To get 1,000, we used SEO.
Our product helps people get ready for the Duolingo English Test, so our target audience was well-defined from the beginning.
57 karma · joined April 12, 2023
To get 1,000, we used SEO.
Our product helps people get ready for the Duolingo English Test, so our target audience was well-defined from the beginning.
Speaking as a potential user, my biggest hang up is trust. How can I trust that Roundtable’s results are accurate and not the result of hallucination?
One of the powerful things about data is that they surprise you. This is why data integrity is so important (“crap in, crap out” as the old adage goes). But if I get a surprising result from Roundtable, how can I verify it? I think you two are already thinking about this and building features to address it.
I’m also wondering if trying to verify a surprising result from Roundtable is the wrong response…Why would a LLM give me that answer? There may be something useful to understand about why the LLM is “hallucinating.” In terms of features, it may be interesting to see whether Roundtable’s LLM could explain its answer.
The UX could be like having a brilliant but inscrutable research assistant…
Another way to phrase my question would be: What are the cycles that I will likely encounter and what can I do now to be better prepared for them?
But what does “sales” look like in B2C startups, which is my current project?
I wonder how many other people using ChatGPT do not comprehend that ChatGPT can be a confident bullshitter...
I'm surprised that this one case is getting so much attention because there must be so many instances of people using false information they got from ChatGPT.
It's surprisingly easy to build something that no one wants. This is because it's easy to build something that doesn't solve a real problem. For reasons that I still can't figure out, we're very good at fooling ourselves about what are real problems that people experience.
Now when I have a new idea, I force myself to dwell on the question, "What is the cheapest way I can validate that the problem I'm solving is real?" You did that with an ugly prototype and a video. That's quite cheap. Maybe you could've even done it with no product at all!
One recommendation: Add a section on winding down a corporation. This is difficult and easy to get wrong and can be quite costly when things do go wrong :)