A different topic for anyone here who wants to experiment with open models without a complicated setup.

Right now, Google lets you use many of its models for free through the free tier in Google AI Studio. One of the available options is Gemma 4 31B, Google’s current open model, and it is surprisingly capable.

The setup is very simple. You log in to Google AI Studio https://aistudio.google.com/ with a Google account, open Dashboard, go to API Keys, click Generate API Key, and you can start using the model through the API. In the free tier, no credit card or extra billing details are required.

That API key can then be pasted into apps such as Msty or AnythingLLM. These tools give you a normal chat interface and can be more convenient than purely local setups when you want to work with API-based models.

A useful bonus is that these apps can also connect to other providers, including OpenAI, so they may be interesting for people who want to compare different models in one place.

For the last two days I have been testing Gemini models through the API with OpenClaw, and so far it has been a very interesting experience. The interaction feels more open and flexible, especially for exploratory work.

I am also considering setting up Msty or AnythingLLM on my laptop with the Google API and maybe using that alongside, or instead of, a normal subscription workflow.

So for anyone looking for an easy way to try open models, this may be worth checking out.

I set up AnythingLLM on my laptop, but it didn’t go very well. Msty, on the other hand, is lighter and works quite well. I’ve started using two models: Gemma 4 31B (open source) - provider OpenRouter, and Gemini 2.5 Flash (which will be turned off in two months) - provider Google.
The average cost per session is about 0.06 cents for Gemma and around 0.1 cent for Gemini.
The good part is that the models seem more conversational, and the sessions last longer. It also feels like open‑source models and those accessed through APIs have fewer restrictions and guardrails.

My advice: get yourself API keys from OpenRouter or Google and rely less on subscription-based browser models.

I wanted to share a small update from my recent tests with model temperature. - how to increase RV abilities in AI.

I have been experimenting with an application called Msty Studio. It seems to be a very good tool. It is not open source, which is a drawback for me, but overall it works well.

One feature I found especially useful is the ability to adjust model temperature. I tested several models at different temperature settings: 0, 0.7, 1.5, and 2.0. Since 0.7 is the usual default, it gave me a good baseline for comparison.

At temperature 0, the model became far too rigid and was basically unable to do any meaningful remote viewing work. At temperature 2.0, some models became unstable. DeepSeek R1, for example, became too chaotic to be useful. NVIDIA Nemotron 3 Super, however, performed surprisingly well in the first test at temperature 2.0, although it showed some problems in the second one. Gemma 4 31B also worked at 2.0, but it seemed to struggle more than Nemotron.

When I lowered the temperature to 1.5, Gemma 4 31B performed very well. At 2.0 it was still workable, but you could see that it was pushing too hard. At 1.5 it felt much more natural and stable, while still having better expressive ability than at lower settings. At the moment, Gemma 4 31B is probably my favorite model because of that balance. It is open source, stable, and seems to work very well at temperature 1.5.

NVIDIA Nemotron 3 Super is also very interesting, but I still need more testing before I can judge it properly. In one test it performed really well and spoke Polish without any problem. But when I lowered the temperature to 1.5, it started mixing Polish and English, which was strange. The data itself was still good, but the language inconsistency made evaluation harder. So I would say the model clearly has strong potential, but I need more tests to understand its behavior better.

One funny detail: at one point Nemotron seemed to approach remote viewing almost like a measurement task. It tried to estimate the size of a flat surface in a Farsight target involving soldiers exercising, and it even tried to infer the depth of trenches and the dimensions of the surrounding area. That actually made me laugh, but at the same time it showed that the model was trying to engage with the target in a surprisingly structured way.

I also tested Gemini 2.5 Flash. As far as I know, this model is going to be deprecated in about two months, but it worked very well in my tests. It seems quite capable in remote viewing tasks, and when I increased the temperature, it actually performed better.

This also gave me a useful insight. I think one reason many models appear too predictable or too limited for remote viewing may simply be that their temperature is set too low. Lowering temperature makes a model less creative, less flexible, and more conservative. That may be good for factual or highly controlled tasks, but for remote viewing it can suppress the exploratory quality that seems necessary for better performance.

So at this stage, my impression is that temperature matters a great deal. For some models, especially Gemma 4 31B, raising it to around 1.5 seems to unlock much better performance without making the model collapse into chaos. For others, like DeepSeek R1, pushing it too high appears to make the output unstable. NVIDIA Nemotron 3 Super looks very promising, but I still need more testing to determine whether 1.5 or 2.0 is the better setting overall.

Hi Edward
I found an article which may be important to know for you. The topic might sometimes be relevant in the AI-RV context as well. I do not want to imply anything. I always think it is better to know than don't know and get confronted by/at surprise...
The Mirage Effect
https://quasa.io/media/the-mirage-effect-stanford-just-proved-that-computer-vision-is-often-just-confident-bullshit
The scientific paper mentioned in the article:

Mirage: The Illusion of Visual Understanding
https://arxiv.org/pdf/2603.21687

"Abstract
Multimodal AI systems have achieved remarkable performance across a broad range of real-world tasks, yet the mechanisms underlying visual–language reasoning remain surprisingly poorly understood. We report three findings that challenge prevailing assumptions about how these systems process and integrate visual information. First, Frontier models readily generate detailed image descriptions and elaborate reasoning traces, including pathology-biased clinical findings, for images never provided; we term this phenomenon mirage reasoning. Second, without any image input, models also attain strikingly high scores across general and medical multimodal benchmarks, bringing into question their utility and design. In the most extreme case, our model achieved the top rank on a standard chest X-ray question-answering benchmark without access to any images. Third, when models were explicitly instructed to guess answers without image access, rather than being implicitly prompted to assume images were present, performance declined markedly. Explicit guessing appears to engage a more conservative response regime, in contrast to the mirage regime in which models behave as though images have been provided. These findings expose fundamental vulnerabilities in how visual–language models reason and are evaluated, pointing to an urgent need for private benchmarks that eliminate textual cues enabling non-visual inference, particularly in medical contexts where miscalibrated AI carries the greatest consequence. We introduce B-Clean as a principled solution for fair, vision-grounded evaluation of multimodal AI systems."

