I just skimmed the article so perhaps I missed it, but a few thoughts...
1. Unless you have a very solid idea of what you want, tell Claude your objective and ask it to brainstorm solutions.
2. Require it to explain the rationale for its recommendations (potentially including specific references to supporting materials), identify assumptions, risks, and weaknesses.
3. Take the plan, excluding the weaknesses/risks, and plug it into another LLM and ask it to critique the plan, provide rationale, etc.
4. Iteration.
5. Especially since you lack the coding skills to understand the implementation, don't let any LLM build a tool without having reviewed and approved the design. Really, really bad idea.
Did you not run into problems scraping Yelp review data? FWIW I tried something similar to track CAVA reviews but Claude Code couldn't seem to get past Yelp's scraping defenses (which are significant)
In general though I think the core idea here is spot on / insightful
I originalyl wanted to do it across all stores but needed to get yelp API to do it. I don't tihnk there were issues with NYC only stores (which is what I did)
you have convinced me to purchase Claude Co-work. It is quite fascinating. So many applications once you start thinking about it. I even had it write a research paper in your style. It is kind of fun.
This seems like a great way to generate lots of new data but I'm legitimately curious why this would lead to better investment results. If everyone has access to real time pricing data from Sweetgreen or cable it would seem that nobody has an edge to exploit and the market as a whole goes from being very efficient to extremely efficient ... unless I'm missing something?
that's the buffett "if you're at a parade and everyone stands up at their tiptoes" argument, and it's probably correc.t but i think there is alpha if you can figure out new ways to attack interesting problems, and i certainly wouldn't want to have my head in the sand while everyone else is figuring out ways to use these
It’s definitely impressive technology although corporate insiders are the ones setting product prices and there are already mechanisms for tracking their trading activities. On the other hand we know that people like Jim Simons found all kinds of predictive statistical anomalies that are exploitable by machine learning at Renaissance Tech. I will note for whatever it’s worth that the fear of falling behind, missing out, and not being with the times are good psychological hallmarks of a bubble.
I just skimmed the article so perhaps I missed it, but a few thoughts...
1. Unless you have a very solid idea of what you want, tell Claude your objective and ask it to brainstorm solutions.
2. Require it to explain the rationale for its recommendations (potentially including specific references to supporting materials), identify assumptions, risks, and weaknesses.
3. Take the plan, excluding the weaknesses/risks, and plug it into another LLM and ask it to critique the plan, provide rationale, etc.
4. Iteration.
5. Especially since you lack the coding skills to understand the implementation, don't let any LLM build a tool without having reviewed and approved the design. Really, really bad idea.
6. Read #5 again.
Great to see the experimentation. Good luck!
Did you not run into problems scraping Yelp review data? FWIW I tried something similar to track CAVA reviews but Claude Code couldn't seem to get past Yelp's scraping defenses (which are significant)
In general though I think the core idea here is spot on / insightful
I originalyl wanted to do it across all stores but needed to get yelp API to do it. I don't tihnk there were issues with NYC only stores (which is what I did)
you have convinced me to purchase Claude Co-work. It is quite fascinating. So many applications once you start thinking about it. I even had it write a research paper in your style. It is kind of fun.
This seems like a great way to generate lots of new data but I'm legitimately curious why this would lead to better investment results. If everyone has access to real time pricing data from Sweetgreen or cable it would seem that nobody has an edge to exploit and the market as a whole goes from being very efficient to extremely efficient ... unless I'm missing something?
that's the buffett "if you're at a parade and everyone stands up at their tiptoes" argument, and it's probably correc.t but i think there is alpha if you can figure out new ways to attack interesting problems, and i certainly wouldn't want to have my head in the sand while everyone else is figuring out ways to use these
It’s definitely impressive technology although corporate insiders are the ones setting product prices and there are already mechanisms for tracking their trading activities. On the other hand we know that people like Jim Simons found all kinds of predictive statistical anomalies that are exploitable by machine learning at Renaissance Tech. I will note for whatever it’s worth that the fear of falling behind, missing out, and not being with the times are good psychological hallmarks of a bubble.