The field notes
The Daily AI Drop - Evening Edition, October 19, 2026
Podium's agent scaling playbook, the post-password fix, Thomson Reuters legal AI, Greg Isenberg on Meta Muse money, Perplexity's new decider model, AI rewriting PS5 games, and cooking with Codex.
Agents go enterprise, passwords get replaced, and AI starts rewriting video games. Tonight's edition covers eight new videos from the channels you follow.
Podium's Playbook for Scaling AI Agents Org-Wide — LangChain
Walker Ward of Podium explains why making LangSmith the default agent platform across the whole company was a bigger unlock than any single feature. Teams inherit observability automatically, and every new agent follows patterns that are already known to scale. If you own agents at your company, this is five minutes well spent.
Your Password Can Be Hacked: Here's the Fix — pvergadia
Priyanka walks through how passwords authenticate users, why multi-factor authentication can still fall to phishing and man-in-the-middle attacks, and how attackers steal and replay session cookies. The fix: device-bound cryptographic authentication — private keys tied to your physical device, bound to a specific session — so a stolen cookie is useless. A jargon-free explainer of the post-password future.
Thomson Reuters on Building Trust in Legal AI — OpenAI
Camillo Speroni of Thomson Reuters sets the bar for legal AI: output lawyers can verify, cite, and put their own name on. He discusses Thomson Reuters' work with OpenAI to bring HighQ and CoCounsel Legal into ChatGPT Enterprise, connecting the tools lawyers already rely on. Legal AI is moving from experiment to daily workflow — and trust is the whole ballgame.
Meta Muse Will Create 1,000 Millionaires — Greg Isenberg
Greg argues you can use Meta Muse to mine Facebook groups, Marketplace, and Instagram to match buyers with sellers — arbitrage as a business. The idea came after the night-nurse agency he hired admitted it finds candidates in Facebook groups. Meta owns the social graph and the marketplaces; an agent living on top of that graph will mint brokers and micro-businesses at scale.
Perplexity's pplx-decider-v1.2: Is It Worth Using? We Tested It Live — Fahd Mirza
Fahd tests Perplexity's new decision model live — the follow-up to pplx-decider-v1.1-27b, built on the same Qwen3.8-27B backbone — with no editing, so you see the hits and the misses. Routing (which model handles which query) is quietly one of AI's most important problems, and this is the honest benchmark for whether the upgrade is worth it.
AnyPS5 — This AI Rewrites a PS5 Game to Run on Your PC — TheNextNewThingAI
AnyPS5 is an open-source tool that uses AI to read a PlayStation 5 game's files and convert them into a version that runs natively on Windows or Linux. It isn't an emulator — it actually rewrites the game. A signpost for AI-powered reverse engineering, raising questions about game preservation and the legal gray zone of converting games you own.
REA use cases you'll love — TheNextNewThingAI
Andrew Warner tests REA (Reverse Engineer Anything), showing how coding agents can inspect closed-source apps and learn how features work without the source code. Incredible for security research and learning — and unsettling for any business model that depends on keeping software's inner workings secret.
Cooking with Codex: Ep. 1 — OpenAI
A charming new series where an OpenAI dot plans a ramen recipe and organizes prep in ChatGPT Space, with GPT Live guiding the cooking. The real news is under the hood: Codex cloud environments that keep tasks running after you close your laptop, plus new Codex CLI features — voice input, agent management, and forking conversations into separate worktrees.