Microsoft published a field study this week on rolling out Claude Code and GitHub Copilot CLI to its own engineers, and the interesting part wasn’t the productivity number. Adoption spread through peers and managers, not official enablement pushes. Adopters merged roughly 24% more pull requests, but Microsoft is careful to note that merged PRs are only a proxy, not proof the code was actually good. A few of this week’s raises are betting that gap never closes on its own.
🏆 DevTools Funding Wins
Meticulous Closes $15M Series A to Make Developers Trust AI-Written Code
London-based Meticulous has closed a $15M Series A led by Chemistry, with participation from Menlo Ventures and a host of prominent angels from the devtools space. Founders Gabriel and Quentin Spencer-Harper built Meticulous to close the gap between how fast AI writes code and how fast humans can verify it. It analyses a codebase, builds an edge-case checklist, and simulates user flows before and after every change so developers see the visual impact before shipping. Customers include Notion, ElevenLabs, and Wiz, with 5x ARR growth this past year. 🛡️ Well earned, Gabriel and Quentin — read the announcement.
Hermes Agent Maker Nous Research Approaches Unicorn Status With $75M Round
Nous Research, the team behind the open-source Hermes agent, is finalizing at least $75M led by Robot Ventures at a $1.5B valuation, on top of roughly $70M raised previously. Founded in 2023 by Jeffrey Quesnelle, Karan Malhotra, Ryan Teknium, and Shivani Mitra, Nous built Hermes as a rival to the viral OpenClaw agent, shipping it with built-in skills for web search, coding, and image understanding, plus the ability to learn new skills from usage automatically. Hermes has amassed roughly 214,000 GitHub stars. 🦄 Big things ahead for this team — read more at TechCrunch.
Rippling Cofounder Secures $15M Seed for Vorflux, an Autopilot for Software Engineering
Prasanna Sankar, CTO and cofounder of Rippling, has launched Vorflux with a $15M seed round led by Y Combinator. Vorflux stands up a company’s full stack on a real machine, plans and builds a feature end to end: writing code, testing it, having a rival model critique the work, and opening a pull request with video proof it works. Sankar’s bet is that models can run for hours unsupervised, so engineers should supervise by exception, not approve every step. The company runs every frontier model rather than training its own. ✈️ Nice launch, Prasanna — find out more on the Vorflux website here..
Casco Raises Series A at $100M Valuation to Automate Continuous Pentesting
Casco, the agentic security engineer that tests software for vulnerabilities around the clock, has closed a Series A at a $100M valuation led by Standard Capital, bringing total capital raised to $17M (the round’s own size isn’t separately disclosed). Founders René Brandel and Ian Saultz built Casco to replace noisy legacy tools with real runtime testing, already finding critical vulnerabilities across 300+ companies running standard tools and human pentesters. Standard Capital’s Bryan Berg, former Stripe security lead, backed the round. Casco is hiring across engineering, security, and growth at casco.com/careers. 🔐 Solid milestone, René and Ian — read the announcement.
🚀 New DevTools Launched
Skyvern Rewrites Playwright in Rust, Cuts Memory Use 70%
Skyvern (YC S23), cofounded by Suchintan Singh and Shuchang Zheng, has open-sourced Rustwright, a browser automation engine that talks to Chromium directly over the CDP protocol instead of routing through Playwright’s Node driver. The team says removing that driver layer makes the engine 2.55x faster and cuts memory use by 70% in Skyvern’s own runs, while also removing the Playwright automation fingerprint since the driver never loads. A single Rust core backs both Python and TypeScript bindings, with Ruby, Go, and Java planned. It’s alpha and Chromium-only for now. 🦀 Great build, team — check out the repo.
Autosana Launches a UI/UX Agent That Judges Whether Your App Is Actually Good
Autosana (YC S25), cofounded by Yuvan Sundrani and Jason Steinberg, has added a second layer to its end-to-end testing: an agent that watches test runs as a senior QA engineer would, flagging visual problems like overflowing text alongside UX issues like confusing transitions or badly timed prompts. Every finding gets a severity, critical, major, or minor, and recurring issues across devices get grouped into one tracked issue rather than flooding teams with duplicates. The team built it as a layer on top of existing E2E testing, not a replacement. 🎨 Smart addition, team — read the launch post.
Kestrel AI Launches Workflows, Turning Plain English Into Deterministic Ops Automation
Kestrel AI (YC F25) , cofounded by Raman Varma and Evan Chopra, has launched Workflows, converting plain English into a production-ready automation graph spanning incident response, provisioning, CI/CD, and security. The team built it so platform teams can describe a fix in a sentence and get a wired workflow with triggers, conditions, and approval gates in minutes. Critically, the agent that builds the workflow isn’t what runs it: once deployed, execution is deterministic with no model in the loop. It ships with 30+ integrations plus a CLI and SDK. ⚙️ Sharp launch, team — read the full post.
Got launch news or funding to share? Send it to Natalie Harper.
Want more details on the opportunities here or want to register your resume with us, reach out to Becca Combe MIRP CertRP and Louise Ogilvy directly.
