Participated · Seed · Aug 2026

Apollo Atomics focuses on shrinking and mass-manufacturing the steam generator—the largest component in many nuclear plants—by using a compact metal block with needle-thin channels to transfer heat more efficiently. Its design, based on MIT research, enabled the company to build a reactor it says is roughly 40-times smaller than ones using conventional steam generators and to assemble systems in a factory rather than on site. Apollo has built a 40-kilowatt demonstration reactor at MIT and plans to offer reactor sizes from about 10 megawatts-electric up to 300 megawatts. The company projects it can build a 300 MW plant in under 24 months and estimates the reactor will cost four to five times less to produce than existing designs. Apollo targets a levelized cost of electricity near 3 cents per kilowatt-hour and aims for commercial deployment in 2028. It participated in Y Combinator’s Spring 2026 batch.

Participated · Equity · May 2026

SOND was founded in February 2022 by CEO Yadid Ayzenberg (formerly Bose’s Head of Sleep Products) and CTO Amir Lazarovich. Its debut product, Dreambuds, is an in-ear, closed-loop system that senses 12 physiological signals — including respiration, heart rate variability, sleep staging, body position, snoring, and seismocardiography — and streams data to a cloud-based AI sleep coach that selects or generates audio programs in real time. The AI coach can be engaged via a double-tap gesture, supports voice interaction, and draws from a proprietary library of over 500 audio programs while learning what works for each user. The Dreambuds system includes a charging case with Wi-Fi, Bluetooth, an OLED display, physical buttons, and a speaker so the system can run end-to-end without a phone. SOND has run comfort studies and betas, is accepting reservations on its website, and plans a crowdfunding campaign ahead of mass production targeted for Q2 2026.

Participated · Seed · Mar 2026

Sycamore is building an agentic orchestration layer intended to handle end-to-end enterprise workflows, from coding to back-end infrastructure and data integrations. The company’s approach is to design solutions starting from the problem and assemble agents, back-end systems, front ends, or integrations as needed rather than layering agents onto existing workflows. Sri Viswanath, the founder and CEO, brings decades of enterprise platform and cloud transformation experience from roles at Sun Microsystems, VMware, Groupon, and Atlassian. Sycamore has reported early traction with unnamed large enterprise customers. Financially, the company recently closed a $65 million seed round led by Coatue and Lightspeed with broad institutional and angel participation. It is entering a competitive market that includes both startups and large model makers and cloud providers offering agent platforms.

Participated · Equity · Mar 2026

Unreasonable Labs is a Palo Alto-based generative discovery company that pairs large language models with a neurosymbolic map of scientific knowledge to create an operating system for research and development. Its platform synthesizes unstructured data from multiple fields, proposes novel, validated hypotheses, simulates solutions, and designs lab experiments. An orchestration layer can translate discoveries into executable hardware protocols, ingest experimental data, and close the loop from hypothesis to prototype. This cross-disciplinary engine enables, for example, applying aerospace engineering principles to synthetic biology challenges. The company has already launched pilot collaborations with industrial partners in the energy transition, materials science, and pharmaceutical sectors. Fresh capital will be used to grow the technical team, expand operations, and accelerate product development. Beyond its Palo Alto headquarters, the company maintains an office in Cambridge, MA. To date, Unreasonable Labs has raised at least $13.5 million.

Participated · Seed · Feb 2026

Adaption Labs is an AI research company focused on moving beyond static, parameter-frozen models by creating systems that can learn continuously from real-world interactions. Co-founded by Sara Hooker and Sudip Roy, the team emphasizes efficiency and adaptability over brute-force scale, designing models that structurally understand user needs and adjust their behavior on-the-fly without changing core weights. This approach aims to deliver more responsive, resource-efficient AI capable of evolving alongside user requirements. The startup plans to channel recent capital into expanding operations and accelerating product development. Based in San Francisco, the company positions itself at the forefront of adaptive intelligence research, targeting applications where traditional large-scale models are too rigid or compute-heavy. Although no revenue or user figures were disclosed, the sizeable seed investment underscores investor confidence in the technology’s commercial potential.