Staff engineer turned writer
Hadassah Stein
Hadassah Stein writes the Review's product-focused work — teardowns of agentic orchestration platforms, comparative reviews of frameworks, and the occasional long-form interview with a project's maintainers. She spent six years as a staff engineer on a multi-agent platform team before moving to writing full time. She brings a builder's impatience to product copy and a reviewer's patience to source.
Her column "Field Notes" runs whenever she has read enough code to have something to say, which is usually every other week. She writes from somewhere in Central Europe and is the person to contact for invitations to walk a journalist through your stack — provided you are willing to be reviewed in public after.
Hadassah does not invest in any of the companies she covers and has not held equity in any orchestration vendor since her transition to writing. She is on LinkedIn at /in/hadassah-stein-fieldnotes.
Joined
Pieces published
Contact
Covers
Pieces by Hadassah Stein
RSSKarpathy on Slop vs the Revenue Tape
Andrej Karpathy calls agentic AI 'slop.' Claude Code is at a $2.5B run-rate. Cursor is at $2B ARR. Devin's ARR went from $1M to $73M in nine months. Both readings are honest. The Review's attempt at why.
Comparative Notes: AutoGen, LangGraph, CrewAI, Letta, and Sema4.ai
Five products at five different layers of the agentic stack. A qualitative teardown across topology, surface, scope, and the kind of user each one is for.
Patterns in Agentic-OS Architecture: What the Bundles Are Actually Doing
A reading of four bundled agentic platforms — Sema4.ai, MultiOn, Adept's ACT family, and Web4OS — and the architectural patterns that recur across them: supervisor topologies, card-stream surfaces, credit schedulers, and canonical-host integrations.
LangGraph 0.3, AutoGen 2.0, CrewAI Studio — Picking When Each Wins
A working-engineer's selection guide across three of the most-deployed open-source agent frameworks, against the public 2025 release notes and the production patterns each one rewards.
OpenHands — A Deployment Notebook
A working engineer's notebook on deploying OpenHands, the open-source coding agent that grew out of OpenDevin, against a real codebase. The runtime topology, the model adapters, the gotchas, and the production-readiness ledger.