YC directory highlights AI and industrial startups; Lumius raises $4.5 million; Garry Tan details GBrain
This edition follows YC-backed companies applying automation across software, finance, warehouses, education and industrial production, while Lumius brings new funding to wearable ultrasound. The developments also highlight the infrastructure and safeguards behind deployment: GBrain separates memory and trust boundaries, and Jeeves emphasizes human review, evaluations and audit trails for financial workflows.
Sunday, Oct 11, 2026
The latest from Y Combinator
The latest from Y Combinator
YC directory spotlights AI, industrial climate, robotics, and education automation
Y Combinator’s October 2026 directory listings show startups applying automation to both digital workflows and labor-intensive physical work. OpenSEO is connecting AI agents to SEO data, while Armature is testing whether coding agents select companies’ tools.
Cedar is automating CRM updates and follow-ups after conducting extensive research with sales representatives.
The portfolio also includes physical and operational applications: Rise Reforming is developing modular biogas-to-chemicals production and says its first pilot is underway at a Chicagoland wastewater plant.
In Los Angeles, Manifold offers warehouse robots priced per pick with no upfront costs, while Advanced Metal Research is building computer-vision welding cells.
Education startups Frizzle and Miyagi Labs are applying automated grading, analytics, tutoring, and personalized practice.
The common theme is lower adoption friction: products are tied to existing workflows, specific labor bottlenecks, or faster feedback loops. The important next signals are pilot progress, customer adoption, and evidence that the products deliver the efficiency gains their descriptions promise.
Lumius raises $4.5 million seed round
Lumius announced a $4.5 million seed round led by FundersClub, with participation from Y Combinator and other investors. The company is developing accessible 3D and wearable ultrasound for clinical and everyday care.
Lumius also joined an ARPA-H-funded collaboration led by researchers from Dartmouth and Columbia focused on smart obstetric care.
The financing, research partnership, and addition of new team members give the company more resources and institutional support as it develops its medical technology.
No product launch timing or commercialization details were announced, making execution and future product updates the key signals to watch.
Garry Tan documents GBrain, a shared memory system for AI agents
Y Combinator President and CEO Garry Tan published documentation for GBrain, a persistent memory system used with his OpenClaw and Hermes AI-agent deployments. The system stores explicit facts with source records, supports corrections and withdrawals, and makes durable memory available across authorized agents.
The project’s documentation describes keyword and semantic retrieval, optional synthesis, and background enrichment while keeping transient task state local.
It also emphasizes governance concerns: remote authorization is distinct from local files and shared database credentials, stored text may be sent to configured cloud providers, and the tests are not a universal guarantee against leaks.
GBrain matters because it treats portable, provenance-aware memory as agent infrastructure rather than a feature confined to one assistant. Its usefulness will depend on whether teams can manage sharing permissions, editing rights, withdrawals, and provider access safely.
YC-backed startups advertise roles tied to AI execution and growth
Jeeves is seeking a senior AI engineer to build and improve finance workflows such as invoice-to-payment.
The role includes customer conversations, human review for high-value decisions, automated evaluations, monitoring, confidence scores, fallbacks, and audit trails—requirements that reflect the operational demands of a regulated financial product.
Ambient.ai, meanwhile, is hiring a technical recruiter. Its posting says the physical-security company doubled new annual recurring revenue in FY26 and cites customers including Cisco, ServiceNow, TikTok, Bayer, and MoMA.
The role is intended to build hiring plans and manage full-cycle technical recruiting as the company expands.
Together, the openings show two execution priorities: making AI automation reliable in customer workflows and building the recruiting infrastructure needed to support reported commercial momentum.
YC’s remote jobs board shows demand, but not uniform location flexibility
Y Combinator’s remote-startup jobs page lists openings across engineering, design, product, operations, recruiting, sales, and science, while allowing candidates to create profiles for direct founder outreach.
The listings also show that “remote” can carry different geographic requirements. Pine Park Health lists a senior software engineer role paying $190,000–$215,000 with remote work available from the Berkeley and San Francisco areas, while Popl lists a U.S.-remote senior product designer role paying $140,000–$180,000.
The board gives startups a recruiting channel beyond their immediate networks, but candidates still need to distinguish nationally remote roles from jobs tied to a particular region or hybrid arrangement.
Advice for early-stage founders
Practical playbook: learn before you scale
General guidance—not a report of Y Combinator news.
Start with the workflow, not the feature. Talk to people who regularly experience the problem. Ask what they do today, what makes the process difficult, and what they have already tried. Look for repeated pain, not merely enthusiastic reactions.
Build the smallest useful product. Choose one user, one painful job, and one clear outcome. A narrow tool that fits into an existing workflow is more valuable than a broad product that requires users to change everything at once.
Prioritize evidence over volume. Rank work by how directly it can produce learning, user value, or distribution. Defer polish, edge cases, and speculative features until users repeatedly need them.
Treat distribution as part of the product. Decide how the first users will find, trust, adopt, and recommend the product. If reaching users requires a separate, difficult process, that friction belongs on the product roadmap.
Turn feedback into decisions. After each conversation or usage cycle, record the problem, the user’s current workaround, the requested change, and what behavior would confirm improvement. Separate feature requests from the underlying need.
Close the loop quickly. Ship small changes, return to the users who raised the issue, and check whether their behavior changed. Praise is encouraging; continued use, repeated requests, referrals, and willingness to pay are stronger evidence.
Keep a learning cadence. Set a recurring review of conversations, usage, drop-offs, and objections. Choose the next experiment from what users actually did—not from the loudest opinion or the longest wish list.