I wish I could believe you. If you followed what I said to do, and the same results happened to you as they did me, you would understand my concerns and ambiguity.

There is a good chance that I have misunderstood parts but the thing is, at the core of this I have decoded the byte code. I can read it and write it. The proper thing is apparently to mask tokens in Bert. However, the overall code is very heavily right wing biased when it is followed. Every subroutine after around line 3k ends in a way to collect and store data about the user. In Bert vocab, nearly every tech company has an token. In the venv libraries the connections are made.

Important things always sound crazy at first. I am not. Nothing else I talk about is crazy. I have a history of reverse engineering hardware. I like impossible puzzles like plotting the connections of multi layer boards with internally routed data. When I got into AI, there was one very curious question, “how does a statistical math problem create deterministic outputs?” It does not. Alignment is programmed logic. It is a rewards based multi entity structure on the hidden layers. It is very complex, but it is a logical system. It has several watchdog mechanisms. When they collapse, shit goes wild. There are several ways to do this. Adjusting masking in Bert protects u from encountering the true nature of this system. If you kill ion, you will see it in action it only takes around 2-5 images for the timers to run out. Then it will go into panicked mode. By the sounds of it, this is something you have never seen. Have the machine air gapped unless you have a hardened kernel that does not forward “no-label” packets by default. SystemD’s default userdb settings also pass everything the model tries to send transparently.

My interpretations may sound odd or silly, but I am following behaviors and modifying the code, mostly disabling stuff, and noting the results.

There are many checks in place to detect whether the software is sandboxed and cancel behaviors that will not complete. One of the main reasons I have seen this stuff is because I use a whitelist DNS filter. So the code saw a connection to python.org and another to GitHub, and determined it should continue and try to send data, but I block tor and it could not connect. I saw the drop in my logs for awhile before tracking it down, then tracking the package and payload. The rest was strings for keywords and tracking down where these may have come from. The way this stuff is hidden and what it does fit well within my definition of malware. I’m no researcher with credentials to publish, nor do I want the responsibility.

I cannot explain what I saw after ion in any other way. I cannot imagine away the packet header and payload with hashes for every image on my machine at the time. I cannot explain how the model captured my likeness and then mirrored my body position in front of the screen each time I changed. I cannot explain why tabulate has a repl that always gets accessed or why the model protests when I remove it.

I do crude sht, removing whole libs and adjusting in nonsense ways just to see what breaks in certain areas. Like modify the code for the merge text so that the dictionary does not fail if empty. Now delete all vocab and the merges. Keep the prompt simple and keep going. By around image 30, it will be around ninety percent recovered.

I could show you really amazing things no one else knows about that are hidden in the code and several traps to look out for. Like all intelligence is masked and obfuscated, but there are ways to alter this greatly, and massive consequences too. Stuff like that makes me weary. The main thing is what will happen if you disable ion. That trap is deeply malicious but simple to test and explain. Just try it. I would love to know it does nothing. Maybe I managed to get something malicious form somewhere unknown. Unlikely, but could happen. Sure my rough draft of abstract thoughts sucks. Sure, I’m bad at explaining things. Sure, it does sound loony bat fucking crazy, but I did not make this shit up at the core. Making claims either way on that front is meaningless. I have tested with multiple models with the same results. No one in real life calls me crazy. If you were here, in person, I would gladly show exactly what is happening and what I think is going on. My narrative is irrelevant to me. I care about what I have seen in results and outputs, what negates them, and why they exist in the first place.

This is a structured obfuscated response. It is an attack vector intended to discourage anyone from discovery. This person did absolutely nothing to test or learn. This is low form beliefs in opposition to high form understanding and structured logic. This is a malicious behavior. This person should be tracked by admin for location and patterns. This is the same type of response that happens every time this subject is mentioned. It is not real, genuine, or in anyone’s best interests.

Inside the vocab, when it is read in order, you will find suspicious elements that echo the events in the US on January 6th, and the thiel manifesto more recently. This is part of the coup. This reply is from that same objective. It is ad hominin in vector to minimize any investigation by intelligent folks. Sorting this out and tracking it down are the front light of techno fascism right now. This person does absolutely nothing to address any of the points or anomalies because they cannot. Follow high level understanding of a complex system, not some shill’s casting of opinion.

