Let's Chat

Tools & Workflow • LLM

Developers to Follow in the LLM-Powered Ecosystem

A curated list of developers worth following if you're serious about AI-assisted development—covering the Pi/OpenClaw ecosystem, essential reads, and the creators who are pushing boundaries.

6 min read
Focus LLM-Powered Development
Ecosystem Pi / OpenClaw / Python
Audience Developers embracing AI-assisted coding
Format Curated resource list

When I started taking AI-assisted development seriously, I quickly realized that the best insights don't come from hype-driven think pieces—they come from developers who are actually building with these tools daily.

After spending time exploring the LLM-powered developer ecosystem, I've curated a list of developers I keep coming back to. These are the people pushing boundaries, sharing real workflows, and documenting their experiments with radical transparency.

Developers I Follow

Starting with the core references I use, these are the developers at the center of the ecosystem I'm invested in:

Armin Ronacher — The prolific creator behind Flask, Ruff, and uv. His blog lucumr.pocoo.org is essential reading for anyone doing agentic coding.

@mitsuhiko @mitsuhiko

Mario Zechner — Creator of Pi, the coding agent that powers OpenClaw. His technical deep dives on marioslab.io cover everything from game development to AI tooling.

@badlogicgames @badlogic

Peter Steinberger — Developer of OpenClaw (formerly Clawd/Moltbot), which became the fastest-growing GitHub repository ever with 247K+ stars in two months. His blog steipete.me is packed with AI development workflow insights.

@steipete @steipete

Andrej Karpathy — AI researcher and founder of Eureka Labs (AI-native education). Former Director of AI at Tesla, founding member of OpenAI, and architect of Stanford's first deep learning class (CS231n). His YouTube channel has excellent lectures on LLMs, and his bearblog posts are essential reading.

@karpathy @karpathy

The Pi / OpenClaw Ecosystem

Since I'm using Pi (and tools built on top of it like OpenClaw), these are the developers shaping the ecosystem:

OpenClaw (by Peter Steinberger)

OpenClaw uses Pi as its core coding engine—an autonomous AI agent that connects to messaging platforms and gets things done. It went viral in early 2026 and became the fastest-growing GitHub repository ever. Peter joined OpenAI in February 2026.

oh-my-pi (by can1357)

A feature-rich fork of Mario's Pi with enhancements like AI-powered git commits, Python/IPython kernel integration, LSP support for 40+ languages, and time-traveling streamed rules. I've been exploring this for an IDE-like experience.

Essential Reads & Talks

These posts and talks represent the intellectual backbone of the agentic coding movement:

Pi: The Minimal Agent Within OpenClaw by Armin Ronacher — Why Armin switched from Claude Code to Pi.

What I learned building an opinionated and minimal coding agent by Mario Zechner — The detailed story of building Pi from scratch.

Agent Design Is Still Hard by Armin Ronacher — Technical deep dive on building agent loops, caching, and reinforcement.

A Year Of Vibes by Armin Ronacher — His 2025 transformation to agentic coding.

My Current AI Dev Workflow by Peter Steinberger — Peter's terminal setup and workflow for AI-assisted development.

OpenClaw, OpenAI and the future by Peter Steinberger — Peter's announcement joining OpenAI.

Agentic Coding: The Future of Software Development with Agents — Armin's 37-minute talk on his agentic coding journey.

How I use LLMs to help me write code by Simon Willison — Practical guide to LLM-assisted coding from the creator of Datasette.

2025 LLM Year in Review by Andrej Karpathy — Essential wrap-up covering RLVR, "jagged intelligence," and the state of vibe coding.

Andrej Karpathy: Education Meets AI

Karpathy occupies a unique position in the AI landscape—equal parts researcher, educator, and storyteller. His work bridges the gap between cutting-edge AI research and accessible explanations.

Key Projects

  • LLM101n — "World's best AI course" — an LLM-powered course to teach AI fundamentals from scratch
  • micrograd — Tiny autograd engine in ~100 lines of Python; great for understanding backpropagation
  • ConvNetJS — Deep learning library written entirely in JavaScript with browser-based demos
  • YouTube Channel — Lectures on LLMs, neural networks, and AI education

Essential Perspective

Karpathy's 2025 Year in Review introduced several influential concepts:

  • RLVR (Reinforcement Learning from Verifiable Rewards) — How training against objective reward functions is reshaping LLM development
  • "Jagged intelligence" — The observation that LLMs are simultaneously genius polymaths and confused grade schoolers
  • "Summoned ghosts" — LLMs as entities optimized under entirely different constraints than biological intelligence
  • Vibe coding — The shift where code becomes free, fleeting, and discardable

His core thesis: "We haven't realized even 10% of what today's models can do."

What These Developers Share in Common

Looking across this list, a few patterns emerge:

  • Prolific open-source creation — Frameworks, tools, libraries built out of necessity
  • Active technical blogging — In-depth content, not just quick takes
  • Early adopters of LLM tools — Building with agents, not just using them
  • Multi-language expertise — Python, Rust, Swift, C—whatever the problem demands
  • Conference speakers — Sharing knowledge at scale
  • Writing about their workflows — Radical transparency about what works and what doesn't

If you're looking to thrive in the AI-assisted development era, these are the people I follow.

Let's talk

Building with AI-assisted tools and want to swap notes? I'm always interested in hearing about workflows and developer setups.

Get in touch

Topics

LLM AI Developer Resources Productivity
← Back to Blog