Jonathan Siddharth

CEO of Turing, an AI research accelerator and custom model provider for frontier labs and enterprises.

Claims

As speaker

Mentioned in

Key positions and views

The era of simple data labeling is over, replaced by a need for 'research accelerators' that provide complex, strategically engineered data and reinforcement learning environments for training advanced AI agents.

The traditional SaaS business model is obsolete and will be 'sonic boomed' by foundation model providers incorporating agentic capabilities, making data-driven feedback loops the new competitive moat.

AGI and superintelligence will emerge from a 'slow, steady takeoff' driven by continuous progress in scaling laws, rather than a rapid, explosive event.

For most enterprise applications, smaller, fine-tuned language models (0.5B-10B parameters) are superior in speed and accuracy to massive, general-purpose models.

The ultimate trajectory of AI is the complete automation of all knowledge work, which will eventually render traditional consulting and IT services firms obsolete.

Podcast consensus on Siddharth

Points of consensus

▶Both sources confirm Siddharth's company, Turing, works with seven of the eight major frontier AI labs, specifically naming OpenAI, Anthropic, Meta, Google, Microsoft, Nvidia, and Amazon.Apr 2026

▶Siddharth consistently argues that the era of simple 'sweatshop data labeling' is over and has been replaced by the need for 'research accelerators' that can handle complex data and reinforcement learning environments.Apr 2026

▶Across both appearances, Siddharth expresses the view that AGI and superintelligence will arrive via a 'slow, steady takeoff' rather than a rapid, explosive event, emphasizing continuous progress.Apr 2026

▶He consistently posits that for specific enterprise tasks, smaller language models (in the 0.5B to 10B parameter range) are often faster and more accurate than massive, trillion-parameter models.Apr 2026

Points of debate

▶The provided claims are all from Jonathan Siddharth and show a high degree of internal consistency; there are no disagreements or contrasting perspectives presented about him between the sources.

▶Siddharth's position that the traditional SaaS business model is 'completely over' represents a significant point of debate against the established software industry, which largely still operates on this model.Mar–Apr 2026

▶His advocacy for smaller, specialized models for enterprise use cases contrasts with the dominant public narrative and investment focus on building ever-larger general-purpose foundation models.Apr 2026

▶His 'slow takeoff' theory for AGI is a direct counterpoint to the 'rapid takeoff' or 'foom' scenarios debated within the AI safety and futurist communities.Apr 2026

Key themes

▶Obsolescence of Incumbent Business ModelsMar–Apr 2026

Siddharth repeatedly declares the end of established paradigms. He argues that the business models for data labeling companies, SaaS, and traditional consulting firms (like McKinsey and Accenture) are being rendered obsolete by the rise of agentic AI and foundation models.

Investors should scrutinize companies in the SaaS and IT services sectors for their vulnerability to being 'sonic boomed' by foundation model providers, as Siddharth suggests their moats may be less defensible than previously thought.

▶The Evolution of AI TrainingApr 2026

Siddharth outlines a significant shift in how advanced AI is developed. He claims the internet data used for pre-training was exhausted years ago, forcing a move towards strategically engineered data and reinforcement learning in complex simulated environments, which is the core of his 'research accelerator' thesis.

This theme highlights that the key bottleneck and competitive advantage in AI development is shifting from raw compute and public data to the sophisticated generation of high-quality, proprietary training data and environments.

▶Pragmatic Enterprise AI Adoption

While bullish on AI's ultimate potential, Siddharth presents a pragmatic view of its current enterprise application. He notes the high failure rate of generative AI pilots and advocates for using smaller, custom-tuned models for specific tasks like insurance underwriting, arguing they outperform larger models in both speed and accuracy.

Analysts should be wary of the hype around trillion-parameter models as a one-size-fits-all solution; the real enterprise value may lie in the less glamorous but more efficient application of specialized, smaller models.

▶The Inevitable Automation of Knowledge WorkApr 2026

A core tenet of Siddharth's worldview is the eventual and total automation of knowledge work, which he quantifies as a $30 trillion opportunity. He defines superintelligence as the automation of 90% of tasks for 90% of computer-based workers and sees this as a steady, continuous progression rather than a sudden event.

This long-term vision underpins his company's strategy to replace entire industries, suggesting that the total addressable market for agentic AI is not just a segment of the economy, but the knowledge economy itself.

Source episodes

Sentiment over time

Mar 2026

2 bullish(2 claims)

The winner of the superintelligence race will likely dominate multiple major mar...

Apr 2026

8 bullish, 1 bearish, 9 neutral(18 claims)

An MIT study found that 95% of generative AI pilots in enterprises fail.

Changes over time

Approx. 3 years ago

Siddharth claims the internet data used for pre-training large language models was exhausted, marking a critical turning point in AI development.

Pre-Reasoning Models Era

Describes the AI data industry as being dominated by 'sweatshop data labeling,' focused on providing simple, large-scale datasets.

Post-Reasoning Models Era

Following the release of models like O1 and DeepSeek, Siddharth notes a significant industry shift away from simple data to needing complex, strategically engineered data for reasoning and agentic tasks.

Present Day

Siddharth's company, Turing, is expanding its customer base from frontier AI labs to large enterprises and is transitioning its own organizational structure from fully distributed to a hub-and-spoke model with physical offices.

Suggested prompts

How does Siddharth's 'research accelerator' model fundamentally differ from traditional data labeling, and what are its key dependencies for success? &nearr;What are the second-order effects on the software industry if Siddharth's prediction that the SaaS model is 'completely over' proves true? &nearr;How does Siddharth's 'slow takeoff' theory for AGI reconcile with his prediction that $30 trillion of knowledge work will be automated in the foreseeable future? &nearr;Given the high failure rate of enterprise AI pilots, what specific strategies does Siddharth's approach with smaller, fine-tuned models offer to improve the success rate for companies like Disney and BlackRock? &nearr;

Key concepts

AI Development & Training 2 ep Business Model Disruption 2 ep Turing's Business Strategy 2 ep AGI & Superintelligence 2 ep Enterprise AI Adoption 2 ep Knowledge Work Automation 2 ep AI Model Performance & Scaling 2 ep Market Analysis 1 ep

Notable quotes

“[0:00:10 -> 0:00:13] All knowledge work is going to be automated. [0:00:13 -> 0:00:14] It's only a matter of time.”

Jonathan Siddharth · Turing CEO Jonathan Siddharth: Who Wins in Data Labelling & Why 99% of Knowledge Work Will Disappear

“[0:00:20 -> 0:00:22] SaaS as we know it, I think it's over. [0:00:22 -> 0:00:24] I think it's completely over.”

Jonathan Siddharth · Turing CEO Jonathan Siddharth: Who Wins in Data Labelling & Why 99% of Knowledge Work Will Disappear

“[0:00:00 -> 0:00:02] I think the era of data labeling companies is over. [0:00:02 -> 0:00:05] And it's now the era of research accelerators.”

Jonathan Siddharth · Turing CEO Jonathan Siddharth: Who Wins in Data Labelling & Why 99% of Knowledge Work Will Disappear

“[0:46:08 -> 0:46:18] So I think like Turing will probably replace companies like McKinsey, Bain, BCG, and every services company like Accenture, TCS, Wipro, Infosys,”

Jonathan Siddharth · Inside The $2.2B AI Research Accelerator | Turing

Report last updated: Apr 21, 2026

Get started free

Back to Entities Intelligence Report