DEADSTOCK Raises $2.5M

G5 Labs

Ready for the post-developer era? G5 Labs is an MIT CSAIL spinout building the semantic and governance layer for AI-native software development. With the G5 platform, natural language — organized as a structured, formal graph of intent — becomes the source code: compilable, mergeable, and governable for enterprises. Teams are drowning in AI-generated code. A single developer can generate tens of thousands of lines a day, which moves the bottleneck from building to verifying. The fundamental problem: we have slapped AI onto development tools and processes designed for humans. In an AI-native world, the core challenge is no longer maintaining code — it is managing, aligning, and keeping intent consistent. G5 sits above the frontier coding models (Claude Code, Codex, open-weight) as the application layer: it constrains, normalizes, and governs their output. A self-learning, bi-directional compiler converts intent to code and back; merge conflicts are resolved at the level of intent rather than code; every line of generated code stays traceable to the business requirement behind it; and organization-wide policies — GDPR, security, requirement approvals, hard cost controls — are enforced before a single token is spent. Any frontier model, no lock-in. In deployment today: modernizing complex legacy systems for heavily regulated industries — millions of lines of legacy code uplifted into semantic ontologies, rebuilt natively rather than ported, with decades of technical debt left behind. Two ways to start: modernize a legacy system, or replace rigid SaaS with software you own. Founded by AI researchers and large-scale-systems veterans of Google and AWS, with MIT PhDs. CEO Tim Kraska is a professor at MIT CSAIL and co-director of MIT's Generative AI Impact Consortium. Offices in Boston, MA and Palo Alto, CA. We're hiring. Request early access at g5labs.ai.

CorePower Magnetics

Delivering next-generation, high-efficiency, power-dense magnetics engineered from advanced amorphous and nanocrystalline materials and manufactured in the U.S. Achieve up to 10x weight reduction, 5x smaller volume, and 2x lower losses—without rare earths—compared to conventional powder core solutions.