KORDUROYMonday, September 21, 2026 · Run 2 · 4:07 PM ET
Claude Projects Levels Up, Meta Muse Opens to Developers, and the Real AI ROI Problem
4 stories from 1 source
TL;DR
- Anthropic redesigned Claude's projects so you can hand it several tasks at once and it delegates and assembles the work itself, starting in Claude Code.
- Meta opened Muse to developers via new Muse Connectors, as Muse holds the #1 free spot in Apple's App Store.
- 90% of companies use AI but only 6% report capturing real value from it — here's what the ones who do are doing differently.
AI Tips2
Run a personal follow-up system with an AI agent
- Connect the email, calendar, and messaging accounts you want monitored to a tool like Instinct.
- Tell it which relationships matter most (clients, partners, candidates, vendors, key colleagues) and define what counts as a dropped follow-up, e.g. an unanswered email after 3 days.
- Forward important threads to its inbox or add it to group threads so it can track decisions and outstanding actions.
- Ask for a daily or weekly brief showing who needs attention, why, and the recommended next action, and review before it sends anything.
Keep track of my important professional follow-ups. Watch conversations with [people/companies/clients] and flag anything I've promised to do, messages awaiting my reply, and threads that have gone quiet for more than [X days]. Prioritize by urgency and relationship importance. Draft follow-ups when useful, but do not send anything without my approval.
Source: Superhuman AI
Fix your AI spend before it becomes another line-item nobody can defend
- Build a usage dashboard with clear per-employee token caps so spend is visible, not just approved.
- Route routine, high-volume tasks to cheaper models and reserve frontier models for the work that needs them.
- Sort AI use cases into three buckets (daily work, autonomous agents, one-time strategic projects) and pick the most efficient model for each bucket rather than defaulting to one model everywhere.
Source: Superhuman AI
AI News2
Anthropic redesigned Claude's projects for multi-step work
You can now hand Claude several tasks at once, in any order; it outlines the project, delegates the work, and assembles the final product, building shared memory over time. It's rolling out in Claude Code first, with Claude to follow.
Why it mattersWorth testing on a real multi-part deliverable (a research-then-draft-then-review chain) before you assume you still need to manually sequence Claude's work.
Source: Superhuman AI
Meta opened Muse to outside developers
New "Muse Connectors" let companies link their services into Meta's personal agent, which is now the #1 free app in Apple's App Store; Meta Superintelligence Labs also published 50 sample Muse prompts.
Why it mattersA consumer AI agent platform reaching developer-ecosystem scale is a competitive signal worth flagging in any client conversation about assistant/agent strategy.
Source: Superhuman AI
Summarized from Superhuman AI (Sep 21)
Enjoyed this brief? Buy Korduroy a treat
