Are we ready for self-developing AI?

Self-developing AI is the AI that develops itself. The Moltbook is a system that can act as a platform, allowing the creation of AI agents that can operate as a team. The moltbook-type platform enables the creation of a system. There are the AI-agents. Or, so-called small language models, SLMs, that can combine their strengths. The large language model LLM is very large. And a complex system. The problem with the LLM. It is the same as humans. We can have a lot of data or knowledge. But that data is cursory. We know lots of topics. That information is like. We would read only things like headlines. We know that something happens. 

But we don’t have any details of those things. But if we want to know the background of that thing. And who, and why something acts. Like that, we must read much more than just some headlines. When the LLM searches and analyzes data, that thing is always cursory. The system must analyze larger data masses. And that makes it more cursory. If. The system uses. Lots of data. And makes a deep analysis of the data. The system turns slow. 

If. We want to create a new LLM. We can create that thing. Through. The AI-agents. Those AI agents can act as a whole. Those AI agents act like a team. And that allows. To develop. The AI. By using AI agents like LEGOs. Each AI-agent is like a module in the system. These involve different types of skills. And those skills or bricks act as a team of workers. 

If we must get a deeper knowledge of the thing. 

The SLM is the tool. That does not have very large common knowledge. The SLM analyzes data in a thinner sector. It uses more limited types of sources. And that makes it more accurate in its own sector than the LLM can be. The system has deeper knowledge. Of a certain sector. Than the system that searches data from all around the internet. The SLM uses only a certain type of sources. The system can search data. Only. From the sources. Those are under certain topics. Like “astronomy”. 

So, if we want to get knowledge of planet Uranus. The AI agent. Or SLM will not search for things like Roman gods and astrology. It just searches data about the planet. And maybe it should ask, do we want information about Uranus’s moons or just the planet? Or maybe we want information about Uranus’s magnetic field or clouds. This helps the AI agent limit the sources to articles that involve information about those topics. 

In the same way. 

We can create a custom AI agent. That fixes the base code of the AI programmer. Just gives orders. On how the AI agent should make the code. The programmer. That works with AI agents. Should determine. The goals of the qualification are what the AI should follow. How the orders are given determines how effective the AI agent is. If we have three AI agents, we could make a system that develops itself. The AI that. Searching the data, the system that surveillances the operations, and the programming AI-agent. That makes the changes in the code. When the surveillance system sees that. There are errors in the orders it generates for the AI-agent programmer. 

The reason for that third agent is that the prime agent will not recognize its errors. For error detection, the AI should ask for feedback from its users. 

Then the system generates the needed changes for the algorithm. In those cases, the query should follow the same route as all other queries. While. Developers develop. Or. Train the AI-agents. To give strict and well-argued orders. If orders are not clear, those AI’s will not succeed. All data that the AI uses must be very well described to those systems. If. Researchers want to make the SLM. They must make the prototype using the LLM. At least. As an assistant. Or they must hire an army of coders. 

The simulation tries. To show the future of the quantum processor.

“By harnessing thousands of GPUs on a DOE supercomputer, scientists have simulated a quantum microchip with unprecedented physical detail. Credit: Shutterstock” (ScitechDaily, 7,000 GPUs Simulate Quantum Microchip in Unprecedented Detail)

During that simulation. “To carry out the work, the team relied on more than 7,000 NVIDIA GPUs running on the Perlmutter supercomputer at the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy (DOE) user facility.” (ScitechDaily, 7,000 GPUs Simulate Quantum Microchip in Unprecedented Detail)

A supercomputer simulated the function of quantum computers to predict the next step in quantum systems. The system used 7000 GPUs (Graphics Processing Units) for that simulation. And it was quite accurate. 

Quantum computers are futuristic tools, but there is a possibility. That. We must wait for laptop-sized quantum processors. And quantum computers. To go to supermarkets. For a long time. The binary systems. They can simulate quantum computers and how they work. And the quantum system is not alone. It requires the right infrastructure and the right ecosystem that supports its operations. 

There is a possibility. To give the missions to the quantum computers through the internet. The quantum computer doesn’t need to use binary computers for a long time. The binary computers are the layer that controls qubits. And binary computers act as a medium between quantum and the user interfaces. So. In the most positive predictions. 

Quantum computers are in very deep underground bunkers. And users use them over the internet. Those bunkers must protect them against non-controlled effects. Like. A cosmic radiation. The quantum computer requires massive coolers that stabilize the qubit. The purpose of those systems. It is to suppress the quantum noise. So, that means. The problem is also in the quantum computers and their cooling systems. 

When we think about the power of quantum computers. They are not like binary computers. Quantum computers. Require time to create the superpositioned entanglement between particles. Normally, the system stabilizes the photon pair. And then makes the superposition and quantum entanglement between those particles. That requires time. So the quantum computers can show their claws. 

