
Meta is telling engineers to handle Claude Code and Codex with care
THE SO WHAT
Tightening internal rules on Claude Code and Codex shows how seriously large platforms now treat inadvertent model distillation and IP leakage—coding assistants are no longer “just tools,” they’re potential data exfil paths. Any org building proprietary models should adopt similar policies this quarter: define which external AIs are allowed, where, and with what redlines around sensitive code and datasets.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIThe AI job paradox and the missing link in productivity gains
The constraint isn’t model quality, it’s org design — AI that “improves efficiency” but doesn’t change spans of control, handoffs, or decision rights will never show up in P&L. Treat every AI initiative as a workflow rewrite with explicit role changes, or assume the productivity upside is mostly narrative.
Is AI causing layoffs? This report says it's complicated.
If the companies spending most on AI are also hiring faster, AI is behaving like prior automation waves — a reallocation engine, not a pure headcount cut. For operators, the risk is less “mass layoffs” and more talent mix whiplash if you’re not explicitly planning which roles you’ll scale and which you’ll redesign.
Applied AISAP hands AI product oversight to its CEO and COO in a reshuffle
Putting AI product oversight directly under the CEO and COO — and doing a second reorg in one year to do it — is a clear signal that AI is now a board-level product question, not an R&D lane. If you’re a large vendor without similar executive ownership, expect customers to start asking who at the top signs off on your AI roadmap.
Applied AIAfter Harvey, vertical AI’s next $10B winner might be in agriculture
Legal, healthcare, and customer service already have vertical AI players valued from $11 billion to $15 billion — agriculture is the next obvious candidate because it’s data-rich, margin-thin, and workflow-fragmented. If you build in ag, assume the window for generic tools is closing and start mapping where a domain-specific model could own your value chain.