jasong

Pulling the GPZ, 15" CC and 8". After next week I'm supposed to be able to use my shoulder again, probably more profitable to keep and use it, than sell it at this point. If anyone wants to buy the 12" spiral, or 17" spiral send me a message. Will sell any of these individually or together. (10" is sold).

@Lunk Bingo, that movie shaped my life as a kid. 🙂

Mining claims are now interactive - clicking will now give you a list of all claims under that point, basic claim information, and MLRS/SRP links to full claim data. Many claims at once will create a tabbed popup so you can see all at once - much easier than the way MLRS does it. Issue: MLRS mapper is now at least 6 months out of sync with the MLRS database. I contacted the BLM about this last week but they were gone by 2pm, and then took a 4 day weekend after that. I just received a response and they say they have a fix for the sync issue and it will be deployed "when they get the resources". I asked for more definition and will be working to get them to resolve this. Meanwhile: the mapper isn't totally accurate until the BLM syncs the MLRS database to the HUB map stream - be aware that recently closed claims may still display, and some new open claims may not display. However, if you click on the MLRS link for a specific claim, those pages are all up to date and will accurately show a newly closed clam as closed. So make sure to look at a claim's MLRS page for now, if that claim is of interest. Until then: Claims load very fast - I have two mapping layers, a solid layer will show claims across nearly state-wide areas all at once for a quick overview. Zoom in and you'll get the hash marked claim symbols showing placer/lode/tunnel/millsite. I have the seperate to reduce clutter and to adjust opacity on the solid layer. Click any claim and get claim data and the MLRS/SRP links. Note: The MLRS allows uploading exact GPS coords of claim corners, so the MLRS mapping reflects these exact claim locations, where available, instead of just to the closest 1/4 1/4 section (see above). I always seem to break other things without realizing it when I add new features. If anyone finds a problem, let me know. My phone testing so far is minimal so there may be issues there. If I succeed in getting the BLM to sync the MLRS database to the MLRS mapping layers I will let people know here. I intend on getting them to do it, it's a public service and we paid for that big MLRS replacement of LR2000 with tax money so I consider it a moral imperative to actually be working.

My updates didn't play nice with each other with the popups, even though the sites displayed properly (conflicting interactive layers - this caused the popups to display no information even when a mine/sample site was there). Also the menus were getting unwieldy for smaller screens. So for now I made a restriction that only one interactive layer (the clickable ones that generate a popup) can be active at once while I figure out a better way to integrate everything smoothly later. Also redid the menu box system so you can resize each one individually as desired, and added scroll bars. Now I can add quite a lot more layers in without the menus getting larger. ---> I think everything plays nicely now. Apologies for not fully testing at first. I have been messing with creating a full featured mining claims layer instead of just the visual claims map in there now. Meaning - click and the claim name, serial, location data, etc popup. Making some progress towards it, feels doable.

This one took some work to get right, but it's done now. USGS's geochem (like assays) database for rock samples, soil, sediment, and concentrates. This is a truly massive database agglomerating geochem and assay sample locations into one place. Their interface is clunky, mine is easy. Did you see that thread with Aussies talking about a program they pay for that has a bunch of rock, soil and gravel assay/lab data? This is like that, except free and much larger. It's US-based only though. Contains: wilderness study samples, NURE survey, RASS, PLUTO, ATHENA, and a lot of specialized state-level data like Nevada and Alaska state sample surveys. This stuff is very useful for prospecting and exploration, especially looking for odd tracer elements and signatures. Professional exploration companies use this same dataset to augment their own. Too complicated for me to cover techniques here, maybe more in the future. There are 4 types of samples: Rock - usually chip samples from outcrop/lode Concentrates - usually panned samples Sediment - usually drainage samples, from washes to full rivers Soil - usually grab samples of dirt/soil You can use these to find new mineral bearing areas by understanding the transport mechanisms. Rock - extremely local, taken from outcrop. Anomolies come from right here Soil - local, but derived from all immediete outcrop, but can be blown in, etc. Sediment - contains material from the wider drainage area, or from paleo channels draining unknown areas Concentrates - makes it easier to see valuable anomalies that won't show up in a small sample, by concentrating it down. Just like if you have 1 flake of gold in a pan, your chances of seeing it are slim to none, but pan it down and you can see it in the concentrates. Same with other elements. Just be aware that seeing gold in the sample doesn't mean nuggets. Nugget gold is rare, it just means the element was detected. Also, this is largely spectrographic analysis like ICP-MS, AA, etc sent to labs. That means they analyze a tiny thimble sized sample, so results need to be taken in context because it's easy to over or underestimate actual mineral content when looking at such small samples. It's best to view wider pictures and patterns. Caveat: the USGS servers often run slow. If you click and it sticks on "loading" for more than a few seconds you might try another sample, then come back to that one. This is worse on weekends because they throttle their server bandwidth during weekends. Find it here: www.imageryatlas.com