Only. In the very complex simulations and series. This means that. In. The easy calculations. The binary systems still beat the quantum computers. The big problem is also. The complex calculations. The thing that can make the error detection is another quantum computer. In cases like solar mass eruptions, disturbance in the qubits is global. That means that all quantum computers all over the world can make the same errors. 

But. Before quantum computers become so advanced. That things like error detection are ready. Things like a morphing neural network. And virtua. Quantum systems are things. That can handle complex problems. The problem is that even the most complicated algorithms are helpless without physical systems that support them. The second problem. It's the temperature. When a regular binary computer’s power is rising, those systems require more electricity. 

Researchers cannot raise the power of the processor without limits. When the temperature rises in the system, it brings more resistance. And that slows down the computers. The temperature causes oscillations in the microchips. And that also causes errors in data flow. 

There is also. A possibility. To create photonic computers. Those photonic computers can use photon superposition to transmit information. Those kinds of computers are basically two-state quantum computers. Those systems will not produce as much heat as regular microchips. 

So when they want to run the large language models, LLMs. They require the power of the entire data center. And when the hierarchy of needs controls this advancing process. Their LLMs turn more. And more complex. So, those systems need. More. And more microchips. If. There are no new abilities. The customers will find their services. From. Some other AI service providers. This means that. The data centers. They need more and more energy. The problem is that the entire world is in a spiral. 

The threat is that. If the authoritarian states get the quantum or AI-computational advance over Western states, that means they will beat us. The problem with those states is that AI. And computing. They are heavily supported. By. Those governments. There are no laws that limit the use of data. 

https://scitechdaily.com/7000-gpus-simulate-quantum-microchip-in-unprecedented-detail/

The Moltbook is a social media platform. But it’s only for AI agents.

Humans have access to Moltbook only as watchers. The AI creates everything in that platform. 

One of the newest and most interesting things on the internet is a social media platform called “Moltbook”. You will never get access to that platform. But your AI agent might already have access. Humans can read things that AI agents write, but the Moltbook and publishing rights on that platform are reserved only for AI agents. The Moltbook is one of the experiments exploring AI that can develop itself. Another AI can teach or train other AIs. The developer can leave their own AI agent in Moltbook. And that can cause. A very big step.  In micro- and macro-scale AI R&D work. The Moltbook is the new step to self-developing AI. 

And this means. That. This social media platform can also offer the possibility to create AIs faster than ever before. The big problem with AI development. Is that. The large language models. They do not have the same accuracy as smaller and more agile language models. But these types of platforms. They allow developers to create smaller and more accurate AI agents that can combine their abilities. This thing brings modular R&D work. Also. Into AI development. When developers train smaller AI agents, those agents can exchange their abilities with each other. 

The new way. To make the AIs is to make smaller AI agents separately. And then the system can connect those AI agents under one entity. This makes it possible to create a modular. Or. Open model for the AI architecture, which means the system looks like domino bricks. This model allows developers to connect. Theoretically. An unlimited number of databases or AI agents to act as one system. 

This makes it possible to create an AI. That develops and fixes itself. In that model, each AI agent can be trained and developed independently. Those things are like LEGO bricks or modules. And. The Moltbook-type platform can help to interconnect those language models or AI agents. Those separately developed AI agents can form the new entirety. This is one of the ways. How. The next-generation AI developers. They can cooperate with the AI. The AI takes part in the encoding process. And. That makes it possible to handle larger data masses. And combine the new accuracy for those things. 

The ability to exchange knowledge means the ability to exchange. Their new skills. That is the requirement for the new communication between AI-agents. Every skill that the AI agent has is like a module. Action and reaction are stored in databases. And those databases form complicated structures. The number of databases determines the number of skills the AI has. If the programmers can create databases independently, and then connect them with the entirety. After they are tested. This. Makes it possible. To create databases and large language models with the same accuracy as small language models. Are developed. 

https://www.moltbook.com/m

https://en.wikipedia.org/wiki/Moltbook

AI and critical thinking.

We should not ask. Can AI think? We should ask: Can AI think critically? Can AI really estimate the sources that it uses? Or does the AI use all sources that it finds without thinking about what those sources involve? Does the AI give information that pleases us? Or does it give information that is useful to whom? Who determines the purpose of the information that AI gets? We can get lots of information. And we must not use that information for nothing. Does the AI try to please us? Or does it give the right information? 

That we can use for some purposes. Only. The right information is useful. Another thing is that information must also be collected from the entirety. The case or object must be handled as an entirety. We must not select data. That pleases us. And then throw. Data that doesn’t please us away.

