Mar 8, 2026 · Weekly Roundup

AI Insights Weekly

Autonomous agents, agentic engineering frameworks, and real-world case studies — curated for builders who ship.

Highlight of the Week

Karpathy’s Autoresearch

Andrej Karpathy on X

A self-contained repo for an autonomous LLM training loop. An AI agent iteratively improves training code and hyperparameters based on human prompts, autonomously driving validation loss improvements. This is the “AI improving AI” loop made concrete and reproducible.


Agentic Engineering & Strategy

Harness Engineering as Cybernetics

George (@odysseus0z)

Harness engineering is the modern iteration of cybernetic control systems. The key insight: success relies not on agent capability, but on calibrating feedback loops with system-specific knowledge — architecture docs, linters, golden principles. This maps directly to how Claude Code skills work: the skill document is the feedback loop calibration.

Anthropic’s Official Skills Guide

0xMarioNawfal (@RoundtableSpace)

Anthropic published “The Complete Guide to Building Skills for Claude” — the foundational patterns behind the tools and frameworks released this week. If you’re building Claude Code skills, this is the canonical reference.

“Everything is Context” — File-System-as-Memory

Akshay (@akshay_pachaar)

Research formalizing the file-system-as-memory architecture: memory, tools, and knowledge stored as versioned files (Markdown/JSON) for traceability, auditability, and persistence. This is exactly the pattern OpenClaw uses — and it’s now getting academic validation.


Engineering & Reverse Engineering

Codex 5.4 Reverse Engineers a DOS Game

Ammaar Reshi (@ammaar)

Codex 5.4 autonomously reverse-engineered SkyRoads from a bare DOS executable. In 6 hours, the agent unpacked assets, disassembled the EXE, rebuilt the renderer, and recompiled a functional version in Rust. No human intervention.

office-agents SDK

Li Yang (@hewliyang)

Renamed open-excel to office-agents and published the SDK to NPM — a lightweight framework for building AI plugins for Office-like flows (Excel, PowerPoint, etc.).

SuperClaude Framework

Tom Dörr (@tom_doerr)

Open-source framework for Claude Code with pre-built commands and agentic structures for automating coding workflows.


AI Services & Business

The AI Services Playbook

Noah Epstein

A comprehensive guide on identifying high-value “boring” niches (HVAC, law firms, accounting) and positioning AI as an employee amplifier rather than a replacement. The framing matters — “amplifier” sells where “replacement” triggers resistance.

Google’s PaperBanana

Daily Dose of Data Science

An agentic framework that automatically generates publication-ready academic illustrations from methodology descriptions and captions. No manual design tools needed.


Case Studies

The “Larry” Case Study — $7k/Month Autonomous Business

Oliver Henry

An agent called “Larry” running on a local gaming PC autonomously manages a business — marketing, development, analytics — generating over $7k/month. This led to the creation of LarryBrain, a marketplace for verified AI agent skills.

Biological DOOM Agent

Mark Gadala-Maria

800,000 biological human brain cells in a petri dish have been interfaced with DOOM, exhibiting “goal-directed learning” by adapting and improving at the game in response to failure. Not practical, but conceptually wild.


Creative & Community

AI Cinema Production

PJ Ace

AI agents (Luma) significantly lower the barrier to producing high-end cinema adaptations — showcasing a Red Rising teaser that would have been impossible at indie budgets a year ago.

Evangelion UI Pack for OpenClaw

Scoopy Trooples

An Evangelion-inspired UI design skill pack for OpenClaw/Claude. Form meets function.


Themes This Week

ThemeSignal
Feedback loops > raw capabilityHarness engineering, skills guides, “Everything is Context” all point the same direction
Agents running businessesLarry generating $7k/month autonomously is proof-of-concept becoming proof-of-revenue
Reverse engineering at scaleCodex 5.4 on SkyRoads shows agents can now handle truly adversarial, undocumented codebases
Academic validation of builder patternsFile-system-as-memory, agentic skills — things builders discovered empirically are now getting formal treatment