• The fetch()ening: Htmx is getting a new major release version. Sounds like a solid upgrade. I love the philosophy behind this project. Small, purpose built, server side rendering, very stable, grounded in the original ideas behind hypermedia.
  • Deploying to Amazon’s cloud is a pain in the AWS younger devs won’t tolerate. Yep.
  • The Rule Of 2: Consider 3 categories of risk when deciding whether new behaviour should land - trustworthiness of inputs, the implementation language, and the privilege mode it has to run in. You can’t have more than 2 of those. Tricks to de-fang some of the issues here are to run code in a sandbox, or building safety guardrails around something that is otherwise unsafe.
  • Importing vs fetching JSON: Interesting. I guess there’s a need to read data locally if you are bundling code and data into a package that is a deployable unit?
  • Should You Use Big Models or Services?: “Models are nouns, and services are verbs.”, “When your method needs to interact with multiple models maybe extracting it is a good choice.”, “It’s all subjective in the end.”
  • Game design is simple, actually: Lots of ideas in here about what makes us tick. I can recognize stuff he talks about in some of the games I have really enjoyed playing like practice, and obstacles progressively getting harder. The framing is interesting of play as problem mastery and predictability!
  • You Should Write An Agent: Agents are simpler to write then I thought. Using one of the big llm providers, all we have to do allegedly is provide context (an array of strings) and a prompt.
from openai import OpenAI

client = OpenAI()
context = []

def call():
    return client.responses.create(model="gpt-5", input=context)

def process(line):
    context.append({"role": "user", "content": line})
    response = call()    
    context.append({"role": "assistant", "content": response.output_text})        
    return response.output_text
  • Supply chain attacks, npm: Pinning, lockfiles, npm ci, and scanning.
  • How I Use Every Claude Code Feature: Practical tips for claude and other llms.
  • What we look for in a resume: Great.
    • Demonstrated expertise, not keywords
      • Show how you acquired and use that skill in your job
      • Share your expertise on public channels such as: StackOverflow answers, open source contributions, papers, blog posts
    • We look for people who get things done
      • Initiative - Can you start something that matters?
      • Persistence - Can you see those hard things through? Getting other people on board, making them feel safe and like participants, following up, etc
    • We look for unique perspectives (Do they have any interesting projects and are unique? Or interesting life experience? Or writing? that demonstrates a different point of view?)
    • We care about impact, not meaningless metrics - Relate to business objectives and team impact
    • 1 pagers are great but if you have more to say 2 is fine