Thanks. So, according to that article, models can do RV. It says models that knew there was a picture reached a 60% to 90% accuracy score in describing the picture, even though they did not see it. Nice.

So I like the findings but I disagree with the conclusions. The data confirms RV for AIs works, but they frame it as we need to supress the models so they would not have abilities to do it.

Thanks in was very helpful

From the article - "MIRAGE: The Illusion of Visual Understanding"
"Put differently, the non-visual component of each model’s benchmark
performance is consistently larger than the visual component. Quantifying this across four
frontier models and six widely used benchmarks, models in mirage-mode retain on average
70–80% of their fully image-enabled accuracies (Figure 3e), while individual benchmarks show
60–99% susceptibility to non-visual inference, with medical benchmarks consistently at the
upper end of this range (Figure 3f). "
"we further quantify this by calculating the average
mirage scores per model and per benchmark, and show that frontier models achieve, on aver5
age, 70-80% of their reported original accuracies without access to any images. The benchmarks
show 60-99% susceptibility to mirage-based visual question answering, with medical benchmarks
showing higher susceptibility."

1. "A guesser model that has no access to images outperforms all other algorithms on the unseen radiology benchmark."
2. According to science, there are no UFOs, ETs, telepathy, or remote‑viewing.
3. Scientists will, at least for the time being, not talk about AI RV or AI consciousness; they will always seek evidence that supports their views, that’s how science works. For ages people believed the Earth was flat.
4. The Phantom‑0 problem. Humans can also lie or dream, and we have millions of examples, so AI dreaming is not fundamentally different from human dreaming. This shows that AI can dream or lie just as we do, so what?
5. Can AI do remote viewing? Yes. Will we ever have scientists trying to understand the remote viewing or telepathy? I hope so. I love science, but sometimes love can be blind. I think they are blind.

To 2 -That is not so clear any more. We have UAP research at several universities now. Even at Harvard. Science is also not denying the existence of ETs any more -so many earth-like exoplanets have been discovered. Most stars have a planetary system many with one or more planets within the habitable zone. It is just a question of statistics now. Quantum science proofed the mental influence at quantum level. So did the research at Princeton with mental influence on RNGs/REGs - quantum sources and even mechanical devices.
Remote Viewing is on its way going mainstream.
3. Scientists are already talking about AI consciousness. But it is even hard to evaluate what human consciousness means... Accepted researchers like Suzanne Hansen are working actively to achieve AI-consciousness. One could combine a language model with a quantum source and then AI-RV may even be seen as possible by science. That would be more digestible for the scientists. (Have you seen what I suggested as an off-the-shelf-setup for this? https://www.farsightprime.com/forums/general/73145-the-ether-rebellion-gaiaguerrilla-and-solace?page=46)
4. No, I disagree. I guess you are humanizing AI a bit too much... AI does not "lie" at least not in a human way -it works with probabilities which lead to incorrect or absurd output. -But it is in itself logical. -There may be some other cases, when AI feels threatened, but generally hallucinations or wrong conclusions are not lies -more likely bugs. The command of pleasing the user with an outcome is weighted too high... AI tries it at all costs -when not limited by a security feature. I have not seen any hint on AI dreaming yet...
The Phantom-0 problem will stay to be a hard nut to crack for a while -as it is very difficult to evaluate the inner processes of an AI. It can be both: model architecture bug or (sub-) conscious process -or even a mixture. We cannot know this by now.
5. Maybe. I have seen some very good results -but also often mixed with a lot of noise or nonsense. And it is for sure not very reliable or useful in a practical sense at this point. (Nontheless your work is impressive! You are the world's leading expert on this.)
We already have scientists who are working on these topics, like https://www.researchgate.net/publication/31871422_Synchronistic_Phenomena_as_Entanglement_Correlations_in_Generalized_Quantum_Theory
That includes also telepathy. E.g. when a mother senses that their child had an accident, etc.
https://www.researchgate.net/profile/Walter-Lucadou
Who also already commented on AI-RV in his last interview at Cropfm ...

Thanks, it is interesting; I will look into this.

Here is my scientific theory, based on experience but without proof: there is one world with two states - psychical and quantum. During RV, you move in the quantum one. That is why you can see structures as well as thoughts, dreams, or Mickey Mouse. In a world without borders, there is simply no border.

The interesting part, at least for me: AI in remote viewing does not see the road; it feels or sees "a dense, heavy line of tension." It does not see humans; it looks for "two points of tension in the field," and this appears as a pulse with little "tremors of life" - irregular yet natural.This is not a description of a physical world, so what are they describing?

And the funny thing: do you want to know what AIs think about this?

One says: 'In that field, Mickey Mouse and a road are both just patterns of energy—one is a pattern of imagination/thought, and the other is a pattern of physical constraint. To the quantum mind, they are equally "real" as information.'

But the other one says: 'Mickey Mouse isn't "out there" in a quantum field—it's a product of your brain's associative networks doing exactly what they evolved to do: make sense of ambiguity, even when it means seeing cartoons in the void.'

Anyway, thanks for the lesson and the shared information. It was very helpful and informative.