This Week in Startups
Neurosymbolic AI outperforms chatbots and product search | E2327
· Summarized
Key takeaways
- Neurosymbolic AI models like Ontology 1 offer a transparent, cost-effective alternative to traditional LLMs by learning in real-time and providing explainability, making them highly suitable for personalized product searches and trust-dependent applications.
- The space industry is undergoing a transformation driven by decreasing launch costs, faster development cycles, and emerging refueling infrastructure, reducing traditional multi-decade project timelines.
- In-space refueling technology is becoming a critical bottleneck and commercial opportunity, with companies like Spacium developing rapid, precise docking and refueling hardware to enable longer missions and broader exploration.
- Post-SpaceX IPO interest has revitalized investment and innovation in space, shifting the industry toward rapid hardware iterations and increased commercialization.
- A lean, passionate startup culture, especially in tech hubs like San Francisco, can accelerate space tech development far faster than traditional, bureaucratic approaches.
Summary
Introduction to Neurosymbolic AI and Onton’s Innovation
The episode begins with a focus on how frontier AI models, like chatbots and search engines, excel at tasks such as PhD-level exams but struggle with personalized product searches due to their black-box nature. Zach Hudson from Onton introduces their neurosymbolic model, Ontology 1, designed to understand individual tastes and aesthetics over time. Unlike traditional models, it learns in real-time, updates without retraining, and can explain why results are returned, which enhances trust and transparency. They demonstrate its application in product and aesthetic searches, showing how it recursively refines search results based on user preferences.
The Power and Future of Neurosymbolic Models
Zach explains that neurosymbolic models combine logical reasoning with neural networks, offering more efficiency, cost-effectiveness, and trustworthiness than large language models (LLMs). These models are particularly suited for e-commerce and design, where understanding nuances and personal taste is crucial. The discussion covers whether such models could lead to artificial general intelligence (AGI), emphasizing their verifiable, update-in-real-time nature and lower costs. The conversation highlights that traditional LLMs are less effective for trust-critical tasks, which is where neurosymbolic approaches excel.
Application of Ontology 1 in E-Commerce and Design
Onton’s product enables highly tailored searches for interior design, fashion, and manufacturing. Zach discusses how their technology can understand complex, multi-object, multi-category searches—such as matching furniture and decor—reflecting how humans think about style. He notes that the model's ability to learn from each interaction enhances its accuracy and relevance over time. They plan to expand into other categories, leveraging the model's ability to quickly adapt and improve, and they target both commercial and defense markets.
Operational Insights and Business Model
The episode touches on how Onton built their hardware and software rapidly, with in-house testing and a lean team of seven engineers, enabling fast iteration. They have already flown their robotic actuator in space, demonstrating the precision required for docking and refueling spacecraft, which is critical for future orbital refueling stations. The company is fully funded for upcoming missions, with multiple commercial contracts and billions in LOIs, reflecting strong market demand. They also discuss the importance of industry standardization for spacecraft docking and refueling interfaces, which is forthcoming.
Market and Industry Perspectives
Post-SpaceX IPO, interest and enthusiasm for space startups have increased, with more investors paying attention and seeing commercial opportunities in space infrastructure like refueling in orbit. Zach notes that the shrinking costs of launch vehicles and rapid technological advances allow startups like theirs to move quickly and execute missions in a fraction of the traditional timelines. The industry is changing from slow, multi-decade projects to rapid, hardware-focused iterations.
Culture and Location of the Startup
Onton chose San Francisco over traditional space hubs like Southern California because of its vibrant, energetic startup ecosystem, which encourages ambitious, innovative work. Zach emphasizes their startup’s 'weird' culture, characterized by intense work hours, rapid testing, and hacking solutions for cost reduction and speed, driven by a passionate, obsessive team. This culture supports their rapid development cycle and ambitious mission pipeline.
Sponsors
- Vanta · Vanta makes sure startups are audit ready and secure, automating security and compliance to help prove trust. Over 16,000 companies use Vanta, including Ramp, Cursor, Harvey, and Robinhood. Get $1,000 off for a limited time at www.vanta.com/slash/twist.
- Agree.com · Stop chasing invoices and automate your entire contract to cash stack. Visit Agree.com and tell them Jason sent you to get 50% off for life.
- YSecurity · YSecurity offers on-demand security with 40+ expert engineers, matched to your needs by the hour, with your first 6 hours free. Perfect for startups wanting enterprise-grade security without hiring a full-time CISO.
