Lisa Ross lisaross

profile: lr format: blog generated: 2026-04-15T00:00:00Z voice_match: 92 word_count: 1250 topic: "Getting comfortable delegating is going to be crucial for individual contributors as they use AI agent teams, mapping proven management delegation competencies to what every worker now needs" red_thread: "The skills managers have spent decades mastering (trust, clear direction, feedback loops, letting go of perfection) are exactly the skills every worker now needs for AI agent teams, and they're well-studied, not new." goal: "Readers recognize that delegating to AI agents isn't an unprecedented challenge. It maps directly to proven management competencies, and they can immediately apply these frameworks to become better agent managers." citations:

The Loop Is the New Way of Working

How autonomous experiment loops are replacing one-shot AI interactions — and why the skill that matters most is knowing what to measure.

The Shift Nobody Saw Coming

We spent 2024 and 2025 obsessing over prompts. How to write better prompts. How to structure conversations. How to get the best single response from an AI model.

Stochastic Consensus & Multi-Agent Debate: Quality Improvement Patterns for AI Agents

Source / Synthesized Research / 2026-03-28 References: Irving et al. (2018) "AI Safety via Debate" (OpenAI); Du et al. (2024) "Improving Factuality and Reasoning through Multiagent Debate" (ICML); Li et al. (2024) "Improving Multi-Agent Debate with Sparse Communication Topology" (EMNLP); Choi et al. (2025) "Debate or Vote?" (NeurIPS Spotlight); MindStudio Blog (2026); AutoGen Design Patterns (Microsoft); LangGraph community implementations

Summary

When a single AI agent is unreliable on hard problems, running multiple agents and having them argue, vote, or compete often improves output quality substantially. Two distinct patterns have emerged from research: stochastic consensus (independent parallel solvers + aggregation) and multi-agent debate (adversarial or collaborative argumentation to stress-test conclusions). Both stem from the same insight — that diversity of reasoning paths reduces syste

AI Infrastructure Demystified: A Level-Up Guide

For: Technical users who understand AI agents and want to understand WHY things cost what they cost, and HOW to make smarter model and infrastructure decisions.

The Restaurant Metaphor

Imagine a high-end restaurant. This metaphor will carry us through the entire guide.

profile: lr format: newsletter topic: The Minimal Viable Stack—where to start if starting over red_thread: Start minimal and learn how things work together. Iterate as problems repeat. goal: Reader has a concrete starting architecture and philosophy for building up—not a copy of someone else's setup. generated: 2026-01-18T15:30:00Z voice_match: 92 word_count: 1750 status: complete

profile: lr format: newsletter topic: Context Intelligence—the skill of reading what kind of work you're facing and knowing how to approach it red_thread: The question isn't "should I plan?"—it's "what kind of work is this?" Context intelligence is the skill that answers that. goal: Reader understands that context intelligence (recognizing work types, sensing complexity, knowing when rigor pays off) is the meta-skill underlying effective Claude Code workflows, and knows how to apply it in practice. generated: 2026-01-15 voice_match: TBD word_count: TBD status: draft-v2

profile: lr format: newsletter topic: When planning pays off vs when iteration wins in Claude Code workflows red_thread: The question isn't "should I plan?"—it's "what kind of work is this?" goal: Reader understands the distinction between recurring systems (where rigor pays off) and one-off work (where iteration wins), and knows how to recognize which they're facing. generated: 2026-01-15 voice_match: 91 word_count: 1750 status: complete

Agent Preflight Prompt

Paste this after your task brief to surface blockers before long autonomous sessions.

The Prompt

## PREFLIGHT CHECK - Do Not Start Work Yet

Before implementing anything, conduct a thorough preflight analysis. Surface every potential blocker NOW so we can resolve them before you begin.

name	codify
description	Capture a discovery from iteration and turn it into reusable infrastructure
argument-hint	[what you discovered]
allowed-tools	Read, Write, Edit, Glob, Grep, Bash, AskUserQuestion

Codify Command

Capture what you learned through iteration and turn it into reusable infrastructure.

	#!/bin/bash
	# Claude Code plugin sync - run daily alongside brew update

	# Update marketplace catalogs
	claude plugin marketplace update

	# Update each installed plugin
	jq -r '.enabledPlugins // {} \| keys[]' ~/.claude/settings.json 2>/dev/null \| while read plugin; do
	claude plugin update "$plugin" 2>/dev/null
	done