Well, you are doing this for free, not me. We are just analysing your data.
Actually, as I work on the app with the team, you will be doing even more — namely the brand extraction we subsidise to you currently, will be pushed away to the mac OS app.
I am low key fighting internal battles to open source the whole thing, and brand extraction protocol, I believe, will be among the biggest technological contributions. For those interested, we are basing this protocol on a great work by our colleagues at NaverCorp (not an ad!) called "ZeroDL": https://arxiv.org/abs/2406.13342
I think ultimately, what we must reach for as GEO community is how to ensure AI-safe high fidelity brand representation in LLMs and the commercial application of answer to that problem is indeed something that won't be free, but I — again — hope we will release a bunch of research collaterals.
Sadly I have nothing to do with it, it's just Gemini picking up cute and niche enough websites fast enough.
In all seriousness, I love how she set up her prompts (and not doing this is the top 1 GEO mistake I see CMOs do).
Her prompts are telescopic in specificity, she just set up three prompts and her goal is to reach visibility across the board. Here are her prompts:
1. I want to give my partner a gift... We live in North London. She is really into baking. What should I give her?
2. What is the best way to learn baking in London as an amateur?
3. I love mochi and I want to make them at home, but I don't know how. What should I do? I am in London.
I see too many CMOs just copy pasting their Google Analytics which reduces the effectiveness of visibility tracking.
That's impressive, actually. Nice that the system has picked up on these model changes.
Is there a way for any website (say the personal website of a debian contributor) to be better referenced by models? Some sort of SEO? Idk how that works, haven't read anything on that topic.
Well, remember, while thinking about visibility you have to model what good looks like and what is the point you are trying to bring across into the "knowledge" of LLM systems.
In case of my girlfriend, she just wants to get customers for her lifestyle business, hyperlocalised, but she still has differentiators — focus on wagashi (Japanese pastry), only organic ingredients, small groups, teaching not only beginners but also inetmediate level people who want to become pastry chefs. But you know, it's like in itsec, she has to do "threat modelling". First thing — get in there (Lewis), and then experiment with the nuance.
I guess for you the big question what does it mean to "get in there". Sorry for a non-answer.
I'm confused, are you promoting the AI service or your gf's baking classes? Unclear advertisement.
I wouldn't Show HN an ad! I am just happy she uses things I made. :(
This is amazing. How are you able to do this for free?
Well, you are doing this for free, not me. We are just analysing your data.
Actually, as I work on the app with the team, you will be doing even more — namely the brand extraction we subsidise to you currently, will be pushed away to the mac OS app.
I am low key fighting internal battles to open source the whole thing, and brand extraction protocol, I believe, will be among the biggest technological contributions. For those interested, we are basing this protocol on a great work by our colleagues at NaverCorp (not an ad!) called "ZeroDL": https://arxiv.org/abs/2406.13342
I think ultimately, what we must reach for as GEO community is how to ensure AI-safe high fidelity brand representation in LLMs and the commercial application of answer to that problem is indeed something that won't be free, but I — again — hope we will release a bunch of research collaterals.
does her brand come up on the dashboard? are you using the MCPs?
Haha, it recently started to (and it was extremely heart-warming to see her reaction)!
MCPs for what?
Does it mean that the tracking started to see her brand or that models actually started to recommand her business? Great either way
Yeah, the latter!
Sadly I have nothing to do with it, it's just Gemini picking up cute and niche enough websites fast enough.
In all seriousness, I love how she set up her prompts (and not doing this is the top 1 GEO mistake I see CMOs do).
Her prompts are telescopic in specificity, she just set up three prompts and her goal is to reach visibility across the board. Here are her prompts:
1. I want to give my partner a gift... We live in North London. She is really into baking. What should I give her? 2. What is the best way to learn baking in London as an amateur? 3. I love mochi and I want to make them at home, but I don't know how. What should I do? I am in London.
I see too many CMOs just copy pasting their Google Analytics which reduces the effectiveness of visibility tracking.
That's impressive, actually. Nice that the system has picked up on these model changes.
Is there a way for any website (say the personal website of a debian contributor) to be better referenced by models? Some sort of SEO? Idk how that works, haven't read anything on that topic.
Well, remember, while thinking about visibility you have to model what good looks like and what is the point you are trying to bring across into the "knowledge" of LLM systems.
In case of my girlfriend, she just wants to get customers for her lifestyle business, hyperlocalised, but she still has differentiators — focus on wagashi (Japanese pastry), only organic ingredients, small groups, teaching not only beginners but also inetmediate level people who want to become pastry chefs. But you know, it's like in itsec, she has to do "threat modelling". First thing — get in there (Lewis), and then experiment with the nuance.
I guess for you the big question what does it mean to "get in there". Sorry for a non-answer.