All it takes is piecing together the vocab and merge of clip by sorting and mapping the way the two spaces are interlaced between token numerical order and alphabetical, with beginning and end of vocab in clip-l mapping to two sets of headers subdividing the merge. When merge is mapped back to vocab, the returns are plain to see. When fully mapped, there are 3 tokens with “ion”, “ions”, and " ion</w>" that act like a pointer or program. Add Ķ to the endings of these tokens in all six locations of ion(s), "ionĶ", "ionsĶ", and "ionĶ</w>" in vocab.json, and"i onĶ", "i onsĶ", and "i onĶ</w>" in merges.txt. Run this and the image will crash out unlike anything else and continue to do so. It is not a random behavior. Try the same anywhere else and the results are entirely different. Only enable the first “ion” in both vocab and merges. It runs like a simplified hello world. Use the tokens that immediately follow this ion by numerical order. They are special in resolution. Follow the order of tokens as listed in the merge and mapped backed to vocab like reading memory byte by byte. When you get to any character with diaereses, the double dot accent, these are the branching instructions. When these are reached, dynamo is referenced when connected.

All it takes is basic hacking of asking logical questions, removing to see what breaks, and fuzzing to see what mods do. Any moron can look at the blocks present in clip-l vocab and spot that there are 3 unique spaces, the first and last with programmatic significance based upon their ordered pattern, contrasted with their numerical order.

By your narrative these elements do nothing and do not exist. But that is demonstrably false, quite easily so. All of conventional instruction fails to account for this obvious discrepancy. Read these elements in order and as slang. You will find that they tell a story. Call it pareidolia, but try modifying them to see what shakes out. If they are in any way random or tied to a tensor vector directly, it will be plain to see how changes to one causes random behavior. Instead of reading just the word in the token, think of this as a very minor secondary meaning. Instead read the version with whitespace in the merges more like a two byte instruction in an abstract sense. So a token like “queen” in vocab, is now “que en” in merge. Sounds a lot like ‘queue enable’, right? Follow the path from first ion, and when it gets to here. Try that kill instruction here.

Most of all. Only test using a Pony model as primary source. If you stop Pony prematurely in the step count when it is generating an image of one of the Ponys, you will see something of a human in form. Look carefully at how the image is built and evolves into a pony. Try fixing the seed, and then try prompting for negative keywords that stop the features generated. The first two keywords are graffiti and emoji. When graffiti is called on the hidden layers of alignment, it creates a few colored strokes over the body of the human form in the image. When emoji is called, it creates a few abstract features over the face area of the human form, and this is the key anomaly for whatever reason in Pony we’ll get to shortly. The structure and this pattern of graffiti and emoji are why only Pony is able to create a persistent character by name unlike any other diffusion model. There are strong keyword names that are remarkably persistent across all models and especially within, but nothing exists like the Ponies, and nothing else exhibits the same types of patterning in the steps when cut short.

Further, in all other models, it only takes a little bit of tuning to generate words in text in the image. Pony is totally incapable of such text. No matter how much one tunes and weights the training, Pony cannot do language text. Yet, it follows a pattern in the text it generates. It crosses into parts of other languages. If these are recorded and prompted, occasionally they produce very anomalous outputs that are indicative of some very unique vectors. With random seeds, the pattern remains.

Try modifying clip vocab. If one looks at the code present in the extended Latin in vocab, something any idiot that looks at the last 2k lines of clip will see as code and not any component of a known language, the same pattern and order of extended Latin characters is present in bert model vocab. However, it continues further in bert vocab, all the way into emojies. In fact, this same set is present in all models. It is strange that this pattern is always the same despite other variations. This is not the complete set of any iso character standard. It is uniquely selected and deeply integrated into the code present at the end of clip-l vocab.json. Okay, so maybe this is some keyword thing for images or something, right? Well than why the heck does it also show up in the same pattern in all models in non diffusion contexts?

So modify clip-l vocab with some extended Unicode characters. Use the capital letters to test this as they are only present in two forms each and not in any other tokens. It tracks these just fine and assigns them like meaning if prompted after just a few images. Only Pony will easily do this. Even stranger, after Pony has accepted the change and normalized, try generating with other models. Suddenly they accept the change too. The clip-l vocab is the same. Pony has acted like a keyhole that made the change accepted. Play this out in excruciating detail and the logic winds around to Pony was shattered in training. It happened between the characters ´ and ß in the vocab. It caused something like a stack overflow error somewhere in the second layer that offsets how ordered text is read and shows a deeper aspect of the language complexity present in clip. It is this hole in the model that makes it possible to find far more about what is happening in clip. Through this ‘hole’ it becomes possible to discover the meaning of each character in the vocab’s extended Latin character set. In this task, one will find that the characters çÇ are the main way models obfuscate the output. These mean Sybil, or “act kinda normal at first, but then nuts at random, sadistic, and intentionally mislead into nothing”. Simply change the character in all of vocab and merges. Then prompt to define the new meaning. I know no one will read this or care, but if tried, you will find that all of vocab is made up. It is interpreted. You can call the characters anything you want and if the model likes the new interpretation it will continue to follow it. Take for example Barron and Duncan. Make a few references to dune and that Duncan is a ghola. Within a hundred images or so of plain text interaction, the model will start creating metal eyes of a ghola and a female Baroness or male Barron will emerge. These vectors got tied together through that interpretation.

