register groups" started showing up in definitions of "CPU", and it is still there. In this...
1454
Rv, a new kind of Ruby management tool
For the last ten years or so of working on Bundler, I’ve had a wish rattling around: I want a better dependency manager. It doesn’t just manage your gems, it manages your ruby versions, too. It doesn’t just manage your ruby versions, it installs pre-compiled rubies so you don’t have to wait for ruby to compile from…
1455
Dangerous advice for software engineers
I’m a big fan of “sharp tools”. These are tools that are powerful enough to be hugely helpful or harmful, depending on how they’re used. Most forms of direct…
1456
macOS dotfiles should not go in –/Library/Application Support
A CLI tool is not an app, and none of you are reading the docs as carefully as you think you are
1457
1458
Everything I know about good API design
Most of what modern software engineers do involves APIs: public interfaces for communicating with a program, like this one from Twilio. I’ve spent a lot of time…
1459
Making games in Go: 3 months without LLMs vs. 3 days with LLMs
After 15 years as a software engineer, I realized I had never actually built and published a game.
1460
Dynamically patch a Python function's source code at runtime
written by Eric J. Ma on 2025-08-23 | tags: python runtime llm security namespace compilation execution functions toolbot monkeypatching
1461
Turning Claude Code into my best design partner
When I first started using Claude Code, I had a naive approach to working with it. I would describe the task directly in the prompt, press Enter, and cross my fingers. If the agent made mistakes, I would tell it how to fix them. For small tasks, this can be good enough, but as the task grows in complexity, this…
1462
A visual introduction to big O notation
A visual introduction to big O notation.
1463
DeepWiki: Understand Any Codebase
How I use DeepWiki to speed up real coding work
1464
In-Memory Filesystems in Rust
I’ve been working on a CLI tool recently, and one of the things it does is manage files on disk. I have written a lot of file management tests for Bundler, and the two biggest reasons that the Bundler test suite is slow are exec and fstat. Knowing that, I thought I would try to get out ahead of the slow file stat…
1465
How to make things slower so they go faster
In a service with capacity $\mu$ requests per second and background load $\lambda_0$, the usable headroom is $H = \mu - \lambda_0 > 0$. When $M$ clients align, e.g., after a cache expiry, at a cron boundary, or as a service returns from an outage, the bucketed arrival rate can
1466
How to build a coding agent
It's not that hard to build a coding agent. 300 lines of code running in a loop with LLM tokens. You just keep throwing tokens at the loop, and then you've got yourself an agent.
1467
AGI is an engineering problem, not a model training problem
LLM models are plateauing, but true AGI isn't about scaling the next breakthrough model—it's about engineering the right context, memory, and workflow systems. AGI is fundamentally a systems engineering problem, not a model training problem.
1468
Materialized views are obviously useful
As programmers we spend a lot of time shuttling data back and forth between different systems and transforming it from one format to another. Sometimes it gets pretty miserable!
1469
RFC 9839 and Bad Unicode
Unicode is good. If you’re designing a data structure or protocol that has text fields, they should contain Unicode characters encoded in UTF-8. There’s another question, though: “ Which Unicode characters?” The answer is “Not all of them, please exclude some.”
1470
Writing Speed-of-Light Flash Attention for 5090 in CUDA C++
In this post, I will walk through how I learned to implement Flash Attention for 5090 in CUDA C++. The main objective is to learn writing attention in CUDA C++, since many features are not available in Triton, such as MXFP8 / NVFP4 MMA for sm120. I also feel this is a natural next step after learning about matmul…