Paper notes on LLaMA covering its core strategy — train smaller models for longer on publicly available data, optimized for inference — the result where a 13B model beats 175B GPT-3, the RMSNorm/SwiGLU/RoPE architecture changes, the 2048-A100 training scale, and how it kicked off the open LLM ecosystem.
Firecrawl is a data API that turns any website into clean, LLM-ready Markdown or structured JSON. It orchestrates in TypeScript, cleans in Rust, converts to Markdown in Go, races several scrape engines in a waterfall, and drives crawls with a hand-built Postgres queue. We analyze the scrape/crawl/map/search/extract pipelines and contrast them with Browser Use as "two ways to feed the web to LLMs."
SkillSpector is NVIDIA's security scanner for AI agent skills. It vets a skill before install to find prompt injection, data exfiltration, and malicious code. We analyze its structure — a LangGraph map-reduce graph that fans out to 25 analyzers (static patterns, AST, taint, YARA, MCP, LLM semantics) and reduces them to a single risk score — against Superpowers and ponytail, which inject skills.
ponytail packs a single discipline — the "lazy senior dev" laziness ladder — into one SKILL.md and ships it to 16 agents via every mechanism (skill, hook, command, MCP, plugin), proving a ~54% code cut with a real agentic benchmark. We analyze it against Superpowers, and the idea that a skill is a bundle of discipline, distribution, and measurement rather than code.
Browser Use is a Python agent in which an LLM drives a real browser. It pre-chews the page into an indexed list of interactive elements for the LLM, drives the browser via CDP instead of Playwright, and runs the session with an event bus and watchdogs. We look at why this fits LLMs better, picking up from the earlier Playwright analysis.