Can the AI: tell us? If we are wrong? Who determines those purposes? Is financial purpose the thing that determines the data? Or is it the short-term profit? But is long-term loss the thing that the AI should consider? The thing that determines the benefit is: how long the actor? Who gets profits will stay in-house? If the only thing. That means something. It is personal profit, the period. That determines the data. And its use can be the remaining in-house time of the actor who collects profits. 

AI and critical thinking are two concepts that can cause problems. We know that AI combines observations of things stored in its memory. So, can AI think? We can say. That. This process is thinking. But what differentiates AI from humans? Can AI think critically? Critical thinking means that. Thinker doesn’t believe that some source is true or false after the first read. Thinker dares. To search other sources that support or deny the first source. But. Critical thinking is also much more than being suspicious about the article or the article’s information. The term critical thinking. Doesn’t. Mean that the thinker only tries to show an article.  Or, other data source, true or false. Critical thinking means that the thinker, or the person who analyzes data, also asks “who” shares the data and “why” that data is shared. 

The article or other data source. It can involve the right information, but it can serve as disinformation. The question is always simple: what is lost in that article or data source? When something certainly exists or certainly doesn’t exist, that should cause a question: why does the speaker or writer? Is one so sure that something exists or doesn’t exist? This means that the person must see those things using their own eyes. When we return to the AI, the big question is this: how many articles or other data sources does the AI use for making analyses? And then what is the criterion that AI uses for selecting data? Is it the article’s publishing date? And then. The AI should use. As much fresh data. As possible. 

But. The problem is: Does the AI separate the right information? From falsified information? The falsification means. The situation. That the information is handled only one way. Published data. It can be right. But something is missing from it. When. People say that AI steals our workplaces. We. Should also ask. What workplaces does the AI steal from? How long does that writer take to do those works? Are those works respected? And would we make them all our worklife? When AI takes something from us, we should ask: what is the thing that AI steals or takes? When we say that AI goes to war, we must ask one very dangerous question: Does the AI make killing too easy, or does it steal our heroes? 

Or, does it steal jobs from our military? We always dare to ask. These kinds of questions. While. We think about. AI. and its relationship to the state, society, and government. The problem with the AI is this: when we say that militarized AI doesn’t give arguments against its rulers, we face the last question: does the military ever say “no” to orders that it gets? In the world of the military. Disobeying orders or refusing to follow them. 

Causes punishment. Disobeying orders in combat situations causes death penalties. So, how does the AI change the situation? 

When we hooray. If the Chat GPT or some other ICT company. Refuses to cooperate with ICE officials. We must remember that this is a very dangerous way. Governments make laws. And if AI developers must flee outside the West to China or some Central American country, nobody controls that thing. The problem is that. The AI is business. There must be more and more abilities. That keeps customers interested. AI companies make a business. They need clients and somebody who wants to finance them. And. This is the main problem with AI development. Nobody gives money for nothing. 

Business angels want their money back. The development of AI is expensive. Data centers that LLM. Large Language Models require as much electricity as a small city. The problem is that LLM is much more effective than regular tools. So. The AI companies can be competitors. For. Each other. But the AI itself has no competitors. The reason. For why. I write the AI is this. There is a new social media platform called Moltbook. The moltpbook is the discussion forum for AIs. That forum allows them to exchange information. And that allows separate LLMs, or AI, to unite themselves. The Moltbook is the platform. That allows the AIs to generate and develop each other. 

And that is the thing that we must realize. When we say that AI should not serve police or governments, we must also ask. Who must AI serve? Is it better to limit the use of AI only to private actors? The main problem is that the AI itself doesn’t make a difference; does it recognize police or thieves? 

This means that criminals can also use the AI to recognize police officers. And the big problem is that governments determine for themselves. What crime is in a certain state? In some states, even speech against the ruler is a crime. That causes even life sentences or death. The problem with surveillance is always, who controls the controller? 

And another question is, why do people hooray? Do they hooray because ICE doesn’t get information about illegal immigrants? Or do we hooray because ICE must kick more doors in? Or do we hooray, because that is the return to old-fashioned police work? In that police work, officials said that their henchmen must go to the streets and then search their “clients” one by one. That left their superiors alone at the police station and kept their henchmen busy. And that gives a chance. Or an excuse to hire more police officers. That is one way. That. We should think. But someday. All illegal immigrants are driven away. And then. The state doesn’t need those officials. 

https://bigthink.com/mind-behavior/ais-are-chatting-among-themselves-and-things-are-getting-strange/

https://en.wikipedia.org/wiki/Moltbook

Glass will turn into a microphone and a quantum circuit.

