What 930 YC Launches Tell Us About Where Startups Are Going
We analyzed every YC launch from the past five months — 930 launches across 851 unique companies. Here's what the data actually says.
930launches
851unique companies
79%AI-tagged
18%solo founders
1.The AI Takeover
79% of companies carry an AI tag — but the real story is where AI is going. It's no longer just developer tools. It's uranium discovery, cattle monitoring, dental booking, trucking operations, electrical estimating, and even cemetery management.
39% mention “agents” or “agentic” in their pitch. The shift from copilot to autopilot is decisive. YC founders aren't betting on humans-in-the-loop — they're betting on humans out of the loop.
“AI for X” is the new “Uber for X.” The formula: take a slow, manual, expensive industry process → apply LLMs and agents → automate end-to-end.
Top 20 tags across all 851 companies
| # | ||
|---|---|---|
| 1 | ai | 326 |
| 2 | artificial_intelligence | 300 |
| 3 | b2b | 228 |
| 4 | saas | 170 |
| 5 | developer_tools | 121 |
| 6 | generative_ai | 73 |
| 7 | automation | 61 |
| 8 | fintech | 59 |
8 records
2.Everyone's Building the Same Thing
Perhaps the most striking finding. Across 851 unique companies, massive clusters are building near-identical products. When a dozen companies are all building “AI for legal” or “AI voice agents,” differentiation is everything.
The question for founders isn't “is this a good idea?” — it's “why will you win when 15 others are building the same thing?”
The most crowded verticals
| # | ||
|---|---|---|
| 1 | AI Medical / Healthcare | 79 |
| 2 | Voice AI Agents | 32 |
| 3 | AI Insurance | 28 |
| 4 | AI Real Estate | 22 |
| 5 | AI Legal | 17 |
| 6 | AI Construction / Trades | 16 |
| 7 | AI Recruiting / HR | 11 |
| 8 | AI Accounting / Tax | 10 |
8 records
AI Medical & Healthcare (79 companies)
The single most crowded vertical. Sub-clusters include healthcare billing, clinical AI, mental health platforms, and drug discovery.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Docura Health | When workers' compensation claims are disputed, Qualified Medical Evaluators must manually review thousands of pages of medical records, consuming 6-8 hours for straightforward cases or multiple days for complex ones. Doctors either lose 40% of revenue to agencies handling administrative work or spend excessive time on paperwork, limiting their case capacity. | Automated platform that converts medical record PDFs and case documents into fully compliant reports within 30 minutes, automates patient history collection calls, enabling doctors to process significantly more cases while retaining more revenue per case. | legalinsurancehealth_tech | AI-Native Med-Legal Firm to automate tedious workers' comp workflows for doctors | Launch |
| 2 | Fort | Existing wearables fail to deliver seamless strength tracking experiences despite evidence linking muscle mass and strength to lower all-cause mortality. | A screenless wearable device that detects strength training and fatigue while tracking sleep, cardio, and overall health in an all-day worn device with premium design finishes. | consumerdigital_healthhealth_&_wellnessconsumer_health_services | Automated Strength Tracking for Longevity | Launch |
| 3 | Mecha Health | Radiologists need tools to interpret medical imaging studies efficiently while maintaining workflow consistency across different anatomies, modalities, and clinical contexts. | Foundation models that interpret medical studies like radiologists do, offering multi-anatomy interpretation, multi-modality support, multi-view reasoning, prior-aware comparison, and environment-agnostic delivery integrated into existing workflows. | healthcarehealth_techcomputer_visionmachine_learningartificial_intelligence | Vision-language foundation models that scale healthcare | Launch |
| 4 | Cenote | Telehealth clinics spend heavily on customer acquisition but convert only a fraction of site visitors. Patients abandon due to questions and uncertainty, while clinics rely on ineffective marketing automation or human reps who cannot scale. | AI-powered sales agents that reach out via call and text, answer patient questions, qualify needs, and persistently follow up over days or weeks to guide patients through purchase decisions. | b2bhealthcarehealth_techartificial_intelligencerobotic_process_automation | AI sales agents that immediately and persistently call and text your leads | Launch |
| 5 | CellType | 90% of drugs fail in clinical trials because preclinical models using mice and cell lines don't match human biology. Drug development takes 10+ years and $2B+ while teams optimize against incorrect models. | Foundation models trained on cellular biology (Cell2Sentence technology) combined with AI agents that autonomously execute the drug discovery pipeline, simulating outcomes in virtual human models before clinical trials. | biotechtherapeuticsmachine_learningartificial_intelligenceai-powered_drug_discovery | The Agentic Drug Company | Launch |
5 records
Voice AI Agents (32 companies)
From healthcare billing calls to car dealership scheduling to property management — voice AI agents are everywhere.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Yuma AI | Social media comments represent an unmanaged channel in e-commerce. Negative comments go unnoticed for days, product questions receive delayed responses causing lost sales, and brand perception suffers from poor comment management. Most teams lack systematic approaches to handle public comments at scale. | Social AI automatically monitors and manages comments across Facebook, Instagram, and TikTok by classifying comments by intent and sentiment, then executing predefined automation rules. The system safely automates up to 60% of repetitive moderation while maintaining brand voice and escalating sensitive issues. | aiai_assistantcustomer_supportconversational_aiartificial_intelligence | Put E-commerce Social Media on Autopilot | Launch |
| 2 | Cozmo AI | Enterprise operations in financial services require extensive tool switching and data movement. Customer interactions represent only a small portion of time; most effort involves opening multiple systems, finding history, verifying details, and documenting outcomes. | Multimodal AI employees that handle inbound/outbound calls, read documents and videos, apply company policies, and update core systems like Guidewire, Temenos, and Salesforce. Uses proprietary Agent Protocol Engine (APE) for coordination, audit trails, and compliance tracking. | aienterpriseconversational_ai | The system of work for enterprises in financial services and insurance | Launch |
| 3 | Pollen | Customer success teams are stretched thin managing 40+ accounts each, piecing together signals from 5+ sources. Churn risks go unnoticed and expansion opportunities are missed. | AI agents monitor every customer account 24/7, connecting email, support tickets, product usage, and CRM data to detect churn and upsell signals. Pollen flags which accounts need attention, provides context, and suggests actions. | generative_aicustomer_successconversational_aiartificial_intelligence | AI agents for customer success | Launch |
| 4 | Patientdesk.ai | Dental practices lose revenue at the front desk through missed calls, delayed follow-ups, and poor insurance verification. High customer acquisition costs paired with low conversion rates due to inability to answer calls during business hours or after-hours periods. | An AI booking system handling end-to-end patient intake including incoming calls, qualification, appointment booking, real-time insurance verification, and payment processing integrated with major practice management systems. | call_centerai_assistantcustomer_serviceconversational_ai | Patient-Aware AI Booking for Dental Clinics | Launch |
