Top Story

Cowork Is Here. Everything Has Accelerated.

On April 10, Maury dropped a quiet message after listening to an Anthropic interview with Felix Rieseberg — the engineer who built Claude Cowork. His read: the Cowork roadmap was "one month — at most." Achal replied with two words: "Everything has accelerated."

"Listening to interview with Anthropic's Felix Rieseberg. The Claude Cowork roadmap is one month — at most." — Maury, Apr 10

What neither knew in the moment: Cowork had already gone generally available the day before, April 9, rolling out to all paid subscribers on macOS and Windows. Enterprise controls, group spend limits, a Zoom MCP connector, and usage analytics landed at GA. Felix confirmed in interviews that most of Cowork's own codebase was written by Claude itself — with engineers spending less time typing and more time directing.

For an agency building on this stack, this is the moment that changes the shape of what "a team" can deliver.

Build Stack

GitHub ships stacked PRs — and it matters for AI-built features

GitHub's native gh-stack hit private preview on April 13. Stacked PRs break large changes into a chain of small, reviewable layers — each merged atomically. Achal shared it with a note: it's "very interesting specially for big features or working with AI models to build large features but in a smaller chunked out manner." The data backs it: PRs under 400 lines have 40% fewer defects and get approved 3x faster.

Achal Apr 14

Adobe takes a swing at NotebookLM with Acrobat Spaces

Adobe launched Student Spaces on April 7 — a free AI tool that turns PDFs, links, and notes into flashcards, mind maps, quizzes, and audio summaries. No login required to start. It's a clear play against Google's NotebookLM. Kunal spotted it and shared it; the team gave it 3 fires.

Kunal Apr 7

1Password unlocks with passkeys

Rishav flagged 1Password's new passkey unlock — you can now open the app itself with a passkey instead of the master password. Small change, surprisingly satisfying UX shift.

Rishav Apr 14

Signal from the Feed

A week of fires and raised hands

Achal shared a string of tweets through the week — OpenAI, Google DeepMind, Google AI Devs — each pulling a fire or raised hands from the team. No commentary, which often says more than words. Shopify's tweet got two raised hands. Satish shared a Claude AI tweet that drew another. The cumulative effect of a team all watching the same feeds is its own kind of signal.

Achal Satish Apr 7–15

The Business of AI

The NYT got played — and so did everyone who forwarded it

Maury shared the NYT story on Medvi — a 2-person GLP-1 telehealth startup that claimed to be worth $1.8 billion, built with $20K and AI. The team reacted with open mouths and star-struck faces. The problem: the FDA had issued Medvi a formal warning letter six weeks before the Times ran it. No outside funding. No independent valuation. The company turns out to be a patient-acquisition layer over outsourced clinical and pharmacy infrastructure. The story sparked serious pushback. Worth reading the Techdirt response alongside it.

Maury Apr 8

Vibe coding vs. agentic engineering

Achal drew a line worth holding onto: vibe coding is for prototypes. Agentic engineering is reviewed, production-ready code that works and is safe. Simple framing, but useful when you're having the conversation with clients about what AI can and can't own in a production build.

Achal Apr 2

Team Corner

Kunal's AI challenge build

Kunal dropped a screen recording of what he'd been building for the internal AI challenge — 4 fires, 1 star-struck from the team. No details in the thread other than a quick follow-up: "solving for json response." Classic build-in-public energy.

Kunal Apr 7