Intel / video

Morgan Stanley's ALPHALAB: Multi-Agent Research Across Optimization Domains — Brendan Rappazzo

Source: youtube.com
Author: AI Engineer
Date: Published Jul 29, 2026

Why it matters

It shows what an agentic research loop looks like when it has to survive real institutional scrutiny — a strategist agent proposing experiments, worker agents writing code, configuring backtests/evals and running significance tests, with a purpose-built chosen over off-the-shelf frameworks so reasoning stays observable. If you're designing agents that must produce verifiable results rather than plausible text, the strategist/worker split and the emphasis on building the verifiable environment are directly reusable patterns.

Terms in this piece · Glossary

multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.

Key quotes

“it really felt possible for the first time with like with Opus 4.5 and with, you know, these harnesses like Claude Code and Codex”

“LLMs aren't malicious, but they can make, you know, very silly mistakes and if you're optimizing against a bad a bad eval, the whole thing kind of falls apart.”

“this is a lesson I learned over and over, like monthly, working with these models, you have to start with good eval.”

“to me evals and environments are the same thing. It's just you train in environments.”

“I think we're already seeing it with GLM 5.2, and so I really think all of your value as like an enterprise or human expert comes from building environments.”

Read the source www.youtube.com