Hi, very nice approach, been playing around with it in Perplexity this morning.
However, I get some really weird results upon doing: Analyze MAKO MINING. Agent pulls 'current' GOld/SLV values of Gold (XAU/USD) → $2,435/oz and Silver (XAG/USD) → $30.85/oz. Also, AISC of MAKO (refer to Q2 financials) are 1668, with agent consistently useing 1050 as AISC.
This 'dents' my trust in the model somewhat;-) Am I doing something wrong?
Can you educate me on how you came up with the context metadata model? First time I see it in this level of detail - especially interesting it features physical data, like temperature.
Yes, all I’m doing is context engineering. The context window is the temporary data container behind your prompt. For Claude Sonnet it’s like a 400 page book capacity. All the sec filings, reports and investor relations should be in the background of your session. For me it pulls 40-50 sources per query.
Hi, very nice approach, been playing around with it in Perplexity this morning.
However, I get some really weird results upon doing: Analyze MAKO MINING. Agent pulls 'current' GOld/SLV values of Gold (XAU/USD) → $2,435/oz and Silver (XAG/USD) → $30.85/oz. Also, AISC of MAKO (refer to Q2 financials) are 1668, with agent consistently useing 1050 as AISC.
This 'dents' my trust in the model somewhat;-) Am I doing something wrong?
Yeah, that happens when it doesn’t find spot and pulls the year end from SEC. For now just say ‘those prices don’t look current’ and it will correct.
what is the difference between https://github.com/5thgenfinance/5thgenfinance.github.io/blob/main/agents/live/don-durrett-chap10.md
and
https://github.com/5thgenfinance/5thgenfinance.github.io/blob/main/agents/live/perplexity-agent-architect.md ?
The later is the seed agent that will create other markdown agents for you.
it says 404 - page not found
does not contain the path
agents/live/don-durrett-chap10.md should we just copy and paste the text? Is it the same as the .md file?
I will fix thanks!
Link Fixed now!
Thanks for the share - very elaborate.
Can you educate me on how you came up with the context metadata model? First time I see it in this level of detail - especially interesting it features physical data, like temperature.
Yes, all I’m doing is context engineering. The context window is the temporary data container behind your prompt. For Claude Sonnet it’s like a 400 page book capacity. All the sec filings, reports and investor relations should be in the background of your session. For me it pulls 40-50 sources per query.
Thanks for the context - I‘ll have a go with that and will try to fiddle around with this as a base, very helpful!