Haha yeah...in a way AI is no better than we are, and can be worse when it comes to really specialized fields like metal detecting for gold, or how a specific metal detector works. AI isn't superhuman (yet). But the real leverage comes when you realize that having limitless access to the equivalent of an infinite amount of really smart people in their specific fields who will do whatever you want, is pretty powerful too. The problem is when I work with AI, the bottleneck to productivity is...me. I can only manage so many at once. And here shortly we are going to have interfaces that are capable of managing at superhuman time scales these fleets of fairly capable workers. Things are going to change, fast. 1-2 years IMO. Be ready. 🙂

@GhostMiner Did your partner trade the trommel for gold to a guy in Wyoming too? Curious because I just saw this thread and coincidentally acquired a large trommel in trade for gold shortly before you posted this haha. I have no clue what precluded this thread, just saw it and went "wait, what, did I end up with this guys trommel or something?".

You will get the same suggestions a smart human economist might give you. It's biases will reflect the biases of the humans in the training data and the biases of the trainers. They are great unintentional mirrors like that. AI is a lot of things. Creative is not one of them. Not yet anyways. You are unlikely to get anything truly new or groundbreaking out of any current consumer AI models. There is some proof that the ultra specialized (AlphaEvolve, AlphaFold, AlphaZero, etc) models such as what Google is developing are capable of generating new ideas, new science. But they are limited to extremely narrow subjects. Chess. Protein folding. Specific branches of math. Etc.

Long overdue fix: Historic Mines and Prospects are now interactive/informative and easier to see. To use: just enable the layer by clicking on the box, and adjust opacity slider if desired. Click on any mine/prospecit (orange dots) to get a popup with more information. Overlapping points will be tabbed. Some points contain some useful notes, and a full MRDS link is provided at the bottom of the popup if you scroll down. Watch for more improvements here coming in the future. These mines and prospect sites are worldwide. Though not nearly as extensive in all countries as in the US. The speed these load depend on the USGS server load, out of my control unfortunately, sometimes faster or slower. Find it here: www.imageryatlas.com

I just completed an early iteration of a useful prospecting AI app, at least I think it's pretty useful. The problem is that it's way too expensive to run. The software is NOT cheap, even using fairly cheap bulk API calls with the recent price reductions. Basically, each call costs around 50 cents to $3 in AI, bandwidth, compute, and data storage, just to give a rough idea to people what costs are. I can't offer it free or even demo it, because if 1000 people used it I'd owe like $1500 overnight...it's a problem. I will show people a pared down version of the app I built is in a week or so I think. But it may be an idea of what sorts of things you can do with AI related to prospecting, or even general research for relics, treasure hunting, etc. Beyond just using it like a search engine on steroids and asking it questions. A point I want to make after much consideration: @Steve Herschbach addressed the problem with decreasing ad revenue as AI feeds user site content directly rather than bringing traffic to the site. I am fairly certain we are going to see a lot more paywalls go up on the internet in general as a result. I'm not certain ad based income streams are going to be viable when everyone has a personal AI assistant that feeds them recommendations directly, and probably knows what they want before they know they want it themselves...how can ads compete with that? Who will pay for old ads in 2 years? Patreon, subscriptions, donations, etc are probably going to be more necessary going into the future if ads continue to get used less.

