What differentiates winning AI training-data providers from commodity talent marketplaces?

Jonathan Siddharth · Dec 1, 2025

Turing is not a talent marketplace; it is a company training superintelligence by supplying the data pillar to frontier labs.

A talent marketplace merely matches talent to opportunities, whereas superintelligence requires research, compute and data, and Turing owns the data pillar.

4:33 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Jonathan Siddharth · Dec 1, 2025

Frontier labs now need a data partner with research DNA that can proactively anticipate paradigm shifts, rather than a passive labelling vendor.

Paradigms change fast — a year ago nobody was discussing reinforcement learning, then o1 shipped in December and DeepSeek launched in January and now everything is RL environments.

12:18 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Jonathan Siddharth · Dec 1, 2025

The era of data labelling companies is over and it is now the era of research accelerators.

Labs want a proactive partner that can reason about which types of data will help the models and make recommendations, plus a platform of the world's smartest humans and domain experts to build RL environments.

13:50 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Jonathan Siddharth · Dec 1, 2025

The data provisioning market will reward players with deep AI research capability and the ability to adapt quickly

The pace of AI research is so rapid that the dominant paradigm shifts within a year — RL environments only spiked in the last twelve months after o1 and DeepSeek, and a year from now it could be something totally different

55:30 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Brendan Foody · Jun 1, 2026

The ability to train models and hire the best researchers in the world is one of the largest things differentiating Mercor and Surge from other data providers

Both companies have executed on staying close to research, which he sees as the key differentiator in the category

69:34 20VC: Mercor CEO on Why Application Layer Companies Have No Defensibility, The Model is the Product | Token Spend Will Exceed Headcount Spend in 5 Years | The True Cost of Hiring AI Researchers in the Valley Today with Brendan Foody

Brendan Foody · Sep 15, 2025

RL environments are the new human-data type that all the labs are moving toward, distinct from the RLHF bucket where Surge leads

Scope: Mercor does comparatively little RLHF work

55:56 20VC: Mercor: From $1M to $500M in 17 Months: The Fastest Growing Company in the World | How to Think About Margins and Revenue Sustainability in AI | Why Evaluation Benchmarks in AI are BS Today with Brendan Foody

Joelle Pineau · Nov 3, 2025

The bigger trend in the data market is the shift from labeling data to crafting environments that produce new tasks, rather than vendors adding implementation services

Scope: she agrees some implementation/integration work is also happening

35:13 20VC: Cohere's Chief AI Officer on Why Scaling Laws Will Continue | Whether You Can Buy Success in AI with Talent Acquisitions | The Future of Synthetic Data & What It Means for Models | Why AI Coding is Akin to Image Generation in 2015 with Joelle Pineau

Jonathan Siddharth · Dec 1, 2025

The move from chatbots to agents requires a totally different kind of training data than the SFT and RLHF used to train chatbots.

Agents execute multi-step workflows and call functions in real business settings, so they must be trained in mini world models where trajectories can be verified; reinforcement learning is now the dominant paradigm.

5:28 20VC: Scale, Surge, Turing, Mercor: Who Wins & Who Loses in Data Labelling | Is Revenue in Data Labelling Real or GMV? | Why 99% of Knowledge Work Will Go and What Happens Then? | Why SaaS is Dead in a World of AI with Jonathan Siddharth @ Turing

Edwin Chen · Jul 21, 2025 · hedged

A lot of the other companies in the data-labeling space are not really technology companies but body shops, or body shops masquerading as technology companies

Scope: about competitors in Surge's space

0:00 20VC: Scaling to $1BN+ in Revenue with No Funding: Surge AI | The Most Insane Scaling Story in Tech |

Edwin Chen · Jul 21, 2025 · hedged

A lot of companies in the data-labeling space have no technology at all — no way to measure or improve data quality — and are effectively body shops passing recruited workers along to AI labs

They have no platform or way to measure or improve data quality; they recruit warm bodies off resumes and pass them to frontier labs, so they cannot A/B test algorithms, tools or worker-screening methods

Scope: about competitors in the human-data space, unnamed

13:25 20VC: Scaling to $1BN+ in Revenue with No Funding: Surge AI | The Most Insane Scaling Story in Tech |

Edwin Chen · Jul 21, 2025

Customers ultimately want high quality data and do not want to work with 'body shops', which is why demand shifted to Surge en masse

The space is large and many teams were on Scale only for legacy reasons; when they moved they already knew who the biggest and best provider was

35:55 20VC: Scaling to $1BN+ in Revenue with No Funding: Surge AI | The Most Insane Scaling Story in Tech |

Brendan Foody · Sep 15, 2025

The data market is transitioning away from the crowdsourcing paradigm of low- and medium-skilled labelers toward a sourcing-and-vetting paradigm of elite domain experts — Goldman bankers, McKinsey analysts, FAANG engineers, top doctors and lawyers.

Early LLMs needed people who could write barely grammatical sentences; today's frontier work requires experts who can build the highest-complexity data on Earth and work directly with researchers.

10:43 20VC: Mercor: From $1M to $500M in 17 Months: The Fastest Growing Company in the World | How to Think About Margins and Revenue Sustainability in AI | Why Evaluation Benchmarks in AI are BS Today with Brendan Foody

Brendan Foody · Sep 15, 2025

The method of improving models has shifted from shovelling large volumes of low-caliber data to curated, thoughtfully built datasets from extremely high-caliber people.

This transition toward environments and high-complexity data is what underpins Mercor's trajectory.

14:22 20VC: Mercor: From $1M to $500M in 17 Months: The Fastest Growing Company in the World | How to Think About Margins and Revenue Sustainability in AI | Why Evaluation Benchmarks in AI are BS Today with Brendan Foody

Brendan Foody · Sep 15, 2025

A proprietary referral-based supply base plus matching infrastructure that routes 10x experts to the right opportunities is extremely difficult for competitors to recreate.

Because outcomes are power law, finding and correctly placing the 10x contributors creates outsized customer value that can't be easily copied.

15:15 Proprietary referral network and matching infrastructure is the moat

Brendan Foody · Sep 15, 2025

Mercor's business sits at the intersection of labor marketplaces and AI research, unlike crowdsourcing companies that hide their workers and pay low rates.

It pairs a core competency in finding world-class people with deep research partnership with frontier labs.

16:48 Transparent high pay labor marketplace integrated with research partnership

Brendan Foody · Sep 15, 2025

The single most important factor in this market is having phenomenal people and treating them extremely well, which drives referrals and real frontier improvement.

Mercor's average marketplace pay rate is $95/hour versus roughly $30/hour at Scale and Surge, reflecting a radically different approach to talent.

22:32 Treating contributors exceptionally well drives referrals and frontier quality