TL;DR
Claude plugged into Microsoft 365 everywhere, Swyx said Devin one-shot blog-to-code, and Peter Steinberger called out GitHub’s API as still not built for agents. Aaron Levie hit the context wall, while Garry Tan shipped a DX review tool from his own stack.
Microsoft 365 connectors are now on every Claude plan, so you can pull Outlook, OneDrive, and SharePoint into the chat instead of juggling tabs. It’s a practical move: Claude is turning into a more useful work assistant by sitting closer to the docs and email people already live in.
He says he keeps slamming into GitHub API quota limits, which is a pretty blunt sign the platform still assumes human usage, not agentic workflows. In the same vein, he’s arguing Anthropic’s latest moves hurt the ecosystem while giving credit for the small fixes that soften the blow for API users.
He says a $10B company suing an app for "looking and feeling" similar is the wrong move, and doubles down on Bevel Health as the better answer: free, accessible health for everyone. The jab at WHOOP makes the point pretty clear — compete on product, not lawfare.
He says AI agents won’t magically erase management overhead: they’re only as good as the context, tools, and human oversight around them. Until agents can reliably self-escalate and operate inside a broader system, people will still do the real mental work — which is why the “jobs disappear” take is off.
He’s releasing `/plan-devex-review`, a new GStack tool to help teams create better developer experiences, inspired by Addy Osmani’s DX framework. The subtext: he builds for himself first, then turns the internal tooling into something other founders can use.
He argues that when design and product are fused too tightly, design gets demoted, exploration gets shallow, and brand suffers. His cleaner split: keep brand and product design close, and treat PMM as a product concern instead of mashing design and PM together.
He says he’s hit “agentic self improvement”: copy-paste a blog post or tweet into Devin and it can one-shot the implementation. The surprising part is that it worked even though this feels way out of distribution for Gemini Flash Lite underneath — and that’s the real signal.
She says the new weird trend is distilling colleagues, influencers, and even exes into agent skills. It’s a sharp read on where AI agents are headed: less generic assistants, more personalized behavior packs people can swap around.
AI’s boom is real—but the buildout will still get messy
The Takeaway: AI isn’t a hype cycle so much as an 80-year research backlog finally cashing out.
Marc Andreessen, cofounder of Andreessen Horowitz, frames AI as the culmination of decades of serious work rather than a sudden miracle. He’s been in the field since the late ’80s, remembers the Lisp-and-expert-systems era, and argues that the industry has always moved in waves: “summer, winter, summer, winter.” What changed is that the old arguments against neural nets and transformers have been answered by reality. In his view, the real unlocks were AlexNet, then transformers, then the recent leap into reasoning and agents.
His sharpest point is that the current wave is no longer just demo magic. Coding is the proof. Once AI can beat strong programmers in meaningful tasks, everything else becomes a downstream problem. That’s why he calls this an “eighty year overnight success.”
Still, he’s not naïve about the business side. He compares today’s GPU and data-center frenzy to the telecom overbuild of the late ’90s: demand was real, but the capital got ahead of itself. The difference now is that the biggest spenders are Microsoft, Amazon, Google, Meta, Nvidia, OpenAI, and Anthropic—not fragile startups. The boom is real, but the path will be uneven, expensive, and full of companies that get crushed before the market settles.
Peter Steinberger called AI code review the killer dev workflow, while Thariq said Claude Code turned files into a workbench. Cursor bet on specialized models plus brutal RL infra, and Anthropic detailed how it contained Claude across products.
Steinberger said AI agents need lean skills, not bloated prose; Levie argued AI won’t kill jobs, it’ll raise the bar. Yang said Codex tested itself while Claude still owned frontend, and Garry Tan pushed hard evals as the way agents got better fast.
Levie warned AI demos fool CEOs; the hard work starts after. Tan said high agency plus taste beat everything, while Yang urged builders to build the system that builds the MVP first. LeCun kept the pressure on: LLMs help, but they’re not the road to real intelligence.
Levie said AI kills tasks, not jobs, so roles just expand. Masad said Replit’s AI stack can ship MVPs in a weekend, while Tan showed Qwen fine-tunes in hours and personal AI got real. No Priors said Cerebras won by betting on speed before the market cared.
Levie said AI will surface more bugs, making humans the bottleneck. Steinberger argued GitHub’s native PR limits kill bot policing, while Tan pushed the “ship the 60%” rule and Swyx said AI coding should harden codebases, not just move faster.