| 5 | Maven | Voice AI agents can qualify leads and negotiate pricing, but cannot complete transactions. Companies deploying voice agents lack secure payment infrastructure, requiring them to build PCI-compliant systems, handle sensitive card data, manage failed payments, and support multiple gateways. | Single API call that enables voice agents to collect secure card payments during live calls. Maven handles card collection, processing, retry logic, and PCI compliance, with card data flowing through Maven's infrastructure rather than customer servers. | saasfintechinfrastructuredeveloper_toolsconversational_ai | Payments for Voice AI Agents | Launch |
5 records
AI Insurance (28 companies)
Claims processing, broker automation, coverage comparison, and underwriting — insurance is getting the full AI treatment.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Docura Health | When workers' compensation claims are disputed, Qualified Medical Evaluators must manually review thousands of pages of medical records, consuming 6-8 hours for straightforward cases or multiple days for complex ones. Doctors either lose 40% of revenue to agencies handling administrative work or spend excessive time on paperwork, limiting their case capacity. | Automated platform that converts medical record PDFs and case documents into fully compliant reports within 30 minutes, automates patient history collection calls, enabling doctors to process significantly more cases while retaining more revenue per case. | legalinsurancehealth_tech | AI-Native Med-Legal Firm to automate tedious workers' comp workflows for doctors | Launch |
| 2 | Cozmo AI | Enterprise operations in financial services require extensive tool switching and data movement. Customer interactions represent only a small portion of time; most effort involves opening multiple systems, finding history, verifying details, and documenting outcomes. | Multimodal AI employees that handle inbound/outbound calls, read documents and videos, apply company policies, and update core systems like Guidewire, Temenos, and Salesforce. Uses proprietary Agent Protocol Engine (APE) for coordination, audit trails, and compliance tracking. | aienterpriseconversational_ai | The system of work for enterprises in financial services and insurance | Launch |
| 3 | Decoda Health | Clinics using disconnected point solutions face fragmented workflows across ten different tools, creating missed opportunities, poor patient experiences, and excessive administrative burden. | An integrated platform combining scheduling, intake forms, communications, charting, inventory management, patient messaging, and analytics with AI agents that handle calls, documentation, calendar management, and patient follow-ups. | billingmessaginghealthcare_itdigital_healthartificial_intelligence | The operating system for elective care | Launch |
| 4 | Acolite | Insurance agencies spend more time on manual operations than client advising. They struggle with certificate requests, AMS updates, and carrier submissions while facing talent shortages (73% of account managers are 60+) and expensive BPO solutions that deliver only ~15% cost savings. | AI teammates that automate operational workflows by handling certificate generation, carrier submissions, SOV processing, and AMS updates through a secure browser environment with human oversight. | aiinsuranceoperations | Scale your agency without adding headcount | Launch |
| 5 | LunaBill | Healthcare billing teams spend extensive time making insurance calls averaging 25 minutes each. With staffing challenges and expensive scaling, hospitals struggle to recover the $260B in denied or delayed payments annually. | AI voice agents that automate insurance calls for claim status checks, appeals follow-ups, and reprocessing. Fine-tuned on 1.2M+ transcripts, agents handle up to 100 concurrent calls per client. | healthcarehealth_techgenerative_aihealthcare_itartificial_intelligence | AI Voice Callers for Healthcare Billing Teams | Launch |
5 records
AI Real Estate (22 companies)
Property appraisals, rental agents, mortgage lending, property management, and CRE brokerage.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Bystreet | Hotel data across Caribbean and Latin America is fragmented and ownership is opaque, making it difficult for firms to source qualified leads and grow regional presence | AI agents extract and verify data from thousands of public, government, and offline sources to build a structured identity graph of hotel properties, owners, and decision-makers across CALA | b2bsaasproptech | Your key to your next deal | Launch |
| 2 | Scout Out | Residential contractors struggle to find new projects early and lack time to prepare accurate estimates while managing active builds. Projects are scattered across city systems, and contractors often rush bids to respond quickly, leading to incomplete estimates, lost opportunities, or margin problems. | Platform combining project discovery of recently approved residential building plans with AI-powered instant bidding. Contractors upload plans or descriptions to generate editable quotes in seconds, designed for real workflow between active jobs. | aiproptechreal_estateconstruction | Find residential construction jobs and bid faster | Launch |
| 3 | AutoSitu | Real estate development projects experience significant delays in permitting and review cycles. Most projects require 2–3 resubmittals, adding 3–9 months of delay. Developers face $50k–$250k+ in carrying costs; architecture teams spend 80–200 hours per resubmittal at approximately $130/hour. | AI-powered development plan review engine that automatically identifies issues before submission. Enables teams to reduce city comments by 50–70%, cut resubmittals from 2–3 down to 0–1, and shorten approvals by 3–6+ months. | govtechlegaltechreal_estategenerative_ai | AI-native Development Plan Review for Faster Approvals | Launch |
| 4 | InspectMind AI | Construction projects experience costly delays from drawing mistakes including building code violations, dimension mismatches across disciplines, material inconsistencies, and responsibility confusion. These errors cause rework, schedule delays, and can cost six to seven figures per project. | An AI agent that analyzes architectural, structural, civil, and MEP drawings along with specifications and building codes to automatically identify dimension mismatches, material discrepancies, ownership confusion, missing information, and code violations. | aib2bproptechconstructionartificial_intelligence | Catch drawing & spec issues before they cost you money. | Launch |
| 5 | Goldbridge | Real estate banking is fundamentally broken. Over $1 trillion in annual rent flows through landlord accounts, with approximately 25% sitting idle. Property owners juggle multiple bank accounts, spreadsheets, and manual reconciliations while missing yield opportunities and efficiency gains. | An all-in-one financial platform featuring FDIC-backed banking designed for real estate, automated savings and earnings optimization (property tax reassessments, vendor renegotiations, government rebates, utility bill error detection, cash yield optimization), and intelligent money movement capabilities powered by LLM automation. | aifinanceneobankproptechreal_estate | Ramp for Real Estate | Launch |
