Madison's Morning Memo

Archives
Log in
Subscribe
October 28, 2026

Lenny's Personal Agent Showdown and GitHub's Agent-Native Developer Data

AI PM Research

Wednesday, October 28, 2026

Lenny's Personal Agent Showdown and GitHub's Agent-Native Developer Data

Lenny ran the first real head-to-head benchmark of personal AI agents (Grok, Dot, Muse, Claude Duet) across actual tasks, GitHub's Octoverse data shows developers are now shipping more commits and more PRs per contributor thanks to agents, and Noah Shinn keeps shipping Instinct features (budgets, Action Button, iMessage groups) faster than almost anyone can track, which together paint a pretty clear picture: the agent autonomy question from our open-questions file is being answered in public, one shipped feature at a time.


Instinct - Shipping Autonomy Controls Faster Than the Market Can Evaluate Them

Noah Shinn shipped three distinct Instinct features in a 48-hour window: a Stripe/Link-powered spending budget so users can authorize Instinct to make purchases up to a set limit, Action Button integration on iPhone (press-and-hold to task Instinct without leaving your current app), and expansion of Instinct-in-group-chats to all iMessage users, including people who don't have the app.

GitHub - Octoverse 2026 Shows Commits and Pushes at All-Time Highs as Developers Build With Agents

GitHub's annual Octoverse report for 2026 shows commits and pushes reaching new highs, with the headline framing being that developers are shipping more because they're building with agents, not despite them. This lands alongside GitHub Universe 2026 coverage describing changes "at every layer of the stack."

Lenny Rachitsky - Benchmarking Personal AI Agents Across 9 Real Tasks

Lenny Rachitsky ran Grok's Bot, Dot, Meta Muse, and Claude Duet through 9 real tasks and scored them against each other, explicitly framing the results as "which one runs my work and which one runs my life." This is the first comparative, task-based evaluation across the major personal-agent entrants that isn't coming from one of the vendors themselves.

Also today

  • Simon Willison: "The October inflection point for vibe game development" flags a shift in AI-assisted game dev quality/quantity this month
  • Vercel: Vercel Agent sessions now cost 67% less, plus Agent can validate alerts before paging, a meaningful pricing and reliability change for agent-based ops tooling
  • Notion: "Agents, rebuilt to do more for less" — Akshay Kothari details smarter model routing, long-running agents, bring-your-own-AI, and a new credit system (recent)
  • Meta: Muse is now available in Singapore, continuing its country-by-country rollout as "the world's first personal AI agent built for everyone"
  • Anthropic: Early progress update on using Claude to accelerate drug development, with six research partnerships underway
Read the full update →

AI PM Research · Daily updates on AI products and strategy

Don't miss what's next. Subscribe to Madison's Morning Memo:
← Newer GitHub Universe, Notion's Credit Reset, and Agent Harness Design Older → Notion's AI Credit System and LangChain's Internal Agent Playbook
ai-pm-research.netlify.app
Powered by Buttondown, the easiest way to start and grow your newsletter.