Perplexity launches Frontier Research to build an automated scientist

Nov 16, 2026
The research effort, backed by more than 10,000 GPUs, will work on agents, self-generated training data and post-deployment learning.

Perplexity has launched Frontier Research, a research effort aimed at automating scientific research, according to its research site. The company describes the goal as building an automated scientist: systems of agents that search across thousands of sources, go deep on single threads, propose experiments and critique hypotheses.

Perplexity says the work needs a new science of its own, with models that generate their own training signal, keep learning after release and coordinate across devices and data centres as one system. It says it will publish papers, models, data and benchmarks, and that the effort will be backed by more than 10,000 GPUs.

Frontier Research is a research programme rather than a released product, Superpower Daily notes. Its agenda covers agent coordination, model training and post-deployment learning, reinforcement learning, computing systems and evaluation, including whether generated training data can outperform real data and how to tell scientific understanding apart from memorisation, which Perplexity calls an unresolved problem.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like Daily AI Brief.

Also, consider following us on social media:

Free newsletter

Daily AI Brief

Daily report covering major AI developments and industry news, with both top stories and complete market updates

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more
Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.