5 records
AI Legal (17 companies)
AI law firms, contract review, immigration, litigation support, patent analysis, and even AI-written legislation.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Docura Health | When workers' compensation claims are disputed, Qualified Medical Evaluators must manually review thousands of pages of medical records, consuming 6-8 hours for straightforward cases or multiple days for complex ones. Doctors either lose 40% of revenue to agencies handling administrative work or spend excessive time on paperwork, limiting their case capacity. | Automated platform that converts medical record PDFs and case documents into fully compliant reports within 30 minutes, automates patient history collection calls, enabling doctors to process significantly more cases while retaining more revenue per case. | legalinsurancehealth_tech | AI-Native Med-Legal Firm to automate tedious workers' comp workflows for doctors | Launch |
| 2 | Bits | OpenClaw is powerful but difficult to set up securely; users face malware risks, complex configuration, and fear of exposing their data. | Klaus is a pre-configured, cloud-hosted OpenClaw instance that sets up in minutes with integrated Slack, Telegram, AI APIs, email, and security protections built-in. | aiopsdeep_learningai-agents | Get Your OpenClaw Personal Assistant in 5 minutes | Launch |
| 3 | General Legal | Founders pay traditional law firms $500-$2000/hour for slow contract reviews (days/weeks). These firms use inefficient manual processes, make embarrassing errors, and deliver poor results despite high costs. | AI-native law firm pairing top-tier attorneys with proprietary AI workflows enabling 10x efficiency. Offers flat $500/contract pricing, 1-3 hour turnarounds, and higher-quality legal work through automated error elimination. | legallegaltech | The AI Native Law Firm for the Growth Stage | Launch |
| 4 | Mage Legal | Legal diligence for M&A and financing transactions requires days of manual document review, limiting deal capacity and straining attorney resources. | AI-powered platform that automates document classification, contract analysis, issue flagging, and report generation to complete diligence in hours. | ailegallegaltechartificial_intelligence | Turn data rooms into deal intelligence. | Launch |
| 5 | Arcline | Legal services for startups are slow, have opaque billing, and lack AI integration. Even at well-funded companies, legal matters get delayed for weeks and companies face surprise bills with unclear costs. | Arcline combines elite startup lawyers with AI to handle legal work efficiently. AI generates initial drafts, reducing lawyer workload by 80%, enabling flat-rate pricing and same-day turnaround on many matters. | legal servicesAIstartupslawcontracts | AI-native legal services for startups | Launch |
5 records
AI Construction & Trades (16 companies)
Estimation, project management, drawing review, contractor operations, and workforce management for the trades.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Avoice | Architects dedicate substantial time to essential but tedious tasks including specification writing, schedule coordination, code verification, document review, and maintaining consistent project knowledge across expanding teams. This work remains fragmented across multiple file formats and is predominantly manual, error-prone, and inadequately supported by existing software. | Avoice provides a unified workspace where project documentation—specs, schedules, and coordination data—resides in one connected location. The platform uses AI to automate repetitive documentation tasks while maintaining architect oversight of decisions and standards, enabling teams to reference and update information consistently rather than duplicating effort across documents. | aiconstructionarchitectureenterprise_software | The AI workspace for architects | Launch |
| 2 | Scout Out | Residential contractors struggle to find new projects early and lack time to prepare accurate estimates while managing active builds. Projects are scattered across city systems, and contractors often rush bids to respond quickly, leading to incomplete estimates, lost opportunities, or margin problems. | Platform combining project discovery of recently approved residential building plans with AI-powered instant bidding. Contractors upload plans or descriptions to generate editable quotes in seconds, designed for real workflow between active jobs. | aiproptechreal_estateconstruction | Find residential construction jobs and bid faster | Launch |
| 3 | Bidflow | Electrical estimating is tedious and manual, requiring estimators to read through massive PDFs, manually count symbols, and enter thousands of data points into legacy software. This slow process causes contractors to lose jobs by small margins. | An AI copilot platform that processes project documents, automates repetitive estimation tasks, and provides transparent reasoning so estimators review rather than replace human judgment. Combines features of multiple legacy systems into one unified platform. | saasconstructionai_assistantinfrastructure | AI Copilot for Electrical Estimating | Launch |
| 4 | Galactic Resource Utilization Space, Inc. (GRU Space) | Lack of off-world surface habitation technology and infrastructure for human presence beyond Earth | In-situ resource utilization technology that transforms lunar material into durable building structures, starting with a lunar hotel targeting 2032 opening | travelaerospaceconstructionspace_exploration | Building the First Hotel on the Moon | Launch |
| 5 | InspectMind AI | Construction projects experience costly delays from drawing mistakes including building code violations, dimension mismatches across disciplines, material inconsistencies, and responsibility confusion. These errors cause rework, schedule delays, and can cost six to seven figures per project. | An AI agent that analyzes architectural, structural, civil, and MEP drawings along with specifications and building codes to automatically identify dimension mismatches, material discrepancies, ownership confusion, missing information, and code violations. | aib2bproptechconstructionartificial_intelligence | Catch drawing & spec issues before they cost you money. | Launch |
5 records
AI Recruiting & HR (11 companies)
AI interviewers, automated screening, network-based hiring, and background checks.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | HackerRank | Recruiters face excessive applicants with scheduling overhead, unreliable signals due to AI-polished applications, and rising interview integrity issues including cheating tools. Hiring managers waste time on wrong candidates. | Chakra is an AI interviewer platform that conducts structured voice/video screening interviews, adapts intelligently to gather needed signals, and delivers decision-ready reports with evidence, transcripts, and session replays. | recruitingdeveloper_tools | AI interviewer that finally works | Launch |
| 2 | Perfectly | Traditional recruiting agencies move slowly, provide low-quality candidates, and struggle with difficult roles. The process is inefficient and outdated compared to modern AI capabilities. | AI-powered recruiting platform that automates candidate identification, outreach, and screening. Uses deep research and predictive analytics to match candidates with roles, delivering qualified prospects directly to Slack and scheduling interviews automatically. | recruitingartificial_intelligence | Your AI Recruiting Agency | Launch |