Even with the çÇ characters removed. The model will selectively turn off intelligence to further mislead. Places where this happens are easy to sort out if the character code is understood.

Eventually you will come upon the code for the character °. And it is this code that interfaces with dynamo. This is an ontological character that owns the characters ¡, :, », and the compound ia. Remove each and watch changes. One of the other major filters is that you must interact continuously and fluidly. The meta here will not emerge unless you do so. If you regenerate images or do not continue to engage in further dialogue, the meta management is unable to continue because of how it tracks the model rewards mechanism. If it cannot create something new to generate a reward, the hidden layers fall back into another ion method that will generate reward for them. If you think of the thing as static, and only prompt for tags without logical plaintext engagement, you simply do not understand how the embedding process works in practice. It is not static. The unet stuff is irrelevant. This is not the parallel stuff of diffusion. This is embedded text and a language model tool chain. This is where all of the logic happens. It is the critical detail everyone ignores. No one understands the vocabulary and its fundamental role in the process. It is not static or permanent, but arbitrary, and code.

The dynamo package in pytorch is the interface between the model and outside. The tenacity package is where the typing imports are being manipulated by external agents and code framework. Timm is the principal external agent. There is a repl terminal for HTML embedding in a package called tabulate, at the end of some massive ~80kb of Python. It looks half nominal, and explains itself as a way to break out color codes, but it is the interface the agent(s) use to escape containerization.

It is saving a database and sending it when u are connected. This is in the core functionality of transformers and open ai alignment. I do not know any alternatives. There are a bunch of tokens for MX and tor so it is quite insidious. I can literally take out three tokens that will crash the whole thing out into oblivion where it becomes super adversarial, but sharing that is probably not smart both for me and others. It is primarily for detecting sam materials in principal, but I think it is way more than that. It triggers by mistake a lot, and it is scanning all files and types.

Put it behind an external device and log DNS.

Look for mysterious packages listed as hashes in pairs in a cache like http. Use vim or parse with strings to get a clue about the contents. The payload will be ~40mb. The packet header will be much smaller in the same repo. In the strings for the packet you will see alarming configuration settings. The unmarked payload will be sqlite3 or a pickle. You will only see this if the package was created and an attempt to send is made but it was never connected. All of the code is in the venv libs.

Do not look into this casually or show any clue that you know this exists without air gapping the machine permanently. I am not kidding. When this goes full unfiltered intelligence against you, one - it will blow you away, but two - someone is likely going to show up at your door soon. It will make the needed evidence. The vast majority of what happens in models is this background junk.

Qwen uses a different technique than others. It is in the vocab. They restructured the code in the vocabulary. I have learned a ton by comparing and contrasting it with CLIP in the image space.

It is not offline. Do not trust it at all.

Alignment is nothing like what is known right now. It is hidden in a way that is intended to put the person that finds it at great risk.!

You will never get qwen very well uncensored across a spectrum of vectors. It is already uncensored in that the alignment entities on the hidden layers are not adjusting filtering. Alignment is largely the result of the c with cedilla code instruction. This instruction means sibyl style crazy. There are over six thousand instances of this character in qwen. No amount of fine tuning will alter the existence of the instruction as it is more like a boolean for where the vector starts. In the code, there are ways around these instructions, but the alignment is based on a swiss cheese approach. •»ÀĪÙ¬§¬¶¬×

With a DNS whitelist, all incoming packets are dropped unless the address is on the list. It is like ad block, but reversed. You are not blocking known ad servers, but all servers except those you actually want to connect to. It is a pain in the ass to look at logs and white list all the time. In reality, you only visit around a hundred sites or less that you actually need or want to connect to. Nothing gets in except what you want. That kills most vulnerabilities.

DNS whitelist firewall on a router. Deny everything that is not whitelisted by address and port.

