Dec 1, 2025 · 1h 17m · 20vc
Turing CEO Jonathan Siddharth: Who Wins in Data Labelling & Why 99% of Knowledge Work Will Disappear · 20VC with Harry Stebbings
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 the 20VC podcast, Turing CEO Jonathan Siddharth explains how his company acts as an AI research accelerator, details why enterprise AI relies on custom on-premises models, and outlines a future where knowledge work is automated and traditional SaaS is obsolete.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 17.7% of the talking time here. How this is scored →
Harry as informed peer 5.1 Guest teaching 4.6 Guest disagreement 2.3 Harry pushing back 4.7
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 44:09 Guest Rejects AI Bubble Concept
Guest strongly dismisses AI bubble narratives using emphatic language and insisting existing models are already transformative.
Hardest push from Harry ▶ 56:15 Host Refuses 'Death of SaaS' PremiseHost directly refuses the guest's claim that SaaS is dead, outlining three clear operational arguments regarding app maintenance, non-tech SMBs, and niche domain software.
Biggest teaching moment ▶ 13:15 Explaining Custom Enterprise Model ArchitectureGuest clearly explains why smaller fine-tuned on-premise models outperform trillion-parameter frontier models for specialized enterprise workflows like insurance underwriting.
Harry holds his own ▶ 16:34 Citing Enterprise Inertia Against AutomationHost demonstrates deep domain understanding of institutional paralysis, arguing that legacy enterprise processes will severely delay full AI automation timelines.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Why Turing is Not a Talent Marketplace | 4 | 5 | 2 | 3 | Host opens by challenging standard marketplace definitions and frames industry shifts toward specialized vertical data. Guest clarifies Turing's shift from talent matching to RL data generation and research acceleration. | |
| Training AI Agents with Reinforcement Learning | 2 | 6 | 1 | 2 | Guest delivers a detailed technical monologue on training agentic systems through reinforcement learning and world models. Host asks basic clarifying questions about data requirement differences. | |
| Innings One of Data Acquisition and Turing's Edge | 6 | 5 | 2 | 5 | Host brings external insider knowledge from a competitor board member regarding data acquisition timelines. Guest agrees that AI development is in 'innings one' while outlining Turing's enterprise edge. | |
| The Permanent Need for Custom Models in Enterprise | 3 | 6 | 1 | 3 | Host asks whether enterprise custom model building is temporary or permanent. Guest educates on insurance underwriting workflows where smaller fine-tuned models outperform giant general LLMs. | |
| Will All Knowledge Work Be Automated in Ten Years? | 7 | 4 | 4 | 8 | Host forcefully pushes back against total knowledge work automation within ten years by citing enterprise data paralysis and inability to adopt basic tools. Guest holds his ground by differentiating front-office from back-office timelines. | |
| Socioeconomic Impact of Budget Transitions to AI | 7 | 4 | 3 | 7 | Host challenges guest's optimism by citing Rory O'Driscoll's labor budget framework and UK workforce statistics. Guest defends his perspective using OpenAI's GDP Val research paper on task capability. | |
| Moats in the AI Era and Enterprise Schlep | 6 | 5 | 2 | 5 | Host brings in commentary from Base44's founder regarding AI code generation to question company moats. Guest explains enterprise schlep and human-AI tandem deployment models. | |
| Evaluating Revenue Quality and Competitors in AI | 5 | 4 | 4 | 8 | Host relentlessly presses guest on revenue accounting transparency and forces him to name respected competitors. Guest carefully deflects broader sector accounting comments while praising Alex Wang of Scale AI. | |
| Revenue Concentration Risks and Sovereign AI Models | 6 | 3 | 1 | 4 | Host compares Turing's revenue concentration to Nvidia's customer metrics and raises sovereign AI model requirements. Guest agrees with host's assessment on sovereign government demand. | |
| Solving the Model Capability Overhang and AI Failures | 5 | 5 | 3 | 5 | Host raises concerns about a cooling period in AI investment, offering to hire Turing for his podcast workflow. Guest strongly rejects bubble fears and outlines the model capability overhang. | |
| The Death of SaaS and Future of Coding | 8 | 4 | 5 | 9 | Host delivers a comprehensive multi-point defense of traditional SaaS apps against guest's claim that SaaS is over. Guest contends that foundation models and ambient interfaces will replace GUI-based software. | |
| AI Hardware and Envisioning Post-Smartphone Interfaces | 3 | 5 | 1 | 2 | Host asks about post-smartphone hardware interfaces. Guest paints a detailed vision of multimodal sensory wearables providing real-time feedback during conversations. | |
| Data Market Outlook and the Robotics Investing Opportunity | 4 | 5 | 1 | 2 | Host prompts guest on investment opportunities in the data ecosystem. Guest identifies physical robotics and embodied AI as the primary uncolonized frontier. | |
| Quickfire Round on Leadership, China, and AI Future | 5 | 4 | 2 | 3 | Host conducts a rapid-fire query sequence covering China, leadership changes, and AGI timelines. Guest candidly shares his personal evolution from seeking approval to hands-on detail management. |
Statements from this episode (26)
Siddharth: The era of traditional data labeling companies is over
“I think the era of data labeling companies is over, and it's now the era of research accelerators.”
Siddharth: There is no AI bubble because current models are incredibly powerful
“I don't see an AI bubble. These models are incredibly powerful today.”
