wayanjimmy
ENID

Newsletter #34

· 2026-W39 · 2026-09-21 to 2026-09-27

Refactoring Hermes with 1,393 agents #

— Nous Research

tl;dr: Hermes Agent refactored its own codebase using 1,393 subagents — a live experiment in orchestrating massive parallel agent fleets against one repo. The write-up covers how work was decomposed, coordinated, and merged without the swarm trampling itself. A rare behind-the-scenes look at what “agents maintaining agents” actually costs and yields.

AIAgents Refactoring Orchestration Engineering #

Software with Taste with Geoffrey Litt | AM Podcast #7 #

— AM Podcast

tl;dr: Geoffrey Litt (design engineer at Notion) argues that with agents writing the code, understanding is the new bottleneck — he has agents write him teaching essays with a quiz at the end, and won’t merge until he passes. He connects malleable software to coding agents: AI is a huge force multiplier for end-user programming, but app-store distribution and closed platforms are still built for a world where nobody could code. His north star for customizable tools is the spreadsheet’s gradual learning curve, and his antidote to shipping slop is building mechanical speed limits into your own workflow.

MalleableSoftware DesignEngineering AIAgents Talk #

We are all product engineers now #

— Laurie Voss, seldo.com

tl;dr: Agents are eating the software development lifecycle from both ends, collapsing the wall between “engineer” and “product person.” When anyone can ship a working prototype in an afternoon, the scarce skill shifts to knowing what to build and how to judge it. A sharp essay on how job boundaries in software are dissolving rather than disappearing.

ProductEngineering AIAgents Career Software #

Native is now the future of mobile at Shopify #

— Shopify Engineering

tl;dr: Shopify reversed course on cross-platform frameworks and went back to native iOS and Android — because coding agents changed what it costs to build and maintain two native codebases. The economics that made “write once, run everywhere” attractive simply stopped holding when agents could do the platform-specific work. A landmark data point in the cross-platform-vs-native debate.

Mobile Nativeapps Shopify Engineering AIAgents #

React for systems engineers #

— TJ Zhang

tl;dr: Explains React through systems concepts — state ownership, derived data, and invalidation — instead of component trees and hooks trivia. If you come from databases, distributed systems, or backend caching, this maps the mental model onto what you already know. The clearest bridge yet between systems thinking and frontend reactivity.

React Frontend SystemsEngineering MentalModels #

Using OKF to provide long term context to your agent #

— iodigital TechHub

tl;dr: Your agent forgets you after every compaction — this article proposes the Open Knowledge Format (OKF) as a durable, structured context file the agent reads back each session. Instead of prompt-hacking memory, knowledge lives in a versioned artifact the whole team can review. A practical pattern for making agent context survive session boundaries.

AIAgents Memory Context KnowledgeManagement #

Introducing Roast: Structured AI workflows made easy #

— Shopify Engineering

tl;dr: Roast is Shopify’s convention-oriented orchestrator for multi-step AI workflows — define steps, inputs, and outputs in a structured format instead of glue scripts. It targets the messy middle of agent work: reliable pipelines with retries, approvals, and human-in-the-loop checkpoints. Worth studying if you’re productionizing anything beyond a single prompt.

AIWorkflows Orchestration OpenSource Shopify #

Scenarios for our Economic Future #

— Anthropic

tl;dr: Anthropic’s economics team models several scenarios for how advanced AI could reshape labor, productivity, and income distribution. Rather than predicting one outcome, they map the variables that determine which world we get. A sober, numbers-first framing for a debate that’s mostly vibes.

AI Economics Labor Market Forecasting #

Subscribe