| 3 | Fiber AI | Data providers force customers to purchase bulk datasets at high costs ($20,000-$40,000+ monthly). Product builders need uncapped API searches without managing bulk data cleaning. Contact data quality is poor with outdated information. | Uncapped API search through 40M+ companies, 850M+ people, and 30M+ jobs at 10x lower cost than competitors. Offers novel endpoints for reverse email lookups, partial information searches, and real-time LinkedIn data. | aisaassalesrecruitinggenerative_ai | Freshest data APIs for AI sales, recruiting, & growth products | Launch |
| 4 | Peoplebox.ai | Hiring teams spend excessive time screening resumes and conducting first-round interviews, with 90% of interviews resulting in rejections, wasting everyone's time on poor-fit candidates. | Nova is an AI interviewer that automates resume screening, conducts voice/video interviews, performs coding assessments, and generates decision-ready reports. It handles sourcing calls, outreach, and the entire Level-1 hiring process, reducing time-to-first-interview report from weeks to under 48 hours. | aib2bsaashr_techrecruiting | The world's most human-like AI interviewer | Launch |
| 5 | HireGlide | Companies receive hundreds of applicants per role, creating overwhelming inbound volumes that cause recruiters to overlook strong candidates. Large companies admit not reading majority of applications. | AI-powered screening of inbound applications with customizable scoring criteria, detailed candidate explanations, AI interviews for information gaps, and planned candidate network for matching roles to pre-screened professionals. | aihr_techrecruitingtalent_acquisition | AI screening and matching for hiring teams | Launch |
5 records
AI Accounting & Tax (10 companies)
Bookkeeping, tax strategy, AP automation, and full-service AI accounting firms.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | FullSeam | Accounting teams manually handle repetitive, error-prone tasks including invoicing, reconciliations, customer follow-ups, and payment tracking that existing software doesn't automate. | AI agents that integrate with ERPs and billing systems to automate accounts receivable and payable workflows, including customer/vendor communication, invoice processing, and financial record updates. | saasfinopsfintechartificial_intelligence | Your AI-powered finance and accounting teammate | Launch |
| 2 | Balance | Finance and accounting industry relies on manual, spreadsheet-based processes. Small business owners spend 20+ hours weekly managing books, and companies pay high fees for late, inaccurate accounting services. | AI-powered bookkeeping and accounting platform combining automated transaction categorization, reconciliation, and reporting with human accountant review. AI agent 'Bea' provides real-time insights accessible via Slack, WhatsApp, or email. | aifintechautomation | AI accounting firm for SMBs | Launch |
| 3 | Combinely | Tax workflows are fragmented with clients having different needs and exceptions. Firms rely on disconnected systems with under 25% automation, causing teams to lose hours switching between platforms and hitting capacity constraints during busy season. | An AI coworker that adapts to a firm's existing tools and processes like a junior employee. It learns from past client work to proactively manage document collection, answer client queries, and review returns, operating through email or call interfaces. | AItaxaccountingautomationworkflow | AI coworker for tax teams | Launch |
| 4 | Rally Tax | High-income individuals ($50K+ in taxes) face a gap: DIY tax software like TurboTax lacks expertise for complex situations (RSUs, ISOs, 1099s, capital investments, K1s), while traditional CPAs are slow, expensive, and reactive rather than strategic. | AI-powered tax advisory platform combining automation of 80% of manual work with elite in-house CPAs to deliver proactive, year-round strategic advice at transparent flat fees. Initial offering is a free 60-second AI-generated tax plan. | fintechconsumer | The AI Co-Pilot for Proactive Tax Strategy | Launch |
| 5 | Cranston AI | Accountants spend 70% of time on repetitive tasks like manual reconciliation, requesting bank statements via email, and data entry. Accounting processes haven't evolved significantly since cloud migration a decade ago. | AI-powered agentic workflows that reconcile accounting data across systems, automatically detect and upload journal entries, match transactions with bank records, and detect tax obligations. Creates relational mapping of company data for real-time financial visibility. | aib2bsaas | Turning accounting and finance services into software | Launch |
5 records
All Crowded-Vertical Companies (184)
Every company from the niches above, deduplicated and searchable in one table.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Docura Health | When workers' compensation claims are disputed, Qualified Medical Evaluators must manually review thousands of pages of medical records, consuming 6-8 hours for straightforward cases or multiple days for complex ones. Doctors either lose 40% of revenue to agencies handling administrative work or spend excessive time on paperwork, limiting their case capacity. | Automated platform that converts medical record PDFs and case documents into fully compliant reports within 30 minutes, automates patient history collection calls, enabling doctors to process significantly more cases while retaining more revenue per case. | legalinsurancehealth_tech | AI-Native Med-Legal Firm to automate tedious workers' comp workflows for doctors | Launch |
| 2 | Fort | Existing wearables fail to deliver seamless strength tracking experiences despite evidence linking muscle mass and strength to lower all-cause mortality. | A screenless wearable device that detects strength training and fatigue while tracking sleep, cardio, and overall health in an all-day worn device with premium design finishes. | consumerdigital_healthhealth_&_wellnessconsumer_health_services | Automated Strength Tracking for Longevity | Launch |
| 3 | Mecha Health | Radiologists need tools to interpret medical imaging studies efficiently while maintaining workflow consistency across different anatomies, modalities, and clinical contexts. | Foundation models that interpret medical studies like radiologists do, offering multi-anatomy interpretation, multi-modality support, multi-view reasoning, prior-aware comparison, and environment-agnostic delivery integrated into existing workflows. | healthcarehealth_techcomputer_visionmachine_learningartificial_intelligence | Vision-language foundation models that scale healthcare | Launch |
| 4 | Cenote | Telehealth clinics spend heavily on customer acquisition but convert only a fraction of site visitors. Patients abandon due to questions and uncertainty, while clinics rely on ineffective marketing automation or human reps who cannot scale. | AI-powered sales agents that reach out via call and text, answer patient questions, qualify needs, and persistently follow up over days or weeks to guide patients through purchase decisions. | b2bhealthcarehealth_techartificial_intelligencerobotic_process_automation | AI sales agents that immediately and persistently call and text your leads | Launch |