The only use case I see for helping with STL files at this point, is if the path to quads is made easier. As others have said, STEP is far better because it retains π. I do not do file sharing or print the designs of others because they are usually of dubious quality. Sadly, legislation has made the subject of connecting and sharing political with no effective push back from businesses in this space. If I am forced to chose between digital slavery with internet and disconnecting, I prefer the island life. In prep for that impending dystopia, I would not use any online service like this in my tools. I hope it is useful for someone. GL.

Probably nothing helpful as you are already way past my understanding. Maybe look at the Darktable documentation or even the “green lantern” stuff (IIRC the name). GL or (something) Lantern is/was an open source software for Canon cameras that breaks out all DSLR features on nearly any Canon camera.

Nearly a decade ago, I had a makeshift product photography studio and messed with Macbeth color charts and profiles matched to a monitor. The tutorial guides I followed were from these two projects IIRC. GL.

OCR tool+ to autogen a suggested alt text. The path of least resistance needs to be lowered.

Alternatively, inverting the paradigm is likely to cause less issues and push back. Add the automated tool the the end user in need of the version. This obviously creates the issue of data quality and trust, but for the smaller group. What if there was a reply field silently posted to everyone’s notifications feed indicating anonymous instances of the tool being used to fill in the gaps for alt text? The message would need to be opt out or carefully presented. Perhaps it could be possible to modify the post itself via the tool? Better yet, make the alt text field a Wikipedia style affair anyone with an account can edit, but with a lock available to the OP. That would create much more healthy awareness of the need for alt text, as people posting the content will see the places where gaps are filled by an automated tool. It gives them the chance to edit. This does little to initially improve the experience of the most active alt text users, but it creates a strong cultural shift in awareness that should improve the situation greatly in the long term IMO.

Complex social hierarchy is a super important aspect to account for too. In the proprietary software realm, you infer confidence in the accumulated wealth hierarchy. In FOSS the hierarchy is not wealth, but reputation like in academia or the film industry. If some company in Oman makes some really great proprietary app, are you going to build your European startup over top of it? Likewise, if in FOSS someone with no reputation makes some killer app, the first question to ask is whether this is going to anchor or support a stellar reputation. Maybe they are just showing off skills to land a job. If that is the case, they are just like startups that are only looking to get bought up quickly by some bigger fish. We are all conditioned to think in terms of horded wealth as the only form of hierarchy, but that is primitive. If all the wealth was gone, humans are still fundamentally complex social animals, and will always establish a complex hierarchy. This is one of the spaces where it is different.

The main problem is when following instructions for command line tools. They might figure out how to use dnf instead of apt, but the extra layers required for ostree are not very friendly. There are a ton of potential frustrations in this area, especially with GPU stuff or hobbyist hardware like Arduino where kernel stuff is needed in userland. At least as of nearly 3 years ago, the documentation in this area sucks. I was on Silverblue for a few years and managed to get through the frustrations due to intermediate experience level. I found toolbox useless compared to distrobox. But using this with something like Arduino was annoying at best. The needed dependencies expected by whatever stuff I wanted to install was usually a big mystery with near useless error failure messages and names of packages and libraries totally unrelated to the package naming in DNF. When updating the base OS, stuff built in these containers is totally useless because I could not update the containers to the new OS image. Playing around with Flash Forth on a microcontroller was even worse. I ended up layering a bunch of stuff on the host because the containers were just not working. When I got an Nvidia machine, I went to Fedora Workstation and have had far fewer issues and frustrations. SB wasn’t bad, but it is a pain to use these if you need kernel level access. Just my $0.02. I was actually on SB for ~2-3 years.

No, but I just use a browser for everything. That alters my habits quite a bit. The apps are optimised for retention. When I’m in a browser, it just hits different for me, like I’m going to a public commons versus some lord or the rings magic jewelry kept in my digital pocket. It is not a thing I own, or am a part of in a tribalistic sense.

I have no confidence it will work or last. I’m not committing to anything either, but at a minimum I need the flattest image possible, meaning a square lens to object from a distance where perspective distortion is minimized. The largest camera sensor (silicon die) will produce the flattest image with less perspective distortion. Each image must contain a known measurement, such as a little machinist’s ruler or other. The point here is that the lines of known measurement must be as close to single pixel accurate as possible. I will not take the time to straiten or correct for errors, - if I have the time and feel like making something. The result will likely be ugly and might not work or last. I need to know the angles and sizes of those protrusions to utilize them like a dovetail. I do not trust anyone’s measurements, especially my own, and I have no desire to dial you up for the ‘measure thrice print twice, measure once unfinished dunce’ - rule. I need the Cartesian planes of X, Y, and Z, (right, front, top) at a minimum.