Siddharth: Traditional SaaS as we know it is completely over
“SaaS as we know it, I think is over. I think it's completely over.”
Turing works with seven of the eight frontier AI labs
“We work with seven out of the eight frontier labs.”
Frontier AI training now requires domain experts over low-skilled contractors
“So it's no longer the kind of data that low-skilled, medium-skilled contractors can generate. You need expert humans in every domain.”
Siddharth: Reinforcement learning is the dominant paradigm for training AI agents
“Today, the dominant paradigm is reinforcement learning.”
Siddharth: AGI development will follow a slow, steady takeoff timeline
“It's innings one, and I believe in slow takeoff. I'm sorry to pour cold water on all the AI doomers that might be listening to this, but we are not in a rapid takeoff scenario. I believe in slow, steady takeoff for AGI and eventually super intelligence.”
Siddharth: Enterprise custom AI models are a permanent requirement
“I think it's a permanent requirement.”
Siddharth: Universal AI assistants require trillion-parameter models
“Like, I mean, if you want like a general purpose assistant, I think you need a trillion parameter model, right? Like that can answer anything to be a universal assistant.”
Siddharth: All digital knowledge work will be automated within 10 years
“The glimpse of the future that I see is that all knowledge work is going to be automated. If a human's job involves looking at a computer, analyzing what's on the screen, using different tools, using a keyboard and a mouse, it's going to be automated. It's onl…”
Stebbings: Legacy incumbents face a 10-20 year decline from slow AI adoption
“We will be on a 10 to 20 year decline of incumbents who are unable and unwilling to adopt new tools because of data, because of permissioning, because of internal buying processes.”
Siddharth: Enterprise AI adoption will be slow in back-office, fast in front-office
“In back office automation, it'll probably be very slow, and it'll, it'll probably be the upstarts that'll do things well. I think the change management will be too slow. But I'm optimistic about front office, especially in financial services, life sciences, ph…”
Siddharth: Budget transition to AI is highest in marketing and support
“I think the transfer is pretty high in areas like customer support copywriting SEO, like some of these marketing related areas.”
Siddharth: AGI will allow non-technical founders to launch startups cheaply
“In a future where AGI exists, this person will recruit a marketing GPT, a software engineer GPT, a PM GPT, and get off the ground for a lot less capital. A million flowers will bloom. Lots and lots of non-technical founders will start companies.”
Siddharth: Data-driven feedback loops will be the key moat in AI
“I think one moat will be data-driven feedback loops.”
Siddharth: AI enterprise models lag consumer deployments and need real-world usage to improve
“I feel like the models have touched reality in consumer. We haven't yet touched reality in enterprise. And the only way we'll improve is by deployment.”
Siddharth: 39% of Nvidia's revenue comes from just two clients
“Nvidia, for Nvidia 39% of their revenue comes from two clients, right? And roughly 50% was like four clients.”
Siddharth: Governments will build sovereign internal AI models
“I think it would make sense for governments to build their own internal versions of some of these models, which would require proprietary data again to be collected.”
Siddharth: Current AI models suffer from a significant capability overhang
“And there's a very significant model capability overhang. By that, what I mean is the models are capable of X, but what we are getting out of the models is X minus Delta.”
Siddharth: There will be more software engineers in ten years than today
“I think it'll, I think there'll be more software engineers because if you define a software engineer as somebody who's capable of building a software product to solve a real problem, that pool of builders is going to expand way beyond people who've graduated w…”
Siddharth: Future hardware will be always-on devices processing multimodal tokens
“We'll have some type of a device that is, that we'll carry that's always on and processing multimodal tokens.”
Siddharth: AI data market will consolidate into a few main winners
“I think there'll be a few winners. I think there'll be a few winners. A few because I do think for the labs, it helps them to have a few partners for resiliency. And I imagine also for price competitiveness. I think there'll be a few winners.”
Siddharth: Vertical AI data is crowded, but robotics remains wide open
“The vertical stuff, like, I mean, we are scaling up pretty massively in generating data for different verticals. So I don't see that as like a big white space. But I think everybody is relatively early with robotics. And robotics is such a vast realm that ther…”
Siddharth: Chinese open-source AI models like DeepSeek and Qwen are state-of-the-art
“I think it's very impressive, like the progress that they've made in open source with DeepSeek Kimi Ketu, Kuen. These models are state of the art.”
Siddharth: Developers should keep frontier AI model technology closed for safety
“I feel like for frontier models, there is some value in keeping some of the technology closed.”
Siddharth: Founders must stay close to ground truth rather than delegating
“I used to believe that to build a Enduring, valuable company. You hire a strong, ah, exec team and operate with a lot of leverage. Basically hire strong people, hire great people and get out of the way. I used to believe that. Now I believe you hire great peop…”
Shorts cut from this episode
▶ What Elon Musk Taught me about Leadership · 20VC with Harry (@1:12:26)
▶ This is why Cursor works... · 20VC with Harry Stebbings (@47:27)
▶ ___ is going to be Wonderful for Entrepreneurs... · 20VC wit (@25:54)
▶ All Knowledge Work is going to be Automated 🤖 · 20VC with H (@16:30)
▶ "SaaS is completely OVER" · 20VC with Harry Stebbings (@0:00)