Say It So They Buy ItEpisode 6
White Box AI and the Trillion-Dollar Trust Problem
Ricky Sun studied computer science at Qinghua, built a career as a Silicon Valley IT veteran, and then spent the last seven years betting against the large language model wave entirely. His company Altipa builds real-time graph databases with ontology.
About this episode
Some founders chase the mainstream. Ricky Sun studied computer science at Qinghua, built a career as a Silicon Valley IT veteran, and then spent the last seven years betting against the large language model wave entirely. His company Altipa builds real-time graph databases with ontology, and its sister company AI Checks puts that foundation in front of end users as a white-box, explainable alternative to transformer-based AI.
In this episode Mark Gordon and Ricky sit down to work through the question a lot of deep-tech founders are asking right now: how do you take a horizontal, infrastructure-level product and actually get it in front of the right buyers?
Ricky shares what it took to pick a first vertical (and get it wrong), why enterprise buyers can only think scenario by scenario rather than technology first, and the reasoning behind running two companies instead of one. Mark brings the go-to-market side, including why the best product rarely wins without a sharp, persona-specific narrative that removes the need for buyers to imagine how they'd use it.
A grounded conversation for any founder trying to sell something genuinely new to buyers who don't yet have the vocabulary for it.
Thanks to Ricky for a genuinely great conversation.
Also in the conversation
- Why unifying two competing graph database schools into one product changes the buying decision.
- How AI has collapsed onboarding time from months of engineering work to overnight builds.
- What a real GTM Clarity Score of six-out-of-ten in generative SEO actually signals.
- Where the gaps are: CRM automation and tech stack depth, even with strong scores everywhere else.
In their words
“99% of customers, they can only think from application layer, scenario by scenario, persona by persona. They don't think about technology”
Ricky Sun · 7:19
“In terms of every single query, our system's cost is about 90 even to 99% lower than the cost for a large launch model”
Ricky Sun · 27:00
“It's better to be really good at go to market and telling your story sometimes than it is to actually have the best product”
Mark G · 9:20
“Previously, we probably need an eighth engineer to build a multiple month. Now we just need AI to build maybe just for 12 hours straight overnight”
Ricky Sun · 16:44
Listen here
Questions this episode answers
Why did Ricky Sun target the financial industry first?
He believed that over the past 400 years the financial industry has always been the frontier of adopting the latest technology, so he expected banks and insurers to appreciate the product. What he underestimated was that tier one banks, regulators and exchanges are a thousand times bigger than his startup, which made it hard to even find the right person to talk to.
What is white box AI and why does it matter for enterprises?
White box AI means the inputs, process and outputs are fully traceable, auditable and explainable end to end. Ricky argues that without that explainability, enterprises cannot be comfortable making decisions, and today's large language models have exactly that problem.
What does the sales funnel actually look like for a deep tech infrastructure product?
Ricky described reaching out to roughly 1,000 people, getting maybe 50 to 100 replies, five to ten who try the product, and two or three who move into a POC and a commercial procurement process.
Why does the company run two businesses, Ultipa and AI Checks?
Ultipa is the IT infrastructure layer, the real-time graph database with ontology, and AI Checks licenses it and pays a yearly fee. AI Checks is the end-user-facing white box AI solution, so customers do not have to understand what a graph database or real-time causality search is.
How did AI change their customer onboarding?
Over the last twelve months they used AI to generate synthetic and realistic data sets, simulate customer scenarios and personas, and produce step-by-step guides. Solutions that used to take a team of engineers several months can now be built overnight, and one person can serve multiple customers globally.
What is Ricky Sun's objection to transformer-based AI?
He names two broken fundamentals: trust, because hallucinations and black box behavior make outputs unexplainable, and economics, because training and operating costs across millions of GPUs do not deliver an ROI or total cost of ownership that makes sense for a trillion-dollar industry.
Keep listening
More episodes
Turning Word Of Mouth Into A Go To Market Engine
September 23, 2026
Seeing Your Business the Way Buyers Do
September 16, 2026