Weyland Joyner

Functional Programming in Python, Part II: Intervals

Python

programming

functional-programming

This is a continuation of my series on Python programming from a functional perspective. In the first post we looked at sliding window problems, here we’ll deal with intervals.

Functional Programming in Python, Part I: Sliding Windows

software

programming

functional-programming

The default way of writing Python in the wild, particularly for data structures and algorithms style problems, involves starting with loops and conditionally updating one or more independent state variables based on multi-line control flow.

Building a FineWeb-style dataset for LLM pretraining

LLMs

data

Python

Let’s start by building a web crawler that returns a sitemap, then we’ll use trafilatura to deal with retrieving and deduplicating content.

Deep Learning for Biology, Part I: Protein Language Models

biology

LLMs

deep-learning

This notebook is based off of Charles Ravarani and Natasha Latysheva’s excellent recent O’Reilly book, Deep Learning for Biology.

RL Basics: OpenAI’s Gymnasium and the CartPole environment

reinforcement-learning

programming

import gymnasium as gym
import torch
import torch.nn as nn
import torch.optim as optim
from torch.distributions import Categorical
import numpy as np
import matplotlib.pyplot as plt

[Building Agents, 3] I Guess Handoff is the Most Important Feature

agents

context-engineering

ai

The point of an agent is that it should have a tight, simple core loop for tool calling and returning a response to the user. At this level the tool call, or sequence of tool calls, is the unit of work. Once you get above this level of abstraction, you’re looking at the thread itself as the unit of work, and here’s where you get into some design decisions within the agent harness. What should we do to keep our threads short and focused?

[Building Agents, 2] Treat Agent Threads like Unix Tools

agents

context-engineering

ai

In other words have them do one thing, well, and succinctly.

[Building Agents, 1] Getting Started

agents

context-engineering

TypeScript

I ended up porting Thorsten Ball’s post on building a basic coding agent to Typescript¹. At first I was using Ink for a React component-based terminal UI, but that got restrictive. Fortunately I found Mario Zechner’s Pi agent project.

Unleashing the Agents on Github

software

agents

“I wish I could see the prompts that led to changes”

Kernighan & Pike & Unix & Agents

software

agents

“Every Unix Command Becomes a Startup”

More on Context - Steve Yegge’s Beads

software

agents

Steve Yegge has a great post about Beads, his agent-first issue tracker. He goes into reasons a plain markdown tracking system doesn’t work at scale, and his findings shed some light on how agents approach their work with limited context.

Digital Twin Universe(s)

software

agents

StrongDM introduces the idea of a Digital Twin Universe to validate end-to-end “scenarios” for software built (they prefer the term “grown”) in “dark factories”¹ by agents.

Thoughts on Context Engineering

software

agents

“Think about context windows as a Commodore 64 with a small amount of memory”

Ralph Wiggum, Gas Town, etc

software

ai

Steve Yegge identifies 8 stages of AI-assisted coding adoption.

Kleppmann on Local-First

software

distributed-systems

This is Martin Kleppmann, author of Designing Data Intensive Applications, talking about Local-First Software.

Carl Bosch’s Nobel Prize Speech

books

history

How did you go about productizing a chemistry experiment in the early 20th century?

a16z on Forward-Deployed Engineers

software

consulting

There’s been a good deal of buzz about FDEs lately. That’s a job that I’ve been doing under some alias since the mid-2010s.

[DuckDB] The Lost Decade?

software

data

This fascinating post confirms some hunches I started to have in the late 2010s when everybody was overengineering their ETL pipelines.

On Writing

writing

meta

Why I’m starting this blog