
Lessons from Nick Frosst
The items below are things Nick Frosst said on podcasts about Cohere, language models and the limits of AI. — Nick Frosst on The BetaKit Podcast, Cohere's Nick Frosst on sovereign AI and Star Trek (June 2026).
On The BetaKit Podcast, Frosst said: "Really, when people talk about AI, they mean statistical models that take in text and predict more text". — Nick Frosst on The BetaKit Podcast, Cohere's Nick Frosst on sovereign AI and Star Trek (June 2026).
On Make it Click, Frosst said: "You're going to do so much better if you're working on something you are genuinely interested in if you're working on something that you think is optimal". — Nick Frosst on Make it Click (June 2026).
On Make it Click, Frosst said: "Either it's going to be a failure and that's going to be really hard, or it's going to be a success and that's going to be just as hard". — Nick Frosst on Make it Click (June 2026).
On Make it Click, Frosst said: "Your ability to notice a difference in your income or your wealth caps out pretty low". — Nick Frosst on Make it Click (June 2026).
On Make it Click, Frosst said: "Everybody is trying to figure out how to process information better, get better outcomes, get more efficiency". — Nick Frosst on Make it Click (June 2026).
On Make it Click, Frosst said: "I do think that there's lots of stuff that LLMs enable people to operate at scale in a way that they couldn't before". — Nick Frosst on Make it Click (June 2026).
On The BetaKit Podcast, Frosst said: "You can develop your own agency and autonomy without infringing on the agency and autonomy of other people". — Nick Frosst on The BetaKit Podcast, Cohere's Nick Frosst on sovereign AI and Star Trek (June 2026).
On The BetaKit Podcast, Frosst said: "We make a tech that can be a fundamental interface for a computer, and you can use it to do the things that are boring". — Nick Frosst on The BetaKit Podcast, Cohere's Nick Frosst on sovereign AI and Star Trek (June 2026).
On The BetaKit Podcast, Frosst said: "When I hear people talking about putting ads into a thing that's trying to help you think, it makes me uncomfortable". — Nick Frosst on The BetaKit Podcast, Cohere's Nick Frosst on sovereign AI and Star Trek (June 2026).
On The BetaKit Podcast, Frosst said: "We expected that this technology would be really influential and would define how people use computers". — Nick Frosst on The BetaKit Podcast, Cohere's Nick Frosst on sovereign AI and Star Trek (June 2026).
On The BetaKit Podcast, Frosst said: "There's consistently reports that come out that personify statistical models in a way that I think is disingenuous". — Nick Frosst on The BetaKit Podcast, Cohere's Nick Frosst on sovereign AI and Star Trek (June 2026).
On Terms of Service, Frosst said: "I don't think the solution to the loneliness crisis is for people to talk to language models". — Nick Frosst on Terms of Service with Clare Duffy, Could AI Become Smarter Than Humans? (February 2026).
On Terms of Service, Frosst said: "There are things in your life that you definitely should get a neural net to do, a language model to do". — Nick Frosst on Terms of Service with Clare Duffy, Could AI Become Smarter Than Humans? (February 2026).
On Terms of Service, Frosst said: "It's great to have a question and then have a model synthesize a bunch of different sources to give you an answer". — Nick Frosst on Terms of Service with Clare Duffy, Could AI Become Smarter Than Humans? (February 2026).
On Tech Disruptors, Frosst said: "I think one of the things that's really interesting about this technology is that there's no magic in large language models that isn't the data it's trained on". — Nick Frosst on Tech Disruptors, Cohere's Path to Differentiation in LLM Race (July 2024).
On Tech Disruptors, Frosst said: "I don't think there's reason to believe that just taking the same approach and scaling it up will lead to AGI". — Nick Frosst on Tech Disruptors, Cohere's Path to Differentiation in LLM Race (July 2024).
On Tech Disruptors, Frosst said: "I think hallucinations without a retrieval augmented generation system to rely on an external source of truth is a barrier to deploying these models into production". — Nick Frosst on Tech Disruptors, Cohere's Path to Differentiation in LLM Race (July 2024).
On Tech Disruptors, Frosst said: "If you're wanting to build a chatbot to answer questions about your HR situation, you don't want it to just make up policies". — Nick Frosst on Tech Disruptors, Cohere's Path to Differentiation in LLM Race (July 2024).
On Tech Disruptors, Frosst said: "When we think about how big to make a model, we're thinking about enterprise deployment". — Nick Frosst on Tech Disruptors, Cohere's Path to Differentiation in LLM Race (July 2024).
On 20VC, Frosst said: "Synthetic data has helped a lot and it's made models better than they would be if they didn't have access to it". — Nick Frosst on The Twenty Minute VC (20VC), How Cohere Compete with OpenAI and Anthropic (September 2025).
On 20VC, Frosst said: "A lot of it is still getting good quality data and then making good quality synthetic data from your good quality real data". — Nick Frosst on The Twenty Minute VC (20VC), How Cohere Compete with OpenAI and Anthropic (September 2025).
On 20VC, Frosst said: "That's one of the really interesting things about LLMs, is that they're really phenomenal, they generalize really well, but they don't generalize as well as you might think". — Nick Frosst on The Twenty Minute VC (20VC), How Cohere Compete with OpenAI and Anthropic (September 2025).
On 20VC, Frosst said: "There's a lot of misinformation, and I don't think the hype is necessarily helpful for getting to the truth". — Nick Frosst on The Twenty Minute VC (20VC), How Cohere Compete with OpenAI and Anthropic (September 2025).
On 20VC, Frosst said: "I think you will arrive to work and you will sit in front of a computer and you will predominantly use language to interact with that computer". — Nick Frosst on The Twenty Minute VC (20VC), How Cohere Compete with OpenAI and Anthropic (September 2025).
On 20VC, Frosst said: "It turns out, if you go talk to a lot of companies who wanted to deploy models into production, they were bottlenecked on deploying because they don't have enough GPUs". — Nick Frosst on The Twenty Minute VC (20VC), How Cohere Compete with OpenAI and Anthropic (September 2025).