I added nationwide USGS geologic maps to compliment the Macrostrat geology maps. These are not interactive like Macrostrat, just images. But often more detailed in some areas, especially structural (faults, folds, shears, etc). In some areas these show mines and prospects not depicted on topos. USGS maps are US only. Macrostrat has global coverage, including Greenland and Antarctica. To use: Simply click the box next to USGS (see green arrow in attached image), adjust opacity as desired. These load slower than Macrostrat, wait 5 or 10 seconds. I have some downtime due to surgery, so I finally have some time to work on updates. Next up is a nation-wide soil, rock, and panning concentrate assay map layer with interactivity. I have the mapping done, the interactivity is proving difficult due to the sheer amount of data and so I'm working on alternatives. This chemical/assay data is very useful for exploration. After I get this working, the next step will be to make it searchable, so you can map gold trace anomalies, or enter elements/minerals of interest and have it map high/low levels occurences of them. Coming soon... www.imageryatlas.com

If you are running cloud based AI in a browser then not really. Cloud AI is stuff like ChatGPT, Gemini, Claude, DeepSeek, etc. All the compute is a data center somewhere remotely. Your local hardware doesn't need to do much more than run a browser (or a normal app on your phone). If you are running local AI then absolutely the hardware makes a difference. CPU and RAM are not particularly important in terms of local AI, it's all about GPU and VRAM. I don't know a lot about Macs but I think they (or some of them) can share RAM with the GPU, which allows them to run larger models than on an equivalent PC which are largely limited by VRAM, which is capped at 24-32GB for the top of the line Nvidia cards. Most full AI models are 600+GB, so require many Macs or many video cards chained together in order to load into memory, or you need to run a smaller, quantitized model. You can run some small distilled and quantitized 1GB or 3GB models on stuff like a Raspberry Pi 4, so I wouldn't be surprised to see these sorts of models on phones and consumer electronics in the future. Including detectors, if there was a need, which there isn't despite what people want to believe right now. Not for LLM's anyways, which is what most people think of as "AI". But I bet we see some usable augmented reality glasses within another year that will be useful for detecting, like tracking your path in realtime for detector gridding, overlaying map layers as you explore, etc. I fully expect to watch my phone phase out in favor of something newer/better like AI enabled glasses in a year or two anyways.

@mn90403 Yes, you can set up local AI's. Until recently it wasn't really feasible because anything powerful enough to be more useful than an intern required equipment that cost far more than an intern costs. Now though, I can run a pretty competent local AI off a consumer Nvidia RTX 4090 and a server about equivalent to a good gaming rig. It's not quite there yet, but close. Running local, all your data is your own. And training your own AI, all your weights and biases are your own unless you choose to share them. No cloud involved.

It's not BS. They didn't say that data was public. There is a massive accumulation of private data within companies. The companies themselves can benefit from AI-assisted analysis of their data. Anyways even ignoring that, there actually is a good amount of public data too in literally every mineral belt I've ever investigated. From older company files that got donated to local geologic surveys, to the national NURE geochem dataset, to state well log data, to state and federal remote sensing data, satellite data like IR, helicopter based magnetic/radiometrics, assays, spectrometry, and on and on... Not to mention the raw amount of public geologic mapping. And even stuff like historical databases and reports, or old hand written field journals. Also, some companies acquire other company's data. Or acquire other companies entirely, including all their data. Hiring a team of geoscientists to interpret it and reanalyze it according to the first company's geologic and economic models is incredibly time consuming and expensive. AI can help this. There is so much that it's simply not possible for humans to find all the patterns in it. This can be assisted with AI, this is one thing AI is actually great at doing - finding patterns that humans understand and taught it, but within datasets that are simply to large for humans to make sense of in various aspects or see the bigger picture in, or lack the tools to merge them together in a way that patterns emerge.

$8500 This is one of the most versatile GPZ's available in the US. I am recovering from shoulder surgery. Selling this to pay off med bills. This setup found me over 3lbs of nuggets, so this thing works awesome. It's older than 2 years, so out of warranty I think. I detected a lot, so it has a lot of various scratches and dings. One WM12 is almost new, the other has a speaker that went tinny but still works as a backup. 2 extra lower shafts (one is missing one of the bushings) and an extra upper shaft. I have the box, harness, j struts etc unused (new) still, just can't reach the box yet to bring them down. Coils: 8", 10", 12" spiral, 15" CC, 17" spiral. Probably $6k in coils alone if you buy them new. Comes with one patch lead to run the X Coils, and I (think?) I have the stock 14" coil still from which I created the patch lead (you will have no need to use it ever again with these X Coils though). I have a bunch of extra coil covers too, also in a box I can't reach yet with only 1 good arm. I will add pics of this stuff later. I don't want to separate the coils unless I have to, people have asked already. In about 1 week after my shoulder heals a bit more where I am comfortable to drive I can meet anywhere in Wyoming, in some parts of Colorado, Montana, SLC, or maybe Winnemucca if you want to do the deal in person. This package would be expensive to ship.