- Published on
Cognition, Cursor, and Vercel launch the same AI agent review feature, and WorkOS is valued at $2 billion
- Authors
- Name
- Hashim Warren
- @hashim_warren
Launch of the Week
Three apps share launch of the week honors because they rolled out the same powerful feature within days of each other.
AI agent startup, Cognition announced Devin 2.2 with this exciting capability: “Devin runs through your app and sends back screen recordings so you can review every detail of its work.”
That release should have excited developers everywhere. But rival Cursor premiered “Cloud agents [that] produce merge-ready PRs with artifacts (videos, screenshots, and logs) that make it possible to quickly review their changes.”
Two companies have agents that can review their own work and record videos. Then Vercel stole the show with a free tool and skill that has the same functionality. I used it on hypeburner.com and it worked well.
The tough lesson for us all is that we can no longer ship innovative and useful features and gain market attention. There’s no longer any technical moat, so our GTM motions need to become the focus.
Enterprise Watch 🏆
App builder v0 gets the 🏆 this week for a new feature they unrolled without much fanfare. They added a new team setting to allow v0 to open Github pull requests on behalf of members without GitHub accounts. This enables non-technical members of a team to make changes to an app and submit it to be reviewed without friction.
This is the kind of small but thoughtful feature v0’s Sales team will likely lead with on a demo.
Funding and Acquisitions
Enterprise auth platform, WorkOS raised an astounding $100 million series C, and is now valued at $2 billion. CEO Michael Grinich says in the announcement letter that WorkOS has the primitives to be ready to serve auth to a new class of user - AI agents.
I like the essay the team penned a few weeks ago about how complex the problem is of giving AI agents access to your business resources.
Announcement
GTM Trends
Developer tool companies are now publishing their own benchmarks for how well AI models use their tech.
The first I saw do this was Vercel with Nextjs. But now Stripe and Nuxt have run their own AI tests.
Vendor evals achieve two goals. First, they attract attention from the wider developer community. Engineers discuss and debate which AI models are the most powerful. And next, it helps customers use the best AI agent for the job. For example, gpt-5.3-codex is the best model for building Nextjs projects, but if you’re using Nuxt you’ll see more success with Claude Opus 4.6
More Developer Tool News
✨ AI framework, Mastra debuted a new coding agent, Mastra Code, that has the distinction of long term agent memory.
✨ Figma alternative, Paper launched a desktop app and MCP server.
✨ Payment platform, Bolt is partnering with merchant of record, Toffee to break into the gaming apps market.
✨ Fintech platform, Plaid announced a new fundraise at a valuation of $8 billion. That’s a 31% jump from their fundraise last year.