Oct 10, 2025 · 49m · sourcery
Inside The $2.2B AI Research Accelerator | Turing · Sourcery with Molly O'Shea
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
In this episode of Sourcery, host Molly O'Shea interviews Turing Founder and CEO Jonathan Siddharth to explore how Turing grew into a $2.2 billion AI research accelerator, the technical recipe for training Artificial Superintelligence (ASI), and how enterprises can build proprietary intelligence.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Molly holds 20.2% of the talking time here. How this is scored →
Molly as informed peer 3.9 Guest teaching 5.8 Guest disagreement 1.5 Molly pushing back 1.1
speaking balance: gold is Molly, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 28:24 Dismissing AI Doomerism and Defending GPT-5
Jonathan rejects the conventional doomer narrative and skepticism around recent frontier models, forcefully stating that GPT-5 is awesome and dismissing rapid takeoff theories.
Hardest push from Molly ▶ 43:25 Pushing on Sustainable Revenue vs Service RevenueMolly challenges Jonathan with classic venture capital skepticism, pressing him on how Turing can prove durable revenue when competitors like Scale and Mercor are viewed as service-based.
Biggest teaching moment ▶ 23:00 Explaining Self-Play RL in Verifiable DomainsJonathan educates the audience and host on how frontier labs shifted from human imitation learning (RLHF) to experiential self-play reinforcement learning.
Molly holds their own ▶ 32:01 Citing Kalshi Prediction Markets on Safety PausesMolly demonstrates deep market awareness by citing specific data points from Kalshi prediction markets showing odds of safety pauses falling from 40% down to 9%.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Molly as informed peer | Guest teaching | Guest disagreement | Molly pushing back | Why |
|---|---|---|---|---|---|---|
| What Is Turing & The 4 Pillars of Superintelligence | 3 | 5 | 1 | 0 | Molly opens the interview asking what Turing is following her viral tweet. Jonathan outlines Turing's role as a strategic research accelerator advancing frontier labs across the four pillars of superintelligence. | |
| Defining ASI & Automating $30 Trillion of Knowledge Work | 1 | 6 | 1 | 0 | Molly asks for a definition of ASI. Jonathan delivers an in-depth explanation using a four-dimensional knowledge work matrix representing thirty trillion dollars of automatable work. | |
| Sponsor Announcement: Brex | 5 | 6 | 2 | 1 | Molly references Nat Friedman deciphering ancient scrolls due to data exhaustion and asks how Turing differentiates among market leaders. Jonathan explains their two-by-two positioning and the post-o1 transition from data factories to strategic research accelerators. | |
| Stanford Roots, AI Research DNA & RL Gyms | 4 | 5 | 1 | 0 | Molly asks about Jonathan's Stanford research background and how it led to Turing. Jonathan explains their research DNA, early meetings with OpenAI regarding GPT-3, and their development of RL gyms. | |
| The Recipe for Frontier Models: Pre-Training vs. Post-Training | 2 | 8 | 1 | 0 | Molly asks for a step-by-step breakdown of how frontier models are built. Jonathan delivers an extensive, detailed lecture covering pre-training, supervised fine-tuning, RLHF reward models, and self-play reinforcement learning in verifiable domains. | |
| Promotional Segment: Turing Intelligence | 6 | 6 | 3 | 2 | Molly brings up David Sacks' tweet questioning AI progress and cites Kalshi prediction markets on safety research pauses. Jonathan strongly pushes back on AI doomers, defends GPT-5, and explains why rapid takeoff is a misconception. | |
| Enterprise AI: Fine-Tuning Proprietary Intelligence | 4 | 5 | 1 | 1 | Molly asks how Turing addresses enterprise adoption gaps and what verticals they serve. Jonathan explains why enterprises need custom proprietary intelligence rather than off-the-shelf general intelligence. | |
| Sponsor Announcement: Carta | 6 | 5 | 2 | 5 | Molly presses Jonathan on how he convinces investors that Turing's revenue is durable software value rather than low-multiple service revenue. Jonathan mounts a detailed defense comparing Turing to Nvidia and legacy IT consulting firms. |
Statements from this episode (19)
Siddharth: Frontier AI Training Demands Complex Reasoning and STEM Data
“The race to HGI is on, and the data needs of these models have shifted. These models need incredibly complex data for training them to get better at coding, advanced reasoning, STEM, etc.”