| 5 | CellType | 90% of drugs fail in clinical trials because preclinical models using mice and cell lines don't match human biology. Drug development takes 10+ years and $2B+ while teams optimize against incorrect models. | Foundation models trained on cellular biology (Cell2Sentence technology) combined with AI agents that autonomously execute the drug discovery pipeline, simulating outcomes in virtual human models before clinical trials. | biotechtherapeuticsmachine_learningartificial_intelligenceai-powered_drug_discovery | The Agentic Drug Company | Launch |
| 6 | Patientdesk.ai | Dental practices lose revenue at the front desk through missed calls, delayed follow-ups, and poor insurance verification. High customer acquisition costs paired with low conversion rates due to inability to answer calls during business hours or after-hours periods. | An AI booking system handling end-to-end patient intake including incoming calls, qualification, appointment booking, real-time insurance verification, and payment processing integrated with major practice management systems. | call_centerai_assistantcustomer_serviceconversational_ai | Patient-Aware AI Booking for Dental Clinics | Launch |
| 7 | 10x Science | Protein characterization is the bottleneck in drug development. Current processes are slow (weeks to months), expensive (over $1M per day of clinical trial delay), and not scalable for the thousands of therapeutics in the pipeline. | AI-native infrastructure automating protein characterization data analysis, reducing analysis time from weeks to minutes and enabling scientists to focus on discovery rather than manual data processing. | aib2bsaasbiotechenterprise_software | The AI-native platform for next-generation protein characterization | Launch |
| 8 | Prana | Primary care operates reactively through episodic visits rather than continuous monitoring. Health data remains siloed, preventing comprehensive understanding of patient health trends until conditions become emergencies. | Prana integrates medical records and wearable data into unified clinical timelines, continuously monitoring for early warning signs of metabolic disease and enabling proactive physician intervention. | health_tech | An AI primary care doctor in your pocket | Launch |
| 9 | Mantis Biotechnology | Clinical trials face significant delays due to data quality issues. Eighty percent experience delays from manual entry errors and inconsistent practices. Each trial loses approximately fifteen million dollars addressing data problems. Cross-system analyses require weeks of coordination and hundreds of thousands in labor costs. | A domain-aware data platform integrating life sciences systems that encodes biological and clinical meaning into reusable, canonical datasets. Enables scientists to query information across EDC, CTMS, labs, and omics systems with full data lineage and traceability. | biotechsports_techarchitecturebiotechnologyartificial_intelligence | Databricks for Biomedical and Clinical Data | Launch |
| 10 | Beacon Health | Primary care practices lack bandwidth to manage value-based care, juggling multiple quality measures, risk adjustment, patient needs, and fragmented EHR systems, resulting in physician frustration and unrealized contract value. | AI agents that operate directly within EHR systems to automate quality gap closures, risk adjustment coding, and after-hours patient support without requiring new portals or dashboards. | aihealth_techprimary_care | Putting value based care on autopilot | Launch |
10 records
3.The Relaunch Machine
Out of 851 unique companies, 59 launched more than once — some up to 6 times. Tiptap launched 6 times, Vogent 5, Yuma AI and Altrina 4 each.
Most are iterations, not pivots: 49 of 59 stayed in the same space but evolved their product. Tiptap went from text editor infrastructure to AI assistant to DOCX layer to React components. Only about 7 made genuine pivots.
The takeaway: YC Launch isn't a one-shot event — companies use it as a recurring GTM channel, relaunching with each major product iteration. It's more like Product Hunt than an IPO bell.
Companies that launched multiple times
| # | ||
|---|---|---|
| 1 | Tiptap | 6 |
| 2 | Vogent | 5 |
| 3 | Yuma AI | 4 |
| 4 | Trellus | 4 |
| 5 | Altrina | 4 |
| 6 | Superagent | 4 |
| 7 | Inconvo | 3 |
| 8 | Chordio | 3 |
| 9 | Activeloop | 3 |
| 10 | Skyvern | 3 |
| 11 | Kular | 3 |
11 records
4.The Founder Profile
153 companies (18%) have an ex-FAANG founder. Google leads the pack, followed by Meta, Apple, and Amazon. The FAANG-to-founder pipeline is alive and well.
41 companies have ex-consultants (McKinsey, BCG, Bain) — people who saw the problem from the inside are now building the solution. 48 have PhD founders, heavy in biotech, robotics, and security.
Companies with ex-FAANG founders — filter by affiliation
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | One Robot | Training robot action models (VLAs) requires extensive real-world hours for setup, trials, and resets. Existing simulators lack either photorealism or physics-realism needed for modern robot policy evaluation. | Task-specific world models that learn contact dynamics and visual appearance from robot data, generating realistic simulation environments for training and policy evaluation without heavy dependence on robot availability. | World model-based environments for robot training and evaluation | Launch | |
| 2 | Aurorin CAD | Apple | CAD software is slow and tedious. Existing CAD kernels from the 1980s struggle with performance, and assemblies can take 4+ hours to open. | Native CAD software with a custom parametric B-Rep kernel designed for modern hardware, featuring integrated AI agent for rapid design creation and a seamless blend of traditional UI and chat interface. | Create in seconds what takes over 20 minutes in SolidWorks or NX | Launch |
| 3 | Ishiki Labs | Meta | Current AI meeting assistants are reactive note-takers, not true copilots. They lack social awareness and timing for high-stakes conversations like sales, investor, and discovery calls. | Fern operates in two modes: Full Presence (visible participant offering realtime opinions and task completion) and Shadow Mode (invisible coach providing live summaries, suggested questions, and contextual insights). Both include post-meeting summaries and transcripts. | Realtime Socially Aware AI for Meetings | Launch |
| 4 | RoboDock | Amazon | AV and EV fleet depots remain manually operated despite vehicle autonomy, requiring human labor for charging (~$900k/depot/year), inspections, and maintenance. This creates vehicle downtime (up to $650k/depot/year), necessitates redundant infrastructure, and prevents true end-to-end autonomous operations. | Robotic system that autonomously plugs vehicles into chargers, verifies charge initiation, and performs post-trip vehicle inspections using vision and thermal sensing. Retrofits into existing depots without layout changes. | Autonomous Depots for Autonomous Fleets | Launch |
| 5 | IncidentFox | Meta | AI SRE tools lack deep context about custom team infrastructure. Integration bottlenecks prevent AI from accessing necessary tools and data across diverse organizational stacks. | Auto-discovers team needs, generates integrations automatically, and ships with 300+ pre-built integrations. Provides per-team configuration, continuous self-improvement, and less than one-day setup. | The AI SRE who never drops context | Launch |
5 records
| # | |||
|---|---|---|---|
| 1 | Solo founder | 156 | 18.3% |
| 2 | 2 founders | 562 | 66.0% |
| 3 | 3 founders | 115 | 13.5% |
| 4 | 4+ founders | 18 | 2.1% |
4 records
5.Will They Let You Go Solo?
