Y Combinator held a Summer 2025 Demo Day last week, showing off the latest batches of over 160 startups.
Like recent batches, the majority of startups have presented AI-centric solutions. However, a clear evolution was clear. Instead of “AI-powered” products, many companies are now building the infrastructure and tools needed to develop AI agents or the development of them. For example, the batch had a surge in voice AI solutions, focusing on helping new businesses monetize the “AI economy” with advertising and marketing tools.
We spoke with a small number of YC-focused investors to learn which startups found most interesting and which produced the highest investment demand.
Below are the most frequently mentioned:
What it does: AI Startup Stripe
Why it's a favorite: Many AI startups use complex pricing models that often blend flat per-seat subscription fees with usage-based fees, credits, and various add-on costs. Managing complex AI pricing on stripes is a time-consuming manual process. Therefore, Autumn has developed an open source infrastructure that simplifies stripe integration for AI startups. The company says the technology is already being used by hundreds of AI apps and 40 YC startups. Given the dominance of stripes in payments and the explosive growth of the AI market, could AI professional billing solutions be the next major fintech success story?
TechCrunch Events
San Francisco | October 27-29, 2025
What it does: build a vercel for an AI agent
Why it's a favorite: Just as Vercel helps developers deploy and host startups, Dedalus Labs claims that its platform automates the infrastructure to build AI agents, reducing hours of coding for hours of clicks. The company handles complex tasks such as automation and load balancing.
What it does: Crowdsource ranking of atmospheric coded designs
Why is it your favorite: The ability of AI to quickly generate huge numbers of designs creates new problems. Design Arena solves this by crowdsourcing rankings of visuals generated in AI, creating a feedback loop that improves AI models. Large AI labs have value in training models and train them to generate better designs. Because some of the designs are already customers of the design arena.
What it does: A tech-enabled distributor for retailers in Southeast Asia
Why it's a favorite: Getasap Asia was founded three years ago by Raghav Arora, who was only 14 years old. Since then, startups using technology to provide supplies to corner stores, restaurants and large supermarkets in Southeast Asia within eight hours have earned millions of revenue. Getasap Asia has closed the rounds from General Catalyst, according to its website, and has heard that the startup is the highest rating in the entire batch.
What it does: AI engineer fixing bugs in production
Why was it my favorite: Founded by 20-year-old Pablo Hansen, who earned his master's degree in AI last year, Keystone has a mission to reduce software breaks. The company's AI has already rejected seven-figure acquisition offers by finding and fixing bugs in clients like Lovable.
What it does: AI Matchmaker for Female Friends
Why is it your favorite: There's no shortage of dating apps, but Realroots tackles another kind of loneliness. Lisa, the company's AI matchmaker, interviews women and then organizes social experiences to connect with compatible friends. The AI part may have performance, but the conversation with Lisa probably won't realize more insight into the participants than written answers – Realroots may be on something. In last month alone, the company generated $782,000 from 9,000 paid clients, the founder said.
What it does: Automate insurance claims with AI
Why it's your favorite: Solva's AI automates the most routine tasks of insurance registrants, from filling out complicated claims to prevent inappropriate payments. Just 10 weeks after its launch, Solva has already raised $245,000 in its annual recurring revenue (ARR), a figure that investors are excited about.
What it does: Counter drone mini missile
Why it's a favorite: With China reportedly accumulating flocks of cheap drones, the US military faces an urgent need for cost-effective counter-drone solutions. Perseus is developing just that. A small missile designed to shoot down drones for just a small portion of the cost of existing systems. Several branches of the US military have already invited startups to demonstrate their solution, which could lead to large contracts.
What it does: AI foreign language tutor
Why I like it: Apps like Duolingo make language learning accessible and fun, but often lack the key elements of flow ency: consistent conversation. Pingo solves this problem by allowing users to speak to AI that acts as native speakers. The company's unique approach is extremely popular, claiming its founders are up 70% each month, earning $250,000 each month.