O'Shea: Turing Reaches $300M+ Revenue and $2.2B Valuation Profitably
“Three hundred million plus in revenue and profitable. Two hundred and twenty-five million in funding at a last valuation of 2.2 billion. A network of four million engineers and customers that include OpenAI, NVIDIA, Anthropoc, Google, Microsoft, Meta, Salesfor…”
Siddharth: Turing works with seven of the eight frontier AI labs
“We work with seven out of the eight Frontier Labs. We work with OpenAI, Anthropic, Meta, Google, Microsoft, Nvidia, Amazon, anybody that's building a Frontier Foundation model, we are probably working with them already.”
Siddharth: Public internet data for pre-training AI ran out three years ago
“These models ate the internet when they were pre-trained, but the internet data is used up. It was used up like three years ago, right?”
Siddharth: Solving multimodality, reasoning, tool use, and coding yields ASI
“Turing's the world's leading research accelerator, working with all of these frontier labs to advance them along I would say the four pillars of superintelligence which is multimodality, reasoning, tool use, and coding. If you solve these four things, you will…”
Siddharth: ASI means automating 90% of computer tasks for 90% of humans
“And I think of ASI as us having automated 90% of the tasks that 90% of humans do today in front of a computer, right?”
Siddharth: AI will automate $30 trillion of global knowledge work
“This is like, if you look at those four dimensional, if you look at that four dimensional matrix, that's 30 trillion dollars of knowledge work that is going to be automated.”
Siddharth: Reasoning models ended demand for simple AI data factories
“This industry had a huge shift after the reasoning models came out late last year, O-one being the first. The shift is, before the reasoning models, this industry needed simple data. Gobs and gobs of simple data. You needed a data factory. Or, ah, this industr…”
Siddharth: Turing Built Thousands of RL Gyms for Agent Training
“We built thousands of these reinforcement learning gyms to train agents.”
Siddharth: Turing's Agentic Coding Data Stumps All Current AI Models
“We are creating data for coding now, for agentic coding, that would stump all of today's models and agents built on top of those models.”
Siddharth: AI models will solve agentic coding benchmarks in six months
“I would give, give that maybe six months before the models climb that hill, and then we'll generate data that's even harder for the models.”
Siddharth: Verifiable domains allow self-play reinforcement learning to replace RLHF
“Now, for these verifiable domains like coding and math, instead of doing reinforcement learning with human feedback, you can do reinforcement learning. Because you can automatically check when you got the correct answer or not in these verifiable domains. And …”
Siddharth: AI will progress steadily rather than via rapid takeoff
“I don't think rapid takeoff is how things will unfold. I think it's going to be steady, continuous progress. Every step of the way, we're going to keep moving forward and it's going to be great, and it's great for a few reasons.”
Siddharth: Enterprise AI does not require 100% accuracy with human oversight
“I feel like the self-driving car industry did a lot of damage to AI in one particular way, which is it because with self-driving cars, 99% accuracy is not enough. That, that last one percent really matters. But with a lot of AI systems, if you're automating th…”
Siddharth: Enterprise workflow data does not exist on the internet
“To solve real world enterprise workflows, you need, ah, real world enterprise data, and that data doesn't exist on the internet.”
Siddharth: Enterprise AI performs best on 0.5B-to-10B parameter custom models
“It'll probably probably be a smaller model. Oftentimes we see a half a billion to a ten billion parameter model. That's the regime, not a trillion parameter model. And this will be faster, more accurate.”
Siddharth: Custom LLMs will allow investment firms to employ fewer analysts
“I think, like, the best firms are already doing this, ah, to help an investment analyst make better decisions faster, and eventually with significantly fewer people.”
Siddharth: Enterprise AI automation is at 0.25 on a 10-point scale
“If I think of a zero to 10 point scale on the consumer side, we are maybe at a three.
On the enterprise side, we are maybe at a .25 at best, right?”
Siddharth: Agentic AI will replace services firms like McKinsey, BCG, and Accenture
“So I think like Turing will probably replace companies like McKinsey, Bain, BCG, and every services company like Accenture, TCS, Wipro, Infosys, like all of these companies doing vanilla services that's going to be automated with these agentic systems.”