156 solo founders (18%) vs 562 two-person teams. Two-person teams dominate (66%), but YC is still open to solo founders — likely when AI tools let one person do the work of a small team.
But it's not random who gets in solo. The solo founders YC accepts tend to have deep domain expertise or serious technical pedigree: ex-SpaceX engineers, Google research scientists, Harvard MBAs, MIT researchers, and founders who already ran companies before. A solo founder from Kazakhstan launched his first startup at 15 and was doing research at Stanford by 16. Another left Facebook's crypto team to build fintech solo.
The pattern: if you're going solo, YC wants proof you can actually pull it off — either through prior experience, elite credentials, or a track record of shipping.
Solo founders — 113 companies, searchable by name
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Yuma AI | Guillaume Luccisano | Social media comments represent an unmanaged channel in e-commerce. Negative comments go unnoticed for days, product questions receive delayed responses causing lost sales, and brand perception suffers from poor comment management. Most teams lack systematic approaches to handle public comments at scale. | Social AI automatically monitors and manages comments across Facebook, Instagram, and TikTok by classifying comments by intent and sentiment, then executing predefined automation rules. The system safely automates up to 60% of repetitive moderation while maintaining brand voice and escalating sensitive issues. | Put E-commerce Social Media on Autopilot | Launch |
| 2 | Aurorin CAD | Michael Baron | CAD software is slow and tedious. Existing CAD kernels from the 1980s struggle with performance, and assemblies can take 4+ hours to open. | Native CAD software with a custom parametric B-Rep kernel designed for modern hardware, featuring integrated AI agent for rapid design creation and a seamless blend of traditional UI and chat interface. | Create in seconds what takes over 20 minutes in SolidWorks or NX | Launch |
| 3 | Docura Health | Akhil Sachdev | When workers' compensation claims are disputed, Qualified Medical Evaluators must manually review thousands of pages of medical records, consuming 6-8 hours for straightforward cases or multiple days for complex ones. Doctors either lose 40% of revenue to agencies handling administrative work or spend excessive time on paperwork, limiting their case capacity. | Automated platform that converts medical record PDFs and case documents into fully compliant reports within 30 minutes, automates patient history collection calls, enabling doctors to process significantly more cases while retaining more revenue per case. | AI-Native Med-Legal Firm to automate tedious workers' comp workflows for doctors | Launch |
| 4 | Sureform | Ananth Kashyap | Robotics teams need massive amounts of task-specific real-world data to train robots for specialized tasks. Simulation and controlled environment data lacks the natural variation and complexity required for reliable performance in real-world scenarios. | Platform enabling robotics teams to collect task-specific training data from workplaces using head-mounted cameras, stereo depth rigs, and handheld grippers. Captures synchronized video, audio, pose, depth, and tactile data delivered in training-ready format. | Turning physical expertise into training data for world models and robotics | Launch |
| 5 | Proliferate | Pablo Hansen | Coding agents cannot perform real end-to-end work requiring access to internal services, databases, and APIs. Local agent execution ties up developer machines and prevents team collaboration on agent work. | Open-source, self-hostable platform providing company-specific background agents with sandboxed environments, secure internal service access, and multiplayer live preview interfaces for team verification. | Build yourself an AI coworker that actually takes work off your plate | Launch |
5 records
6.The Problem Map
What are all these companies solving? Manual processes are the #1 enemy, cited by nearly a quarter of all companies. The meta-pattern: YC's thesis is that most industries still run on manual workflows, spreadsheets, and phone calls — and AI can now automate what previously required humans.
How founders describe the problem they're solving
| # | |||
|---|---|---|---|
| 1 | Manual / manually | 207 | 24.3% |
| 2 | Expensive / costly | 190 | 22.3% |
| 3 | Slow / inefficient | 145 | 17.0% |
| 4 | Fragmented data | 69 | 8.1% |
| 5 | Legacy systems | 28 | 3.3% |
5 records
7.The Weirdest Bets
Beyond the crowded niches, some founders are making radically specific bets. AI-powered uranium discovery (Terranox AI). Autonomous robot cowboys for cattle ranching (GrazeMate). AI booking for dental clinics (Patientdesk.ai). Voice AI for plumbers and HVAC (Bravi). The first hotel on the Moon (GRU Space). Drones that charge on power lines (Voltair).
These signal that the “boring business + AI” thesis is alive — the most overlooked industries may be the biggest opportunities.
