YC Winter ’26 Demo Day Showcases 16 Startups

Another Y Combinator Demo Day, another wave of AI-first ideas—and this time, nearly 190 teams took the virtual stage. From evaluation tools for frontier models to counter-drone radar and medical translation, these 16 startups stood out for their clarity of purpose and sharp product focus.

Together they reflect where the market is moving: applied AI that plugs into real workflows, safety and security for both the physical and digital worlds, and infrastructure for the data-hungry systems powering it all. Here are the companies to watch from YC’s Winter ’26 cohort.

Benchmarks for robots and voice-first wearables

ARC Prize Foundation is building rigorous benchmarks and prize competitions to measure progress toward general-purpose intelligence. Standardized evals have become table stakes for leading labs—OpenAI, Anthropic, and Google DeepMind all lean on batteries of tests—so a nonprofit stewarding transparent, reproducible measures is a welcome counterweight to hype.

Asimov crowdsources human movement data to train humanoid robots, a crucial ingredient for dexterity and safety. With hardware players like Figure partnering with automakers and Agility Robotics piloting warehouse robots, the bottleneck is increasingly high-quality demonstrations—exactly the “data flywheel” Asimov aims to spin up.

Button Computer is a voice-first wearable that connects to daily tools like email, Slack, and Salesforce. After early stumbles from devices like the AI Pin, the lesson is clear: utility wins. Button’s bet is an invisible interface that actually gets work done, not just novelty interactions.

AI for real-world workflows across key industries

Avoice tackles the back office of architecture, automating spec reviews, submittals, and contract administration. The AEC sector still runs on PDFs and manual markups; plugging AI into Revit- and Bluebeam-heavy workflows could free billable time for design rather than paperwork.

Lexius layers advanced computer vision onto existing camera fleets to detect theft, falls, and anomalies in real time. For operators who cannot rip-and-replace thousands of cameras, retrofitting analytics promises faster incident response and fewer costly claims—an area insurers and risk managers scrutinize closely.

Librar Labs modernizes school library systems with AI-driven cataloging and inventory. With many districts still relying on legacy software and limited staffing, automated metadata enrichment and reconciliation against standards could raise circulation accuracy without raising budgets.

Sonarly brings AIOps to production incidents, correlating signals across monitoring tools, suppressing alert noise, and suggesting fixes. Teams measure uptime by MTTD and MTTR; even modest reductions compound into meaningful savings when every minute of outage burns both revenue and on-call morale.

Security and risk intelligence for modern enterprises

Crosslayer Labs helps enterprises detect spoofed sites and brand impersonation—threats amplified by cheap generative tools. As phishing volumes set new highs according to anti-fraud groups, continuous discovery across domains, TLS certs, and lookalike infrastructure is becoming a must-have control alongside DMARC and takedown workflows.

MouseCat plugs into data platforms like Snowflake and Databricks to spot suspicious behavior and recommend actions. Fraud’s “multiplier effect” means each $1 lost can translate to multiple dollars in operational costs, as cited by industry studies, so earlier detection and automated case triage directly protect margins.

Learning and consumer experiences shaped by AI

Doomersion reframes language learning as short-form video scrolling, aligning with how people actually spend hours on their phones. Microlearning research supports spaced, bite-sized exposure; pairing that with creator-style clips could raise stickiness beyond the typical 2-week app drop-off window.

CodeWisp promises “describe it and ship it” game creation, translating natural language into playable experiences. As Roblox and Unity expand the creator economy, removing code hurdles could unlock a wider funnel of designers and storytellers while still exporting to professional toolchains.

ShoFo bills itself as a structured video library for AI labs and enterprises, focusing on provenance and efficient dataset search. With debates over copyrighted training data intensifying, curated, permissioned collections that can be audited and filtered are increasingly valuable to both researchers and compliance teams.

Health and access solutions improving clinical communication

Opalite Health offers an AI medical translator for clinicians and patients who don’t share a language. Research in leading medical journals has shown interpreter use reduces errors; the opportunity—and responsibility—here is delivering accuracy and privacy that meet hospital standards while shortening time-to-care.

Markets and finance platforms unifying alternative trading

Sequence Markets unifies trading across crypto, prediction, and other alternative venues inside one interface. Fragmented liquidity, chain-specific UX, and shifting rules have kept many on the sidelines; consolidation of routing, pricing, and risk management could make niche markets feel institutional-grade.

Defense and energy technologies for drones and uranium

Milliray builds radar tuned to detect and track small drones, a need underscored by modern conflicts and critical infrastructure protection. Human spotters misclassify targets; compact sensors with robust classification give security teams earlier alerts and better counter-UAS coordination.

Terranox AI applies machine learning to geophysical and geochemical data to locate uranium deposits in North America. As data centers and electrification push baseload demand, nuclear is reentering national plans per energy agency outlooks; AI-driven prospecting could speed exploration while informing environmental assessments.

Why these 16 matter for the next wave of YC startups

Across categories, the throughline is pragmatic AI: not just demos, but systems that plug into the world’s messy edges—legacy cameras, school stacks, hospital rooms, factory floors. If this cohort is a compass, expect the next year of YC alumni to prioritize utility, safety, and data discipline over flash.