Jul 27, 2023 · 39m · no-priors
No Priors Ep. 25 | With Palantir's CTO Shyam Sankar
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Palantir CTO Shyam Sankar joins the No Priors podcast to discuss Palantir's architectural evolution, the launch of its Artificial Intelligence Platform (AIP), and how enterprise ontologies and programmatic agents operationalize AI in mission-critical environments.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 18.3% of the talking time here. How this is scored →
The hosts as informed peer 2.8 Guest teaching 3.4 Guest disagreement 1.3 The hosts pushing back 0.4
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 28:30 Rejection of the popular definition of AI agents
Shyam forcefully rejects common industry framing around autonomous agents planning arbitrarily, stating it fails completely to meet the operational reality of enterprise workflows.
Hardest push from the hosts ▶ 32:58 Sarah presses on Alex Karp's market share claimSarah directly questions the feasibility and meaning behind CEO Alex Karp's bold public statement that Palantir is targeting the entire AI market share.
Biggest teaching moment ▶ 28:30 State machine breakdown of enterprise automationShyam re-educates on how AI automation actually works in companies, explaining that businesses are explicit state machines where agents must be scoped to single state transitions rather than autonomous planning.
The host holds their own ▶ 23:52 Elad on legacy ERP and systems integrator moatsElad demonstrates sharp enterprise tech expertise, articulating how LLM tool calling undermines the implementation and integration barriers protecting vendors like SAP and Workday.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Shyam Sankar's Personal Journey and Joining Palantir | 1 | 1 | 0 | 0 | Elad opens with friendly biographical questions about Shyam's upbringing in Nigeria and his transition to Silicon Valley. The tone is entirely welcoming and conversational. | |
| The Conception of Forward Deployed Engineering | 2 | 3 | 1 | 0 | Shyam explains the concept and philosophy of forward deployed engineering (FDE), contrasting checking boxes with customer value accountability. Sarah asks a practical clarifying follow-up. | |
| Palantir's Core Architecture: Gotham, Foundry, and Apollo | 2 | 4 | 1 | 1 | Elad prompts Shyam to explain Palantir's core software platforms. Shyam provides an in-depth breakdown of Gotham, Foundry's digital twin ontology, and Apollo's autonomous multi-environment software deployment. | |
| Grounding Large Language Models with AIP and Ontologies | 2 | 4 | 1 | 0 | Shyam introduces AIP, describing LLMs as a 'stochastic genie' that requires foundational tools and dynamic ontologies to be grounded effectively in enterprise reality. | |
| AI Tool Chains and Operational Defense and Commercial Applications | 3 | 4 | 2 | 0 | Shyam dismisses chat as a limiting interface and explains real-world defense and auto manufacturing deployments where LLMs manipulate application state and operational maps directly. | |
| Model Ensembles and Evolving Beyond Chat Interfaces | 3 | 3 | 2 | 0 | Sarah and Shyam discuss commoditization of model providers and using ensembles of 'mad genius' models communicating via DSL and JSON instead of natural language chat. | |
| Programmatic Agents, System Integrations, and Enterprise Moats | 5 | 3 | 1 | 1 | Elad demonstrates strong domain knowledge by analyzing how programmatic LLM integration threatens the defensive moats of legacy systems integrators like SAP and Workday. | |
| Enterprise State Machines and Human-Agent Teaming | 4 | 5 | 2 | 0 | Shyam rejects the hype of open-ended planning agents in enterprise contexts, arguing instead for human-agent teaming bounded by specific state machine transitions. Sarah actively synthesizes and references academic research. | |
| Internal Engineering Culture and Palantir's AI Ambition | 3 | 4 | 2 | 2 | Sarah pushes Shyam on Alex Karp's ambitious claim of taking 'the entire market share of AI.' Shyam justifies the thesis around owning the connected decision web at the application layer. |
Statements from this episode (17)
Sankar: Palantir's forward deployed engineer role combines product, customer success, and engineering
“The Ford deployed engineering idea was that the people who are going to be interacting with customers in the field, we're going to be computer scientists. You could actually understand what does the product do today? What does it need to do today? How is it un…”
Palantir Foundry powered US and UK COVID vaccine distribution
“And this is the same platform that was used to build the COVID vaccine response distribution in the U S and the UK same platform that commercial companies were using to manage the supply chain crises”
Palantir operates 550 microservices released multiple times daily
“We have 550 microservices. We're releasing multiple times a day for each one of these services.”