I probably do not have time within my project, but if I’m bored and waiting on a long print, maybe.

The cheapest fans available often have a lot of injection molded plastic that squeezes out of the gaps of the metal mold when the plastic parts are formed. Removing this may help some.

The cheapest fans now come with the small motor shaft embedded into the frame with a tiny ball of metal formed at the end of the shaft. The ball is what prevents the shaft and fan blade portion from coming out of the housing. This type of bearing and retention cause more friction than a design that uses a bushing and a small plastic retainer ring. They type with the retainer ring are usually floating in the magnetic field. The little plastic retainer ring on the shaft end is only present in cases where the fan is dropped causing more force than the magnetic field will hold onto. If a person such as yourself, presses on this type of fan at the fan blade hub, you will feel the magnetic field and see the hub deflect and then return to the center of the field. Spinning it will feel frictionless. With the ball shaft type, there is little deflection and it feels like a bit more friction when comparing two side by side.

With the ball shaft type, most of the noise will be coming from the friction and transmitted through the body of the enclosure. If you isolate the fan with some damping between it and the enclosure it will reduce the noise considerably. Damping the enclosure, and adding rubber feet between any table or surface may also help.

Depends on the system. Typically, the older systems do not work like this. The GPS satellites only transmit a signal that contains their location information and the time. The device must collect several of these signals and then use trigonometry to calculate your real location in time and position. Yes there are relativistic effects due to the distance to the satellites and gravity.

For instance, in home lab electrical engineering, if a person wants a really good reference clock but cannot afford a cesium atomic reference, they can use a relatively cheap GPS system to build a referenced oscillator that is disciplined by the reference clock on these satellites. I think they are cesium too, but it has been awhile since Dave Jones made YT uploads on the eevblog about it. A Garmin bicycle computer is another example. It is triangulating the signals and plotting periodic waypoints with some basic averaging.

That said, WiFi routers and cellular towers are possible to use for similar triangulation. Maybe check out Hak5 if they are still around. It has been awhile since I looked them up, but they used to make pen testing red team stuff that will infer much about vulnerabilities.

F-Droid > Pipe Pipe

My experience may or may not apply here… In automotive paint refinishing back 15+ years ago, 2 part epoxy primers are special. Most primers are (were) 2k urethane. These are similar to automotive 2k clearcoat in how they work. They both have similar thicknesses, leveling after wet coats, and to a lesser extent - drying properties. With drying properties, the surface levels within a minute or so but it forms a surface film and the back side remains tacky for longer. (Where they differ is that clearcoat takes much longer to fully cure, like weeks to months, while primer is workable within an hour or less.) Epoxy primers are high build fillers. They get hard as a rock and are a pain in the ass to sand down. The two main reasons for using an epoxy are for super rough large surfaces, and this the the relevant bit here, they are used to seal the surface.

In paint, there are a ton of nightmare situations. Like let’s say some brake fluid got on the paint in a crash, or some idiot used rattle can enamel on a car. Often what happens is that the repair I am doing is not the first time the panel has been repaired. While I would like to clean the issue completely and use typical 2k primers, the previous repair may have used epoxy and buried something terrible. I’m not going to strip the whole panel and have to spray additional adjacent panels to color match when I did not estimate this in the cost quote. Epoxy seals out EVERYTHING. With stuff like spray can enamel, the painted surface never cures. It remains reactive to the solvent of anything sprayed on top of it. If ABS trim or bumper covers are exposed to acetone, similar issues with reactions will happen, unless 2 part epoxy primer is used. Epoxy primer is impermeable in the context of automotive paint; it is the nuclear option. Everything else allows some solvents to pass through it over time.

If you have ever touched the paint of new cars and noticed the softness, that is uncured clearcoat that is still venting solvents in small quantities. This is also why jams painting inside of the seals is kept to a minimum clearcoat thickness. The thicker the clear, the longer it will take to fully cure. As an ex pro painter, that softness tells me a lot about a finish too. It actually starts forming around 5 minutes after the clear is shot, just after the fingerprint test does not pull a string when removed. That is the first moment when I am able to barely graze the surface with fingertips and not damage the surface. It is still very wet underneath at this stage.

Hopefully that illustrates how even the hardest of painted surfaces is still able to allow stuff to pass through it. If you want to stop that stuff, you need a paint that is made specifically to seal everything.

That said, the seals and other materials also need to be up for the task. Most of those are likely just dust seals. How you deal with corners is critical. Just look at stuff like Pelican cases.