RATLOG joins Y Combinator after prototype test; Musk says he experienced “AI psychosis” roughly six times
This edition tracks ambitious automation projects moving from prototypes and customer claims toward funding, deployment and technical tests, from RATLOG’s simplified rocket engine to YC startups in regulated and physical industries. It also examines the human costs of that acceleration, as Musk describes repeated “AI psychosis” episodes and a coding podcast confronts unequal pressure on developers to retrain.
The latest from Y Combinator
The latest from Y Combinator
RATLOG bets on radical rocket-engine simplification
German rocket startup RATLOG has joined Y Combinator’s Fall 2026 batch after testing its first prototype in early September. The Dresden-based company says it is redesigning rocket engines around roughly 90% fewer parts and is targeting specifications comparable to Falcon 9, potentially with a wider fairing.
RATLOG plans to hot-fire a flight combustion chamber in November and demonstrate commercial traction at YC Demo Day in December. It says it has letters of intent from orbital data-center and constellation companies and is exploring Oman and the UAE for launch sites.
The significance is in the timing of the next proof points: the November test and December customer milestone will show whether RATLOG’s fast, simplified approach can move beyond an early prototype toward a credible heavy-lift business.
YC directory highlights practical AI and automation startups
Y Combinator’s October 15–16, 2026 directory listings feature startups working in compliance, logistics, warehouse robotics and healthcare software. Their descriptions focus on narrowly defined operational products: TovenAI says six AI agents connect with more than 30 systems for compliance workflows; Aviern says it built an autonomous scaled aircraft in two months; and Floracene says a first customer put four internal healthcare tools into production within weeks.
The listings point to a practical startup pattern: solve a specific workflow, integrate with existing systems and show evidence of deployment or customer use. They establish how the companies position themselves, but do not independently validate the performance claims. Future hiring, customer and deployment milestones will provide stronger signals.
Three YC-backed startups turn funding into deployment plans
Readily raised $11 million in Series A financing led by Define Ventures, with Y Combinator and the California Health Care Foundation Innovation Fund participating. The healthcare company says its platform serves organisations covering more than 25 million Americans.
Construction robotics startup Origin raised a $20 million seed round to manufacture and deploy 150 autonomous robots by the end of 2027. Meanwhile, Cosmochute raised $1.5 million in pre-seed funding for a modular lunar rover and transport system and is establishing engineering laboratories at IIMA Ventures’ facility.
These deals matter because the funding is tied to concrete next steps: Readily plans to expand enterprise compliance automation, Origin is financing manufacturing and customer growth, and Cosmochute is moving toward prototypes, engineering hires and future lunar missions. For YC, the development is a portfolio financing outcome rather than a new accelerator announcement.
Advice for early-stage founders
Advice for early-stage founders
General guidance—not a report of YC news
Start with the customer’s existing behavior. Talk to people who recently experienced the problem. Ask what they do today, what they have tried, how often the problem occurs, and who decides whether to pay for a solution. Compliments are weak evidence; repeated behavior, time spent, budget, and willingness to try a product are stronger signals.
Build the smallest useful end-to-end product. Pick one customer, one painful job, and one clear outcome. A narrow product that reliably completes that job is more informative than a broad roadmap. Manual work behind the scenes is acceptable if it helps you learn what should eventually be automated.
Prioritize learning as well as shipping. Rank work by customer pain, frequency, ease of reaching the user, and how much the task will teach you. Choose one meaningful outcome for the week, and defer requests that do not improve the core workflow or answer an important uncertainty.
Treat distribution as part of the product. Identify where your target users already work, search, communicate, or buy. Test founder-led outreach, partnerships, communities, or direct introductions before investing in a larger acquisition engine. A product that is easy to find and adopt can outperform a more capable product that requires customers to discover it unaided.
Use feedback to improve decisions, not just features. Watch users complete the task whenever possible. Separate requests from underlying problems, group repeated issues, and look for patterns across users rather than reacting to the loudest individual. Review feedback regularly, make a small change, and return to the same users to see whether the problem actually improved.
The goal is a tight loop: find a painful problem, build a focused solution, reach the right users, observe what happens, and update your priorities based on evidence.