Autonomy, Supervised
Hey there, investors and innovators!
The current YC batch keeps circling back to a question most software companies haven't had to ask before: what happens once the thing you built starts acting on its own, in the real world, without you standing over its shoulder? This week's crop of launches answers it from a few different angles...coordinating memory so autonomous agents don't lose the plot between sessions, generating the synthetic training ground a robot needs before it's trusted with a real one, laying down the physical infrastructure that lets AI reach places fiber and cell towers never will, enforcing policy on money that moves itself, and automating the local hustle that used to require a human doing outreach one relationship at a time. Different industries, same underlying question: once autonomy is real, who, or what, keeps it accountable, supplied, and connected?
What the Pattern Says
- Autonomy needs infrastructure before it needs applications. The most interesting launches this week aren't the flashy end-user product — they're the connective tissue underneath: shared memory so a person and their AI collaborator stay in sync across sessions, a policy layer that checks every autonomous financial action before it executes. Founders are increasingly building the plumbing that makes it safe to let AI act independently, not just the AI that acts.
- Synthetic data is becoming the unlock for physical AI. Training a robot or an embodied model used to mean filming — expensively, narrowly, and never in enough environments. This batch includes founders treating that constraint as solvable: generate the thousands of environments you could never film, and evaluate a model in worlds that don't exist yet. That reframes a data-collection problem as a generation problem, which changes the economics of building for the physical world.
- The frontier is going geographically and physically wider. Between infrastructure built explicitly to reach where existing networks don't, and AI aimed at coordinating hyperlocal, real-world relationships between businesses and creators, this batch isn't just building for people already online — it's building toward the parts of the world and the economy that are still hard to reach.
- Enforcement, not just intelligence, is the product. Whether it's money moving on-chain or an agent taking action inside a codebase, several founders this week are selling the checkpoint, not the capability — the thing that says yes or no before an autonomous action becomes irreversible. That's a meaningfully different pitch than "our AI is smarter," and it's showing up more and more.
- Batch-wide context: S26 is tracking the shape of its record-setting predecessor — YC's 2026 batches are running roughly 60% AI, up from ~40% two years ago, with B2B continuing to dominate. What's notable this week isn't that AI shows up everywhere; it's how much of it is aimed at making autonomous systems safe, supplied, and reachable, rather than simply more capable.
On the Radar
Demo Day: September 10, 2026. Expect the pace of launches to pick up meaningfully through August as founders get in front of investors ahead of it.
Eight Capital tracks every YC batch, cohort to cohort. Reach out if you'd like the full running list of this batch's launches or a deeper cut on any theme above.
Fund Manager Picks
TryNearby
Local creators for local businesses
Why it's interesting
TryNearby operates a local influence network that automates word-of-mouth marketing for local businesses, with an initial focus on restaurants. The platform matches businesses with a vetted network of over 1,600 local creators who visit and post content monthly, driving discoverability and engagement across social media and review platforms. TryNearby manages creator outreach, scheduling, visits, and content tracking, enabling businesses to focus on their core operations while growing their local presence. Since launch, the platform has generated over 33 million views and completed more than 1,579 creator visits. The company was founded by Yousef Abdelfattah, a veteran of the creator economy.
Moving Atoms
Virtual Reality for Robots (World Models)
Why it's interesting
Moving Atoms, also known as Atoms, is an industrial automation startup founded by Ahmad Saroya and Shehryar Saroya, with a leadership team that includes Travis Kalanick (Uber co-founder, per public reporting). The company develops specialized robots and AI technologies to automate manual work in sectors including food, mining, and transportation. Moving Atoms operates multiple business units, including CloudKitchens, and aims to digitize and automate the physical world to usher in a new era of industrial efficiency and abundance. The company has raised $1.7 billion in funding led by Andreessen Horowitz and includes Uber as an investor. Its stated mission is: "Physical automation to transform industry and move the world."
Mosaic
Defining the frontier of multiplayer AI.
Why it's interesting
Founded in 2019 and headquartered in New York City, Mosaic is the developer of the world's first Strategic Finance platform. The company empowers finance and business leaders with real-time data analysis, collaborative financial planning, and AI-driven decision-making functionality. Mosaic targets SMB and mid-market finance leaders, aiming to transform finance teams into strategic partners within their organizations. The company is recognized for its rapid time-to-value, enterprise-grade functionality, and consumer-grade ease of use. Mosaic’s mission is: “To help companies balance high growth with capital efficiency by delivering predictive AI solutions that elevate the strategic role of finance.”
Exosat
Building a sovereign Starlink
Why it's interesting
Exosat is a satellite company developing a constellation of ITAR-free satellites, designed without U.S.-controlled parts, to enable launches from any nation and bypass traditional launch bottlenecks. The company’s satellite backhaul integrates with existing infrastructure and spectrum, providing direct-to-cell, IoT, and broadband connectivity without requiring new ground hardware. Exosat targets sovereign operators seeking flexible, scalable, and sovereign satellite networks. Founded by Edward Ge and headquartered in Palaiseau, France, Exosat is a member of Y Combinator’s S26 batch. Its mission is “to empower sovereign operators worldwide with ITAR-free satellite connectivity solutions that enable deployment of networks anywhere without dependence on any single nation.”
Verdict Machine
AI cybersecurity for financial institutions' digital assets
Why it's interesting
Verdict Machine is a regulatory compliance and governance technology company focused on blockchain environments. Founded in 2024, the team includes Liran Kogan (CPO, ex-Check Point, ex-Payoneer) and Isaac Grossman (CTO, ex-Fhenix, ex-SCRT Labs). Verdict Machine’s mission is to provide agentic systems that learn an organization’s entire onchain posture and build matching security controls, enabling institutions to control what happens next in their blockchain operations.