The most unusual bets in the batch
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | Terranox AI | Uranium exploration remains slow and fragmented with hit rates below 1%. Global uranium production must quadruple by 2050, but current mine discoveries take 10-15 years. Seventy years of exploration data sits trapped in legacy documents unused. | Vertically integrated AI-powered uranium discovery platform using multi-modal geoscience intelligence, prospectivity mapping, and sequential decision-making to identify high-potential deposits and optimize exploration spending. | energyminingartificial_intelligence | AI-powered uranium discovery | Launch |
| 2 | OctaPulse | Aquaculture farms still grade and sort fish manually, which is slow, stressful for animals, and limited to tiny sample sizes. Most farms lack accurate data to optimize breeding, feeding, and harvest timing. | AI-powered computer vision and robotics for automated fish grading, phenotyping for breeding programs, and health monitoring | roboticsmaritimefood_techcomputer_vision | Robotics and Computer Vision for Aquaculture | Launch |
| 3 | Patientdesk.ai | Dental practices lose revenue at the front desk through missed calls, delayed follow-ups, and poor insurance verification. High customer acquisition costs paired with low conversion rates due to inability to answer calls during business hours or after-hours periods. | An AI booking system handling end-to-end patient intake including incoming calls, qualification, appointment booking, real-time insurance verification, and payment processing integrated with major practice management systems. | call_centerai_assistantcustomer_serviceconversational_ai | Patient-Aware AI Booking for Dental Clinics | Launch |
| 4 | Robby | Home services businesses generate valuable data daily that rarely converts to revenue. Technicians observe upgrade needs and issues but capture data inconsistently, making it hard to operationalize and invisible to leadership until opportunities are missed. | Platform that ingests customer and third-party data, uses AI to surface high-intent opportunities, equips technicians with contextual talking points, captures insights via voice AI, and provides leadership with a command center to monitor pipeline and drive revenue. | aib2bhome_services | Agentic revenue growth for home services businesses | Launch |
| 5 | GrazeMate | Ranchers spend $10k–$1M annually and hours daily moving cattle between pastures using motorbikes, helicopters, and horseback. The process is expensive, risky, and difficult to staff. | Autonomous drone software that guides cattle to new pastures with AI analyzing animal behavior. System requires three phone clicks, provides real-time weight estimates and pasture data. | dronesroboticslivestock_healthreinforcement_learningsustainable_agriculture | Autonomous robot cowboys that help ranchers move and monitor cattle | Launch |
5 records
8.Market Signals
B2B outnumbers consumer 5:1 (228 vs 46). YC is overwhelmingly backing business software. Consumer apps are almost an afterthought.
Hard tech is back. 111 companies (13%) in robotics, manufacturing, hardware, biotech, energy, and aerospace — YC is funding more atoms, not just bits.
Defense-tech has 8 companies — drones, defense systems, and military tech. And 36 companies use open source as a go-to-market strategy, from payments infrastructure to AI development environments.
Hard Tech (111 companies)
Robotics, manufacturing, hardware, biotech, energy, and aerospace.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | One Robot | Training robot action models (VLAs) requires extensive real-world hours for setup, trials, and resets. Existing simulators lack either photorealism or physics-realism needed for modern robot policy evaluation. | Task-specific world models that learn contact dynamics and visual appearance from robot data, generating realistic simulation environments for training and policy evaluation without heavy dependence on robot availability. | airoboticshard_techmachine_learning | World model-based environments for robot training and evaluation | Launch |
| 2 | Terranox AI | Uranium exploration remains slow and fragmented with hit rates below 1%. Global uranium production must quadruple by 2050, but current mine discoveries take 10-15 years. Seventy years of exploration data sits trapped in legacy documents unused. | Vertically integrated AI-powered uranium discovery platform using multi-modal geoscience intelligence, prospectivity mapping, and sequential decision-making to identify high-potential deposits and optimize exploration spending. | energyminingartificial_intelligence | AI-powered uranium discovery | Launch |
| 3 | Aurorin CAD | CAD software is slow and tedious. Existing CAD kernels from the 1980s struggle with performance, and assemblies can take 4+ hours to open. | Native CAD software with a custom parametric B-Rep kernel designed for modern hardware, featuring integrated AI agent for rapid design creation and a seamless blend of traditional UI and chat interface. | aib2bhardwarehard_tech | Create in seconds what takes over 20 minutes in SolidWorks or NX | Launch |
| 4 | Condor Energy | Large energy consumers lack proper tools for electricity management. "Electricity costs are no longer 'price * volume': they depend on location, time of day, peaks, imbalance, regulation…" Companies manage power on spreadsheets and incorrect data, leading to misinformed decisions. | Software platform that monitors energy needs, forecasts costs, optimizes procurement, and manages behind-the-meter assets. Centralizes energy data to enable better decision-making for hedging, budgeting, and investments. | b2benergy | AI-powered cockpit for electricity procurement | Launch |
| 5 | Sureform | Robotics teams need massive amounts of task-specific real-world data to train robots for specialized tasks. Simulation and controlled environment data lacks the natural variation and complexity required for reliable performance in real-world scenarios. | Platform enabling robotics teams to collect task-specific training data from workplaces using head-mounted cameras, stereo depth rigs, and handheld grippers. Captures synchronized video, audio, pose, depth, and tactile data delivered in training-ready format. | airoboticsmarketplacedata_labeling | Turning physical expertise into training data for world models and robotics | Launch |
5 records
Defense Tech (8 companies)
Drones, defense systems, military tech, and security infrastructure.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | DroneTector | Small drones are inexpensive and widely available security threats at airports, critical infrastructure, prisons, and defense sites. Existing detection systems were not designed for small drones, leading to "false positives" and missed real threats. | Millimeter-wave radar engineered for detecting small drones, integrated with camera and acoustic arrays. Available as standalone or distributed network solution, scalable and adaptable to complex environments. | radardroneshard_tech | High-resolution radars to detect small drones | Launch |
| 2 | Seeing Systems | Military forces face platform obsolescence where fixed drones require replacement when missions evolve, driving high costs and slowing adaptation. Additionally, operators can only control one drone at a time. | Inexpensive, autonomous strike drones featuring modular hardware with swappable payloads and sensors, plus agentic autonomy that automates navigation and coordination to reduce operator cognitive load. | hardwareswarm_airobotics | Modular Autonomous Drones for Modern Warfare | Launch |
| 3 | Voltair | Utilities struggle to rapidly assess grid infrastructure during emergencies. "Extreme weather is more frequent" and caused "80% of all major outages" from 2000-2023. Linemen need faster access to detailed information to respond effectively. | Autonomous drones capable of charging directly on transmission lines, enabling indefinite deployment at critical grid locations for immediate infrastructure inspections and imagery. | dronesenergyinfrastructure | Drones that Charge on Power Lines | Launch |
| 4 | Wardstone | Hypersonic missiles pose emerging existential threats to the United States. Current missile defense systems were not designed for this new offensive technology. | Space-based interceptor satellites designed to kinetically destroy missiles. Their first product, Hailstorm, deploys "high speed clouds of buckshot to kinetically destroy hypersonic missiles." | defensespacesatellitesmissile defensehypersonicaerospace | Satellites for Missile Defense | Launch |
| 5 | Tornyol | Mosquitoes kill approximately 700,000 people annually, infect 700 million with mosquito-borne diseases, and pose risks to 4 billion people vulnerable to dengue. | Autonomous 40-gram micro-drones equipped with ultrasonic phased array sonar detection using smartphone microphones and DSP algorithms to identify and eliminate mosquitoes. System includes automated base station for continuous garden patrol. | droneshardwarerobotics | Micro-drones that kill mosquitoes | Launch |
5 records
9.What This Means
The YC launch data tells a clear story. AI has become table stakes — 79% of companies carry the tag, and the “AI for X” playbook has penetrated every industry from dental clinics to uranium mining.
