Discover
what’s next.

HyLight is building autonomous, hydrogen powered airships to inspect power lines, pipelines, and railways. Its 12 meter aircraft use LiDAR, high resolution cameras, methane sensors, and AI to detect infrastructure defects without relying on helicopters.

Founded in 2022 and based near Paris, France, HyLight was started by Martin Bocken, Josef Rokusek, and Thomas Laporte. The founders bring backgrounds in engineering, computer science, and entrepreneurship, with connections to UC Berkeley.

The company just raised approximately $14.5 million in a Series A backed by Rubio Impact Ventures, DRW, and Teampact Ventures, alongside existing investors Y Combinator and Ring Capital.

HyLight currently operates four airships in France and plans to scale to 15 by 2027. Its aircraft capture imagery at 1 millimeter per pixel, with AI identifying infrastructure defects. The company is also expanding its engineering team across aerospace, software, and AI.

Could autonomous airships replace helicopters for critical infrastructure inspections?

LM Studio is building a platform that lets developers run, test, and integrate open weight AI models directly on their own hardware. Its tools support local inference, model management, and agent workflows without relying entirely on cloud APIs.

Founded in 2023 and based in SoHo, New York City, LM Studio was started by Yagil Burowski, a former Apple engineer who worked on AirTag Precision Finding and iPhone Car Key. 

The company has raised approximately $20 million across two rounds at a $100 million valuation. Reported investors include Matrix, Preston-Werner Ventures, and Torch Capital, although the latest round's structure has not been independently confirmed.

LM Studio operates with a lean team of 13 people, including 9 engineers, and is hiring for four roles: Software Engineer, Agent Harness; Software Engineer, Application; Software Engineer, Inference Runtime; and Full Stack Software Engineer. The team works five days a week from its SoHo office in NYC.

Could local AI agents become a serious alternative to cloud hosted AI infrastructure?

TypeSafe AI is building Jev, a new class of AI models designed to make fast, structured decisions inside software. Unlike traditional LLMs, Jev produces typed, probabilistic outputs without generating text token by token, dramatically reducing latency and compute costs.

Founded in 2024 and based in San Francisco, TypeSafe was started by Diogo Almeida, Erik Gafni, and Sasha Sheng. Almeida previously worked at OpenAI and Google Brain, Gafni is a repeat founder, and Sheng was a research engineer at Meta FAIR.

The company just raised an $870 million Series A led by a16z at a $7.5 billion valuation, with participation from Sequoia and DCVC, bringing total funding to $910 million.

Since launching three weeks ago, Jev has reportedly processed 1 trillion tokens and generated $100 million in revenue. TypeSafe is also hiring across software engineering, AI infrastructure, and model development in San Francisco.

Could specialized decision models become the next major breakthrough in AI infrastructure?