Snippets and real life examples of (mis)using AI in the dark corners of the junior mining sector☕
Scanning the market
There is no drought in available podcast, interviews & presentation in this sector. Resource Talks alone produced over 1000h of content to go through. The actual interesting ones, I still listen too in the car to catch all the nuances AND getting a handle on the managers involved. But for most, and especially company updates, I boil it down like this;
Free online tool I use: https://www.youtube-transcript.io/
Taking that one step further are integrated browsers (like COMET here) that do this task on the spot:
This works pretty well as most company updates have a clear question/answer set-up. For interviews with guests, the highlights are just a teaser to evaluate if the content will be to your taste. The actual flavors are hidden in the interview. Don’t be lazy, listen to the full episodes if the guest or topic is to your liking.
In Google sheets, I created a very broad watch list of stocks that track live data (15min delayed via Yahoo finance). AI wrote me a script so that I get a daily mail highlighting the stocks that crossed their 200DMA on the day. You can add any other trigger to this. The alternative is to pay up for a Tradingview account.
Analyzing news
This is trickier. I see more & more post on ceo.ca from gents copy-pasting the ‘AI take’ on assay results. Each project is so unique it’s very very very hard to analyze assay results by comparing to others. The included comments from the CEO and/or GEO, while often useful, are biased as f*ck, so those aren’t a good starting point to draw any conclusions.
Personally I start by scanning how many holes are supposed to be in the release, versus the number highlighted. It’s been a trend to not mention dusters, while these are as valuable as the actual hits (to serious people).
As regular X readers know, I try to run my own models. It seems like some companies made it a sport to make it a pain in the ass to extract their drill data (switching columns in between NR’s, mixing points and comma’s, typos, relabeling hole names, providing the table as a “.JPEG” of bad quality,… you name it, they tried it all!). AI slashed this time consuming task by half or more. Thank you. PROMPT: Extract from Table X column A,B,C,D and provide me a .CVS format (that I can then directly plug into GeoScience).
BTW: It was Charlie Greig saying GT Gold was sold to Newmont for such a high price, because they (both) deeply cared for the datasets. Or as Tom Woolrych likes to say: “All juniors are worth, is their data”.
Deeper Dive
Found a stock you are prepared to spend some more time on? Great.
There are many questions to ask. Verify all ‘interesting’ data points or statements.
If you scan a lot of stocks, after a while you know what your questions will be. So why not automate that? Insert ‘saved prompts’. I use the free extension Text Blaze (see below screenshot). I then go to any AI bot (Gemini, ChatGPT, Perplexity, Grok) and write “Evaluate company X according to / compact”. Et voila, it answers the same questions in the same way each time.
Most often though, we are running one-off questions (and follow-up questions).
One value add is running “comp tables” with peers. Tip: I often specify a few peers myself. What is a silver comp table without AbraSilver?
Another prompt one many are running is their own or adjusted economic models at different parameters (ex. Silver at USD60, CAPEX at +20%, etc.). Again, you can automate that in a /Model prompt for example.
One-offs: Why a stock moved in the past
Pull track record from CEO, GEO and Chairman
Comp table of ‘African peers’ that focus on Heap leaching, with a note on Rainy season;
Make me a timeline for the last 2y between promised events and deliverables for Cartier Resources;
This is basically endless. Have fun getting lost!
Conviction picks
For stocks you follow intensely over time, it can be useful to create your own ‘environment’. Google’s Notebook is ideal for this. You upload recent news releases, market & analyst reports, technical studies, relevant academic papers, etc. Within this newly created eco-system, you can now ask specific questions, run personal reports, highlights, challenges, timelines, NPV calculations, etc.
This will expand over time, but curious as to how you are using it to your advantage?
A final tip from Rick Rule himself: You have to constrain the fuck out of your AI prompts!
Happy hunting!
https://buymeacoffee.com/petepanda
Panda Pete
- Using AI in a mining speculator’s world
- Junior Mining spoilers series
- Niobium – A retail starters pack
- Lithium supply – A retail starters pack
- Panda’s main Lessons Learned for 2022
Discover more from Pete's corner
Subscribe to get the latest posts sent to your email.
