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

Turing CEO Jonathan Siddharth argues data labeling is dead, RL environments are the new frontier, and 99% of knowledge work will be automated within a decade.

• Turing works with 7 of 8 frontier AI labs; Scale AI’s acquisition by the US government flooded Turing with new lab demand overnight.
• The data paradigm has shifted from simple labeling to building RL environments simulating real enterprise workflows across every industry, function, and role — a $30T knowledge-work target.
• OpenAI’s GDP-val paper found today’s best models matched human expert output ~50% of the time on single-step tasks; Claude Opus 4 ranked #1, GPT-5 close behind.
• Siddharth believes in slow, steady AGI takeoff — not rapid — arguing incremental improvement delivers value at every step unlike self-driving cars’ last-1% problem.
• SaaS is structurally threatened on three fronts: companies building custom apps on LLMs themselves, foundation model companies moving into the app layer, and agentic models replacing GUI-driven workflows entirely.
• Nvidia gets ~39% of revenue from 2 clients and ~50% from 4 — Siddharth sees Turing’s frontier-lab concentration as comparable and not a red flag given the scale of Stargate-level compute spend.
• Data-driven feedback loops, not technology, will be the enterprise moat: whoever deploys first discovers model failures first and compounds improvements fastest.
• Siddharth’s biggest belief reversal: stopped delegating to exec layers and now operates flat, staying close to engineers and customers as the source of ground truth.

2025-12-01 · Watch on YouTube