Library
Field notes for people building AI agents.
Field notes for people building AI agents.
Field notes for people building AI agents.
Supervisor workers, pipelines, and when one agent is enough.
Chunking, embeddings, hybrid search, and reranking for domain-specific agent knowledge.
How agents call APIs, use MCP servers, and stay secure when acting on user data.
Traces, evals, and dashboards that tell you why an agent failed before customers complain.
Week-by-week plan from idea to paying pilot customer.
Next.js, Python, LangGraph, Vercel, Supabase — proven stacks for solo founders and small teams.
Unit tests for tools, evals for prompts, and staging strategies that catch regressions.
Streaming UX, queues for long jobs, and handling concurrent users without blowing up costs.
Approval flows, edit-before-send, and audit logs that make agents feel safe to adopt.
Per-seat, per-task, outcome-based, and hybrid pricing with real examples.
Outbound, communities, partnerships, and demos that convert for agent products.
Headlines, ICP statements, and differentiation that cuts through 'AI wrapper' noise.
Self-serve vs. sales-led motion, security questionnaires, and procurement timelines.
GDPR basics, BAAs, audit logs, and what to put in your privacy policy.
Keep customers successful after the sale so churn does not kill your business.
Workflow depth, proprietary data, and integrations that copycats cannot replicate in a weekend.
When $49 databases and $5k MRR beat a pitch deck, and when to take VC money.
Industry-specific constraints, buyers, and agent ideas that fit each vertical.
Task success rate, cost per task, time saved, and the KPIs that prove ROI to customers.
Understand the difference between chatbots, copilots, and autonomous agents — and why buyers pay for agents that complete work.
Tokens, context windows, temperature, and model selection — the basics every agent founder needs before writing code.
Who is buying agents, what budgets look like, and which verticals are moving fastest.
A scoring framework for market size, build difficulty, defensibility, and willingness to pay.
System prompts, tool schemas, and guardrails that keep agents on-task in production.
When to remember, what to forget, and how to personalize without leaking data.