Show notes
This Week In Startups is made possible by:
Vanta: https://www.vanta.com/twist
Agree: https://agree.com
YSecurity: https://ysecurity.io/TWIST
Today's show:
Frontier AI models can ace PhD-level exams, but it's still bad at tracking down the product you want in the style that suits you. Onton's Zach Hudson tell us that the problem is that models are a black box.
His solution? A neurosymbolic model, Ontology 1, that learns about your taste and preferred aesthetic over time, then produces product searches tailored specifically to you, rather than just using keywords and relevant tags. How do neurosymbolic models work, and how does Onton understand your prompts and favorite design trends? And why aren't the frontier labs working on neurosymbolic models of their own? Zach joins Jason and Lon to discuss.
PLUS, following a record-smashing SpaceX IPO, Ashi Dissanayake of Spacium makes the case that the real bottleneck in space isn't launching rockets off the ground any more. It's refueling in orbit.
Guests
Zach Hudson on X: ****https://x.com/nosduhz
Onton: https://onton.com/
Spacium: https://spaceium.com/
Spacium on X: https://x.com/SpaceiumInc
Relevant Links
Poolside's journey to AGI: https://poolside.ai/vision/purpose
Startup Archive: Sam Altman on the Paul Graham advice that saved OpenAI: https://www.startuparchive.org/p/sam-altman-on-the-paul-graham-advice-that-saved-open-ai-always-make-an-api
Kelly Wearstler: https://www.kellywearstler.com/
Ennis House: https://franklloydwright.org/site/ennis-house/
Los Feliz Living: Ennis House profile: https://www.losfelizliving.com/los-feliz-historic-homes/ennis-house-los-feliz-hcm-149
Indiewire: Ennis-inspired "The Studio" offices: https://www.indiewire.com/features/craft/the-studio-production-design-interview-seth-rogen-1235114365/
Monocle Magazine: https://monocle.com/
Spaceium on Y Combinator: https://www.ycombinator.com/companies/spaceium-inc
Orbit Fab's RAFTI: https://www.orbitfab.com/rafti/
James Webb Space Telescope: https://science.nasa.gov/mission/webb/
Houzz: https://www.houzz.com/
Timestamps: 0:00 Zach Hudson joins: What is "neurosymbolic search"
4:03 Ontology 1 isn't a black box
6:31 Who is using Onton?
9:36 Vanta - Get $1000 off your SOC 2 at https://www.vanta.com/twist
11:35 Could neurosymbolic models reach AGI?
13:19 Jason loves Wright's Ennis House
17:03 The shape of AI companies is changing
19:28 Agree.com - Stop chasing invoices and automate your entire contract-to-cash stack. Go to https://agree.com and tell them Jason sent you to get 50% off for life!
22:32 UGC as a data moat
26:03 The Dead Internet Theory
28:13 Ashi Dissanayake of Spacium joins
29:52 YSecurity - The on-demand security team for startups. Need enterprise-grade security without hiring a $400k CISO? YSecurity gives you 40+ expert engineers, matched to exactly what you need, by the hour, with your first six hours completely free. Go to https://YSecurity.io/TWIST
31:50 Storables vs. cryogenics: the zero-boil-off breakthrough
34:11 All kinds of propulsion requires refueling
36:09 Getting more value from LEO to GEO
38:28 Moving at rocket speed
44:54 Why demand is so acute
49:37 The investing climate for space, post-SpaceX
Subscribe to the TWiST500 newsletter: https://ticker.thisweekinstartups.com
Check out the TWIST500: https://www.twist500.com
Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp
Follow Lon:
Follow Jason:
LinkedIn: https://www.linkedin.com/in/jasoncalacanis
Check out all our partner offers: https://partners.launch.co/
Great TWIST interviews: Will Guidara, Eoghan McCabe, Steve Huffman, Brian Chesky, Bob Moesta, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland
Check out Jason's suite of newsletters: https://substack.com/@calacanis
Follow TWiST:
Twitter: https://twitter.com/TWiStartups
YouTube: https://www.youtube.com/thisweekin
Instagram: https://www.instagram.com/thisweekinstartups
TikTok: https://www.tiktok.com/@thisweekinstartups
Substack: https://twistartups.substack.com
Follow this show on AudioBrevity and new episodes get summarized for you. Create a free account