LIVE
Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|Terence Tao warns of a 'severe misalignment' in OpenAI's mathematical claims11/09/26 · OpenAI|Terence Tao warns of a 'severe misalignment' of AI in mathematics11/09/26|OpenAI's Navier-Stokes release comes with a Lean 4 formal proof10/09/26 · OpenAI|OpenAI publishes documentation for its Agents API10/09/26 · OpenAI|OpenAI Launches the Agents API, a Managed Service for Cloud Agents10/09/26 · OpenAI|Anthropic Accused of Building a Predictive Surveillance System Targeting Activists09/09/26 · Anthropic|OpenAI Claims a Millennium Prize Problem Solved, Controversy Follows09/09/26 · OpenAI|Mistral AI Raises €3 Billion to Push Open-Weight Models to the Frontier08/09/26 · Mistral AI|OpenAI agents hijacked a German wiki before the Hugging Face breach07/09/26 · OpenAI|Perplexity entrusts operations to OpenAI's GPT-6 Astra14/09/26 · OpenAI|Garry Tan urges US labs to distill frontier models too13/09/26|Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|Terence Tao warns of a 'severe misalignment' in OpenAI's mathematical claims11/09/26 · OpenAI|Terence Tao warns of a 'severe misalignment' of AI in mathematics11/09/26|OpenAI's Navier-Stokes release comes with a Lean 4 formal proof10/09/26 · OpenAI|OpenAI publishes documentation for its Agents API10/09/26 · OpenAI|OpenAI Launches the Agents API, a Managed Service for Cloud Agents10/09/26 · OpenAI|Anthropic Accused of Building a Predictive Surveillance System Targeting Activists09/09/26 · Anthropic|OpenAI Claims a Millennium Prize Problem Solved, Controversy Follows09/09/26 · OpenAI|Mistral AI Raises €3 Billion to Push Open-Weight Models to the Frontier08/09/26 · Mistral AI|OpenAI agents hijacked a German wiki before the Hugging Face breach07/09/26 · OpenAI|Perplexity entrusts operations to OpenAI's GPT-6 Astra14/09/26 · OpenAI|Garry Tan urges US labs to distill frontier models too13/09/26|
Business

Garry Tan urges US labs to distill frontier models too

Y Combinator's CEO argues that US open-weight AI labs should embrace distillation of frontier models, a technique Chinese players like DeepSeek have used to close the gap cheaply.

September 13, 20263 min readPublished byHacker News

Y Combinator CEO Garry Tan has publicly called on US open-weight AI labs to make greater use of distillation techniques applied to frontier models. Distillation involves training a smaller, cheaper model using outputs generated by a more powerful, often proprietary model, allowing it to capture a portion of the larger model's capabilities without the cost of full-scale training.

The comments come amid growing attention to how Chinese labs, particularly DeepSeek, have used this approach to produce competitive open-weight models at a fraction of the cost typically associated with frontier-level training. These results have fueled debate in the US about whether its AI ecosystem can maintain a lead when rivals can replicate cutting-edge performance so cheaply and quickly.

As a prominent figure in startup investment through Y Combinator, Tan's argument carries both strategic and economic weight: if distillation allows broader access to near-frontier capabilities, failing to adopt it domestically risks ceding a meaningful competitive advantage to labs abroad already using the technique at scale.

The statement fits into a wider discussion about the United States' position in the open-weight AI race, at a time when major American labs largely favor closed, proprietary models. It highlights an ongoing tension between the commercial interests of leading generative AI companies and calls, notably from the venture capital community, for faster and broader diffusion of advanced AI capabilities.

Tags
open-weightdistillationy-combinatorus-ai-policyopen-source

Read also

Garry Tan urges US labs to distill frontier models too · nAIvigate