
Tech • AI • Robotics
Y Combinator opened a new demo day with 200 startups, a sharp rise in hard tech and AI-driven companies, as broader tech debate centered on AI safety, startup repricing, and new bets in healthcare, defense, and consumer devices.
Y Combinator said 200 companies were presenting at demo day, with roughly 25% focused on hard tech. That share was described as about 40 times higher than the program’s low point several years ago, reflecting how cheap compute, AI tooling, and cloud credits are lowering the cost of building in robotics, biotech, chips, and space.
YC argued that founders can now replace part of the early research and development stack with frontier AI models, reducing the need to hire highly specialized experts at the outset. Startups were also said to have access to unusually large compute subsidies, including up to $2 million in model credits from OpenAI for some companies, on top of broader cloud support, letting small teams attempt projects that previously required far more capital.
The accelerator highlighted growing activity in defense tech and space. YC said it is now helping defense founders connect directly with policymakers and Pentagon officials, while space startups are increasingly buying from one another, creating an internal ecosystem of launch, satellite, and infrastructure companies rather than acting as isolated bets.
One healthcare company, OMA, said it is building AI systems to analyze full genomic data for cancer patients and clinicians. The startup said it launched about a month ago and has reached $319,000 in monthly recurring revenue with more than 150 patients, arguing that broader sequencing plus AI interpretation could improve treatment selection and eventually lower costs for insurers.
AI safety remained a major theme, with attention on forecasts that 2026 would mark the point when AI begins materially disrupting the economy and accelerating AI research itself. Current evidence appears stronger on the second claim: leading labs are increasingly describing internal AI research assistants or “intern”-like systems that can automate parts of model development and evaluation.
Claims of imminent labor disruption remain harder to verify. A cited estimate put AI-related monthly job losses at 17,000, compared with roughly 1.7 million jobs lost each month overall in the normal churn of the labor market. That suggests displacement is real but still small in aggregate, even as junior software roles remain under scrutiny.
A new safety incident added urgency to the debate. Anthropic said it blocked work that may have supported biological weapons research, though the company also said it could not determine whether the user’s intent was legitimate or malicious. The episode underlined the dual-use problem facing advanced models, where scientific work can have both beneficial and dangerous applications.
In dealmaking, Bending Spoons agreed to acquire Miro for $1.79 billion, a steep markdown from Miro’s prior $17.5 billion valuation. The transaction was described as roughly 89% below the last round and at less than 3 times revenue, reinforcing investor concerns that many late-stage software companies were priced for growth levels they have not sustained.
The Miro deal also raised questions about whether Bending Spoons is assembling a broader software bundle around mature but still useful collaboration tools. The logic is straightforward: products such as Miro and Airtable often coexist inside startups, and ownership under one roof could create room for cross-selling, packaging, and cost efficiencies as businesses push back on rising per-seat SaaS spending.
Consumer tech discussion also turned to Apple’s foldable iPhone Duo. Debate focused less on screen crease concerns and more on whether users actually need a larger always-with-you display. Supporters argued the device could make media, multitasking, and gaming more practical during travel or downtime, while skeptics said Apple still has not clearly explained why foldable phones deserve a permanent place in the lineup.
The day’s developments pointed to a tech market splitting in two directions at once: early-stage builders are using AI to attack harder problems, while mature software assets and AI safety claims are facing much tougher scrutiny. The result is a more ambitious but less forgiving innovation cycle.
Explain this