Sankar: LLMs need external tools for specialized workflows like orbital simulation
“Like an LLM is not going to know anything about orbital simulation or weapon hearing or Predicting forward inventory, 30 days from now. It's certainly not going to do that well, but with the right tool, it's going to do that quite excellently.”
Sankar: Enterprise ontologies provide semantic compression to make LLM outputs reliable
“You can kind of think about the ontology as having this semantic layer that gives you an incredible amount of compression that you're putting into the context window and allows you to build LLM backed functions in very reliable ways.”
Sankar: The best LLM use cases offer massive upside and harmless failure
“The best use case is going to be ones where
When the LLM gets it right, there's massive upside.
And when it doesn't, it's a no op, right?”
Sankar: Chat is a limiting interface; prompts are meant for developers
“Chat is a massively limiting interface, you know, at the limit prompts are for developers.”
Sankar: Foundation model companies will struggle to adapt to enterprise multi-model ensembles
“Yeah, I think that's probably the correct direction. And I think one that, that, that model companies will have a hard time with, right? Cause I think they need to have kind of one model to rule them all or directionally that's where it needs to go.”
Sankar: LLMs can use APIs to dynamically generate enterprise user experiences
“The whole point is not to give me answers, but you change my app. And then that starts changing how you think about interacting with these things. It becomes a new UI layer. You know, kind of the most extreme version of this is like, why have any UI at all? If…”
Palantir used LLMs to cut a two-month UI feature to two hours
“It's like one of the consequences of a hack we had a number of months ago was that I had an engineer who could build a feature in a couple hours that we had previously scoped. It was on the roadmap. It was a feature that was going to take like two months and t…”
Gil: AI will disrupt legacy ERP moats built on complex data integrations
“I think there's a lot of companies where part of their defensibility is the fact that you basically had to munch specialized data or integrations, you know, that would be the SAPs of the world or different ERP systems, Workday, et cetera. Like a lot of these t…”
Sankar: Palantir targets domain experts, not data scientists, with its AI tools
“I think there's a fair amount of companies I see going after kind of let's call it the canonical data scientists as an archetype of like, I want to fine tune a model and I'm going to go do that. I see a smaller number trying to go after devs as an archetype, b…”
Sankar: Autonomous AI planning will fail in strict enterprise state machines
“The practicalities of an enterprise is there is either an implicit or explicit, and often it's kind of fifty-fifty or some combination, implicit-explicit state machine that represents that enterprise. So the idea that you're just going to have an agent that ki…”
Sankar: OpenAI's GPT-4 is currently unavailable in classified government environments
“We live in a world where we can't count on GPT-IV everywhere. Like we don't have that on classified environments, right?”
Sankar: Enterprise AI value will accrue to application layer owners
“So I, we feel like the value is really going to accrete to folks who own the application layer and the enterprises, and we're going to go after that very hard.”
Sankar: Healthcare now accounts for roughly a third of Palantir's business
“Healthcare is roughly a third of our business. It's certainly, I mean, it's probably one of the fastest growing parts of our business as well.”
Sankar: LLMs shift IT triage from alert severity to solution consequences
“In a world where the LLM can process all the alerts and give you a stage set of actions, now you're prioritizing not on the severity of the alert, but on the possible consequences of the solution. So that, that's already an improvement in the sort function, an…”