---
title: Writing — Knitli
description: Current work on classification markings and an earlier archive of Knitli writing.
url: "https://knitli.com/writing"
type: static
generatedAt: "2026-09-01T17:07:33.503Z"
---

Writing

# What we publish, and who says it

News, updates, and tips about our current efforts to make the classification system move at the speed of decision. Thought and opinion pieces on classification and classification policy from our founder, Adam. There is also an archive of earlier Knitli writing, from our early days building developer tools.

Writing

## What we’re working on and thinking about
2026-08-25
## [I could not mark my own documents correctly](/writing/i-could-not-mark-my-own-documents-correctly)

Eighteen years inside the system, months deep in the problem, unhurried and motivated. My human brain could not reliably get it correct. Marque was already more right than me.

**Adam Poulemanos**
2026-08-10
## [Good enough wasn't](/writing/good-enough-wasnt)

I went quiet in January, published one post in April I never told anyone about, and spent four extra months on something that already mostly worked. Here is what I was doing and why it took that long.

**Adam Poulemanos**

Archive

## Earlier writing

These posts predate Knitli’s current classification-marking focus.
2026-04-16
## [Ghost Repos: The Most Dedicated Projects Nobody's Ever Heard Of](/writing/ghost-repos-the-most-dedicated-projects-nobodys-ever-heard-of)

I queried 5TB of GitHub activity data to find repositories with hundreds of commits, tests, CI pipelines, and releases — and almost no stars. What I found says a lot about how we discover software.

**Adam Poulemanos**
2026-01-21
## [Local Queries; Cloud Quality. Voyage-4's Shared Embedding Space](/writing/local-queries-cloud-quality-voyage-4s-shared-embedding-space)

VoyageAI's new models are incredible, and create opportunities for much cheaper and better vector search.

**Adam Poulemanos**
2026-01-12
## [AI Killed One Open Source Business Model, Not Open Source](/writing/ai-killed-one-open-sorce-business-model-not-open-source)

Content driven open source strategies are failing in the AI era. The rules have changed, but open source is far from dead.

**Adam Poulemanos**
2025-12-29
## [Context Engineering: How We Work Around the Goldfish Problem](/writing/context-engineering-how-we-work-around-the-golfish-problem)

Learn how context engineering helps manage LLM context windows, the limitations of summarization, and strategies to optimize AI coding tools for better performance.

**Adam Poulemanos**
2025-12-15
## [Welcome to the Knitli Blog](/writing/welcome-to-knitli-blog)

Introducing our blog where we share insights on building better developer tools, AI context management, and the future of software development.

**Adam Poulemanos**
2025-10-08
## [From Intelligence Expert to AI Business Leader: A Surprising Path](/writing/from-intelligence-expert-to-ai-business-leader-a-surprising-path)

How an 18-year career in intelligence leadership shaped my journey to founding an AI startup focused on democratizing AI for all users. The same principles of removing barriers and enabling people to do their best work apply—just in a different domain.

**Adam Poulemanos**
2025-10-06
## [Tree-Sitter Grammars Explained: Leveraging Data for Clarity](/writing/tree-sitter-grammars-explained-leveraging-data-for-clarity)

A deep dive into Tree-sitter grammar structures, clarifying confusing terminology and demonstrating how empirical analysis across 25 languages led to a clearer, more intuitive understanding of parsing concepts.

**Adam Poulemanos**
2025-09-25
## [Context and Context Windows: What You Need to Know](/writing/context-and-context-windows-what-you-need-to-know)

Understanding how large language models (LLMs) handle context is crucial for effective AI usage. This post explores context windows, their limitations, and strategies for managing context in AI applications.

**Adam Poulemanos**
2025-09-25
## [I Let Four AI Code Reviewers Fight Over My PRs](/writing/i-let-four-ai-code-reviewers-fight-over-my-prs)

Why sometimes obnoxious and repetitive AI code review is better than no code review.

**Adam Poulemanos**
2025-09-18
## [Understanding Tokens: What They Are and Why They're Important](/writing/understanding-tokens-what-they-are-and-why-theyre-important)

This article explores the concept of tokens in AI, highlighting how they differ from words and are the fundamental units processed by AI models like ChatGPT. It explains the cost implications of processing tokens due to the intensive computations required on GPUs, making them a significant cost driver in AI operations. Understanding tokens is crucial for grasping AI mechanics, pricing, and efficiency. Part 1 of Knitli’s 101 introductions to AI and the economics of AI.

**Adam Poulemanos**

[Some doors open only when a customer asks. Tell us.](/sponsor)