PAI Upgrade Sources - What do you use?

[jdrolls]

Hey all!

I was wondering if anyone would like to share the Upgrade sources you use for your PAI setup for the /PAIUpgrade skill. Would love to see what others have customized too, and what resources you might find helpful vs just a waste of tokens.

Here is what I currently use:

Blogs (all HIGH)

• Daniel Miessler — danielmiessler.com — AI worldview, PAI/agentic infra, security. PAI lineage upstream.
• Simon Willison — simonwillison.net — LLM tooling, prompt injection, Claude deep-dives. Daily.
• Hamel Husain — hamel.dev — evals, LLM-as-judge, error analysis, applied LLM ops.
• Eugene Yan — eugeneyan.com — LLM patterns, product evals, ML system design.
• Lilian Weng (Lil'Log) — lilianweng.github.io — agent architecture surveys, RLHF, test-time compute.

Newsletters (all HIGH)

• Latent Space — swyx + Alessio — AI engineer interviews, infra deep-dives.
• Interconnects — Nathan Lambert — post-training, RLHF, open-vs-closed dynamics.
• Smol AI News — swyx — daily Twitter/Discord/Reddit AI recap.
• Don't Worry About the Vase — Zvi Mowshowitz — agentic-coding series + weekly roundups.

RSS feeds / aggregators (all HIGH)

• HN search: LLM agent — hnrss.org agent-loop/harness discussion.
• HN search: Claude Code — hnrss.org Claude Code chatter (hooks, MCP, plugins).

GitHub trending (enabled — 14-day lookback, ≥50 stars, top 5 by stars)

• Queries: claude code, MCP server, AI agent framework, LLM agent harness, autonomous agent, personal AI assistant, algorithmic trading bot, prompt engineering tooling

youtube-channels.json (8 channels)

• Unsupervised Learning — @unsupervisedlearningpodcast — HIGH — AI+security, Fabric, PAI lineage.
• IndyDevDan — @indydevdan — HIGH — Claude Code mastery, sub-agents, parallel orchestration.
• NetworkChuck — @NetworkChuck — MEDIUM — networking/IT, sporadic AI tooling.
• Cole Medin — @ColeMedin — HIGH — Claude Code deep-dives, hooks, MCP, parallel agents.
• AI Jason — @AIJasonZ — HIGH — production AI eng, evals, multi-agent orchestration.
• Hamel Husain — @hamelhusain7140 — HIGH — LLM evals doctrine, LLM-as-judge.
• AI Engineer — @aiDotEngineer — HIGH — AI Engineer Summit talks, agent reliability.
• Sam Witteveen — @samwitteveenai — MEDIUM — LangChain/LangGraph, agent patterns.

[stephtara]

Ah ah I just have it go through this repo and issues and PRs and discussions, and have tried creating a "PAI expert" agent that tries to separate what is "original PAI" and what are our homegrown foxes and improvements. I also have a local clone of the PAI code for reference.
Maybe with a wider net I'll get better results, sometimes I feel like my DA and I are like "the blind leading the blind" 😰

[jaylarr176]

sometimes I feel like my DA and I are like "the blind leading the blind" 😰

Oh yeah....truer words...