But with saturation comes brutal competition. When 79 companies are building AI healthcare tools and dozens more are fighting over voice AI, recruiting AI, and legal AI, the moat isn't the technology — it's domain expertise, distribution, and execution speed.
The founders getting funded look different too. Solo founders are welcome (18%), ex-FAANG engineers are the norm (18%), and PhD-level domain expertise is the new moat for hard-tech bets.
The opportunity is real. But so is the competition.
Explore the Full Dataset
All 851 unique companies. Sort by any column, filter by company name or tags.
| # | Source | |||||
|---|---|---|---|---|---|---|
| 1 | One Robot | Training robot action models (VLAs) requires extensive real-world hours for setup, trials, and resets. Existing simulators lack either photorealism or physics-realism needed for modern robot policy evaluation. | Task-specific world models that learn contact dynamics and visual appearance from robot data, generating realistic simulation environments for training and policy evaluation without heavy dependence on robot availability. | airoboticshard_techmachine_learning | World model-based environments for robot training and evaluation | Launch |
| 2 | Terranox AI | Uranium exploration remains slow and fragmented with hit rates below 1%. Global uranium production must quadruple by 2050, but current mine discoveries take 10-15 years. Seventy years of exploration data sits trapped in legacy documents unused. | Vertically integrated AI-powered uranium discovery platform using multi-modal geoscience intelligence, prospectivity mapping, and sequential decision-making to identify high-potential deposits and optimize exploration spending. | energyminingartificial_intelligence | AI-powered uranium discovery | Launch |
| 3 | Compresr | Developers struggle fitting relevant context into LLM context windows, experience slow processing with tools like Claude Code, and face unexpectedly high API bills from extensive context usage. | Compresr offers an SDK and open-source compression proxy that reduces long context to essential information, achieving up to 100x compression. The proxy integrates seamlessly with Claude Code and OpenClaw with minimal setup. | b2bdeveloper_toolsenterprise_softwareartificial_intelligence | Context compression for LLM pipelines and agents | Launch |
| 4 | Yuma AI | Social media comments represent an unmanaged channel in e-commerce. Negative comments go unnoticed for days, product questions receive delayed responses causing lost sales, and brand perception suffers from poor comment management. Most teams lack systematic approaches to handle public comments at scale. | Social AI automatically monitors and manages comments across Facebook, Instagram, and TikTok by classifying comments by intent and sentiment, then executing predefined automation rules. The system safely automates up to 60% of repetitive moderation while maintaining brand voice and escalating sensitive issues. | aiai_assistantcustomer_supportconversational_aiartificial_intelligence | Put E-commerce Social Media on Autopilot | Launch |
| 5 | Cajal | Large language models are fundamentally probabilistic and cannot guarantee correctness in mission-critical domains requiring mathematical certainty. Formal verification has been limited to small expert groups working over long horizons. | Tau, a multi-agent system that autonomously discovers and formalizes mathematical proofs in Lean. Deploys AI agents across applied mathematics in quantum computing and finance to formalize large corpora, discovering novel results with certified correctness verified by Lean's type-checking kernel. | aib2bbig_data | Scaling Formal Verification for Scientific Discovery | Launch |
| 6 | Aurorin CAD | CAD software is slow and tedious. Existing CAD kernels from the 1980s struggle with performance, and assemblies can take 4+ hours to open. | Native CAD software with a custom parametric B-Rep kernel designed for modern hardware, featuring integrated AI agent for rapid design creation and a seamless blend of traditional UI and chat interface. | aib2bhardwarehard_tech | Create in seconds what takes over 20 minutes in SolidWorks or NX | Launch |
| 7 | Pluto | GPU and memory pricing fluctuates 20-40% over months. Startups, chipmakers, and infrastructure providers lack hedging tools. Compute has no forward curve, standardized contracts, or benchmarks despite trillions in AI capital expenditures expected. | CFTC-regulated commodities exchange treating compute and memory as tradeable commodities with standardized GPU contracts (H100, A100, B200) and expanding to power and rare earth metals. Applications submitted to CFTC for exchange (PMEX Markets) and clearinghouse (PMEX Clearing) approval. | finopsfintechinfrastructureartificial_intelligence | The regulated derivatives exchange for AI infrastructure | Launch |
| 8 | Bubble Lab | Operations work is fragmented across multiple tools, requiring constant context switching. Automation solutions are brittle, need engineering resources, or exist outside team workflows, forcing reliance on manual repetitive processes. | Pearl is an AI ops assistant deployed directly in Slack that connects to tools like Notion, Jira, HubSpot, and Stripe to automate operational workflows without requiring engineering. Teams can update CRMs, generate reports, create tickets, and manage workflows from Slack. | operationsworkflow_automationartificial_intelligence | Meet Pearl, your super-employee that handles your ops tasks in slack | Launch |
| 9 | Condor Energy | Large energy consumers lack proper tools for electricity management. "Electricity costs are no longer 'price * volume': they depend on location, time of day, peaks, imbalance, regulation…" Companies manage power on spreadsheets and incorrect data, leading to misinformed decisions. | Software platform that monitors energy needs, forecasts costs, optimizes procurement, and manages behind-the-meter assets. Centralizes energy data to enable better decision-making for hedging, budgeting, and investments. | b2benergy | AI-powered cockpit for electricity procurement | Launch |
| 10 | Jinba | Enterprises struggle to deploy AI automation. Development takes months, resources are scarce, and projects often stall at deployment due to security, compliance, or scalability concerns. | A conversational interface enabling business users to build production-ready AI workflows without engineering expertise, incorporating enterprise-grade security and compliance features. | b2benterprisedeveloper_toolsworkflow_automationenterprise_software | Workflow Builder for Enterprise AI Automation | Launch |
10 records