Amjad Masad (CEO @ Replit) argues that we’re in an unprecedented era of rapid wealth creation. He also announced that Agent 4 has turned Replit into an OS of sorts, allowing users to endlessly customize the platform with skills. https://x.com/amasad/status/2039552681493336250
Guillermo Rauch (CEO @ Vercel) shared that Vercel signups are growing at 52% month-over-month, up from 23% and 17% in previous periods. https://x.com/rauchg/status/2039493013043626427
Aaron Levie (CEO @ Box) teased something coming “very soon” from his team, sharing that they’ll “kill me for sharing this” but it’s April 1st so why not. https://x.com/levie/status/2039414479084278239
Garry Tan (President & CEO @ Y Combinator) shared thoughts on local models being “a very very good thing” and posted what appears to be an April 1st joke. https://x.com/garrytan/status/2039568811440128137
Peter Steinberger (OpenClaw) shared that he never uses plan mode, explaining that the main reason it was added to Codex is for “claude-pilled people who strong preference for always planning.” https://x.com/steipete/status/2039551079621566812
Dan Shipper (CEO @ Every) made the case that SaaS isn’t dead—it just needs to become agent-native, pointing to Linear as a great example of this transformation. https://x.com/danshipper/status/2039357127903350960
Peter Yang (Product @ Roblox) shared a provocative thought about how “the combination of mobile and short video has rotted the brains of an entire generation.” https://x.com/petergyang/status/2039563521885901091
Zara Zhang (Builder) shared an aha moment with OpenClaw, replacing her to-do list with “braindumping to-dos to OpenClaw.” She also introduced the “Follow builders” skill for staying on top of AI happenings. https://x.com/zarazhangrui/status/2039599038358814961
Nan Yu (Head of Product @ Linear) shared that Doona is “the Notion of strollers” and asked how many times PMs, sales, or support teams have needed to bother an engineer to find something. https://x.com/thenanyu/status/2039530536050274648
Thariq (Claude Code @ Anthropic) announced a major update: they rewrote the Claude Code renderer to use a virtual viewport, enabling features like split-screen dev and infinite canvas. https://x.com/trq212/status/2039453692592873587
Cat Wu (Claude Code + Cowork @ Anthropic) shared her workflow of using Claude Code on the mobile app to “fire off ideas on the go, then pick them up on my laptop.” https://x.com/_catwu/status/2039421527935033854
No notable posts from Swyx, Nikunj Kothari.
Anthropic Engineering
Claude Blog
Latent Space — “Mistral: Voxtral TTS, Forge, Leanstral, & Mistral 4” with Pavan Kumar Reddy and Guillaume Lample
The Takeaway: Mistral is betting big on enterprise AI by enabling companies to leverage their proprietary data with custom fine-tuned models, while simultaneously pushing boundaries with their new open-weight Mistral 4 model.
Mistral announced several major releases: Voxtral TTS (their first speech generation model supporting 9 languages), Forge (their fine-tuning platform for enterprises to customize models on their own data), and a preview of Mistral 4. The key theme is empowering enterprises—they argued that companies using off-the-shelf closed models are missing out on insights from data they’ve collected for years or decades, sometimes trillions of tokens in specific domains. With Forge, businesses can fine-tune models on their proprietary data without losing control of their IP. Guillaume Lample emphasized that “so much data, sometimes it’s trillions of tokens of data in a very specific domain, their domain, which is data that you will not find in the public on the public internet.”
The conversation also explored the evolving model landscape—debating whether open-weight models will win at the edge versus in data centers, and how enterprises are increasingly wanting to own their model weights rather than be locked into API-only solutions.
https://youtube.com/watch?v=SUjA25ijcNs
Generated through the Follow Builders skill: https://github.com/zarazhangrui/follow-builders