Participated · Seed · May 2026

NanoCo develops NanoClaw, a secure alternative to OpenClaw that runs AI agents sandboxed in containers to limit access to services and credentials. The project went viral after endorsements from figures like Andrej Karpathy and high-profile attention, driving rapid community growth to thousands of users. Founders Gavriel and Lazer Cohen pivoted from their prior AI marketing startup to focus on NanoClaw full time and secured partnerships with Docker and Vercel. Early adopters have included technical executives at companies such as Amazon, Gap, Google, Meta, SentinelOne and Accenture. NanoCo has begun booking enterprise customers and offers implementation services (forward-deployed engineers) to help businesses roll out NanoClaw and provide ongoing support. The company raised an oversubscribed $12 million seed to support its expansion and continued product development.

Participated · Seed · May 2026

Hacktron builds an AI-driven platform that continuously runs attacker-style security tests on every pull request and code change to find deep, exploitable vulnerabilities and reduce false positives. The company's team of founders comes from competitive hacking backgrounds with wins at DEF CON CTF and presentations at major security conferences. Hacktron has responsibly disclosed vulnerabilities in widely used software and infrastructure, including BeyondTrust Remote Support, Next.js, Cloudflare, GitHub, GitLab, and the OAuth2 Proxy project. It has performed security testing engagements for companies such as Perplexity AI and Supabase and maintains a public advisories database. The company generated about $240,000 in revenue in its first nine months. Hacktron plans to use the new funding to grow engineering and research headcount, speed product development, and scale its go-to-market activities.

Participated · Seed · Dec 2025

Raindrop builds monitoring infrastructure that lets engineering teams discover, track and resolve failures in AI agents, replacing generic metrics with small, adaptive models tailored to each product’s behavior. The platform surfaces signals such as “User Frustration” or custom issues like “Agent Stuck in a Loop,” then issues Sentry-style alerts when incidents spike. Raindrop’s background AI agents also investigate patterns, producing step-by-step explanations of failures. The company serves frontier AI customers processing millions of events per day and offers Raindrop Experiments, an A/B testing toolkit for validating fixes across tools, models or pipelines. Funds will be used to meet growing enterprise demand and entrench the product as the default observability layer for AI agents. Founded by Ben Hylak, Zubin Koticha and Alexis Gauba (previous founders acquired by Coinbase, with Ben formerly at Apple’s Human Interface team), Raindrop is headquartered in San Francisco and positions itself as critical infrastructure for sectors like healthcare and financial services where agent errors can be severe.

Participated · Equity · Oct 2025

Unikraft develops a cloud platform engineered for sub-50 ms cold starts, instant autoscaling, and the ability to host tens of thousands of strongly isolated virtual machines on a single server. The technology stems from research backed by NEC Laboratories Europe and the Linux Foundation and integrates seamlessly with familiar developer tools such as Dockerfiles, Kubernetes, Prometheus, and Grafana. Key performance claims include 10–50 ms workload boots, 10,000–100,000+ VMs per server, and the potential to double requests per second at half the cost of traditional VM deployments. Security is maintained through hardware-level isolation, avoiding the trade-offs commonly associated with containerization. The platform is aimed at companies that need to scale AI workloads without latency or cost constraints. With its recent funding, the company plans to expand both operations and product development efforts. Financially, Unikraft has secured $6 million to date, but no revenue or user figures were disclosed.

Participated · Seed · Sep 2025

Human Behavior analyzes raw session-replay video with computer-vision models to surface why users convert, churn, and encounter bugs without instrumenting code. The product summarizes and segments thousands of hours of footage and delivers daily emails highlighting feature usage, bugs, and churn signals to product teams. Customers to date are mostly fast-moving Series A and B startups, and the company says it has been growing 20% month-over-month since launching four months ago. The founders position the approach as an alternative to manual event tagging and traditional analytics tools like Mixpanel and PostHog. Over time they plan to spin the same core dataset into multiple products, including automated QA and embedded IT support, with the ambition of becoming the Datadog of session replay. The team previously built and sold an earlier startup, Dough, for six figures to Employer.com before focusing full time on Human Behavior.