Artificial intelligence was once again the defining theme of the latest Y Combinator Demo Day. Nearly 190 companies from the Winter ’26 cohort presented their startups. These companies are developing products across a wide range of industries, including law, transportation, and healthcare.
Given the size of the cohort and this year’s media format, which featured posted pitch videos rather than a livestream, it was not possible to watch every presentation. Instead, after reviewing all 190 startups, a selection was made based on those that seemed most intriguing. The following sixteen stood out as the most interesting from this overflowing class.
ARC Prize Foundation creates benchmarks to help measure progress toward Artificial General Intelligence. It is interesting as a nonprofit within YC. With organizations like OpenAI, Anthropic, and Google already using some form of its benchmarks, its inclusion makes sense. The foundation aims to inspire more open-source AGI research by hosting competitions and awarding grants. Tracking progress toward general machine intelligence is becoming a matter of historical record.
Asimov collects human movement data to train humanoid robots. People from around the world submit videos of themselves performing tasks, which the company turns into training datasets. This effort is part of a broader movement to advance humanoid robots for uses beyond supply chains and entertainment. Using data to teach humanoids the fluidity of human movement could help them perform tasks in a less rigid manner.
Avoice helps automate the tedious non-design work for architecture firms. It is interesting because new technology targeting the architectural industry is rare. The founders note this market is underserved yet rich in potential. This AI tool automates tasks like reviewing specifications, drawings, contracts, and proposals, freeing architects from administrative burdens.
Button Computer is building a wearable AI device. As the world awaits wearable AI products from other major players, two former Apple employees have launched Button. It is essentially a tiny computer built for AI that connects to apps like email and Slack, operating them via voice command to perform tasks. The race for the next must-have AI hardware is on.
CodeWisp lets anyone build games using AI. The founders say you simply tell the AI how to make a game, and it will create it. This makes a fun and creative process accessible, lowering the technical barriers that have traditionally made game development difficult. It represents a new generation of creative tools.
Crosslayer Labs helps detect website spoofs. The rise of AI-powered tools has made it easier for bad actors to create fraudulent websites. This startup helps customers detect and monitor such online threats, providing protection against this emerging category of internet risk.
Doomersion teaches you languages as you scroll. The app shows users short videos in a TikTok-like feed, but in the language they are trying to learn. It aims to add meaning to the time we spend mindlessly scrolling by combining content consumption with language acquisition.
Lexius embeds advanced AI into existing security camera systems. The technology enables footage to automatically detect and report incidents like theft or falls, replacing a fragmented and manual process. It targets businesses that have cameras but lack intelligent monitoring.
Librar Labs is an AI-powered library management system. This tool focuses on an industry often overlooked by tech: libraries. It currently helps schools with inventory and cataloguing. With little competition for modernizing tools in this space, new ideas have significant potential.
Milliray provides a radar system to help track small drones. Defense tech is a hot category, and this startup addresses a specific need. Human observers can mistake drones for birds and vice versa. This sensor-based system accurately identifies small drones, a capability of growing importance.
MouseCat uses AI to investigate fraud. As AI becomes a tool for scammers, this company analyzes a company’s consumer data from platforms like Databricks to identify suspicious activity and recommend actions. It represents an AI-native defense against AI-powered threats.
Opalite Health uses AI to help healthcare providers communicate with non-English speaking patients. Breaking language barriers in medical settings can be critical. This translator enables better understanding between providers and patients, improving access to care in diverse communities.
Sequence Markets lets people trade across various markets, like crypto and prediction markets, on one unified system. It addresses the desire for a less fragmented trading experience, consolidating different asset classes into a single platform.
ShoFo is building a comprehensive video library, billing itself as the world’s video index. While it serves as a custom video index to help AI labs find diverse datasets efficiently, the concept of a vastly improved search and organization tool for video content is compelling.
Sonarly helps software fix its own production issues. It connects to monitoring systems, reduces alert noise, automatically identifies root causes of problems, and suggests fixes or further actions for engineers. It automates a critical part of the software workflow after code is deployed.
Terranox AI uses AI to find uranium deposits in North America. The founders note that uranium will be needed to power the next generation of nuclear energy, which is seen as a potential power source for the growing number of data centers. This startup applies AI to the critical challenge of resource identification for future energy needs.