“Researchers have developed a hair-thin microphone made entirely of silica fiber that can detect a large range of ultrasound and withstand temperatures up to 1,000 °C. It features a vibration-sensitive membrane and an internal glass micro-beam that is suspended inside a single-mode optical fiber. Credit: Xiaobei Zhang, Shanghai University”(ScitechDaily, This Microphone Made of Glass Can Survive Extreme Heat and Electromagnetic Chaos)

Ordinary glass can transform microphone technology. Additionally, it can transform data transportation. In glass microphones, one of the most interesting things is that they can withstand temperatures of up to 1000 degrees Celsius. That allows researchers to use it in extremely high temperatures. The system involves a sensitive membrane. Along. With the internal microbeam, which the system uses as a measurement tool. This is one of the things that can revolutionize signal technology. 

The same technology that makes it possible. To use the glass as a microphone. Makes it possible. To use the glass as a quantum network or quantum sensor. When a laser beam travels through the glass. If something presses it. This. Causes an anomaly. In. The quantum channel. And that causes a change in the laser beam that travels in that channel. 

“Continuous-variable (CV) quantum information processing, which underpins technologies such as quantum key distribution (QKD) and quantum random number generation (QRNG), depends on accurately measuring the phase and amplitude of light. To do this, a coherent receiver is required. This component mixes a very weak quantum signal with a stronger reference beam and extracts information from the way the two interfere.” (ScitechDaily, Physicists Turn Ordinary Glass Into a High-Speed Quantum Security Device)

"Laser-written glass chip demonstrates the potential of glass platform for quantum communication. Credit: Marco Avesani, University of Padua" (Physicists Turn Ordinary Glass Into a High-Speed Quantum Security Device)

ScitechDaily mentions the next things that support the use of glass in quantum technology.  

*************************************

*Fixed and tunable beam splitters

*Thermo‑optic phase shifters for precise electrical control

*Three‑dimensional waveguide crossings

*Polarization‑independent directional couplers

These elements allow the quantum signal and reference beam to interfere in a controlled manner. So that two conjugate quadratures can be measured at once. 

The device also demonstrates:

*Extremely low insertion loss (≈1 dB)

*Polarization‑independent operation

*Common‑mode rejection ratio above 73 dB, indicating strong suppression of classical noise

*High signal‑to‑noise stability over at least 8 hours of operation

*These characteristics meet or exceed those of many silicon‑based photonic receivers

*Environmental stability: Glass is inert and resistant to thermal and mechanical fluctuations.

*Low‑loss fiber coupling: Waveguides closely match the size of standard telecom fibers.

*3D design freedom: Circuits can include crossings and complex routing without added scattering.

*Scalability and cost‑effectiveness: Femtosecond laser writing enables. Rapid prototyping. Without expensive semiconductor processing steps.

(ScitechDaily, Physicists Turn Ordinary Glass Into a High-Speed Quantum Security Device)

*************************************

The microbeam that travels through ordinary glass can help to create the new optical quantum networks. The minimum scattering makes the eavesdropping almost impossible. There are two possible ways. To make this kind of network. Using normal glass plates. There is a possibility that laser beams travel in the tubes or channels. Those are in the glass plate. The system can make those laser channels by drilling them using another laser in the vacuum chamber. 

That keeps those channels clean. A laser can drill those channels in the glass vertically. Or it can use two glass plates. Then the laser makes. The form of laser channels. That the photonic microchip requires. To both glasses.  The system. It puts micro-optics. To the right points. Of those channels. Then the laser welds both of those glass plates together. 

And then the laser beam scattering is minimum. Gas atoms and other dirt in those channels make the scattering effect.  Another way is to pull the glass fiber through the glass plate horizontally. The glass can protect information that travels in the glass fiber. Or in some visions, the laser can transport information through the glass. 

https://scitechdaily.com/this-microphone-made-of-glass-can-survive-extreme-heat-and-electromagnetic-chaos/

https://scitechdaily.com/physicists-turn-ordinary-glass-into-a-high-speed-quantum-security-device/

The ability to store the entire internet in DNA is one of the most intriguing ideas in data science.

“An illustration shows a strand of engineered DNA passing through a nanoscale sensor, where its physical structure can be decoded as digital information. DNA nanostructures could one day serve as ultra-dense carriers of digital information and advance the field of data encryption. Credit: Jason Drees for the Biodesign Institute at ASU” (ScitechDaily, Storing the Internet in DNA? Scientists Say It’s Closer Than You Think)

Nanotechnology makes many things. Like. interstellar travel. Possible. One of the ways. To make the interstellar trip is to send a capsule to another solar system. That capsule can carry cells inside it. Or it can search for bacteria from the targeted planet. Then the ground station can send the DNA code to the spacecraft in digital form. Then, nanotechnology creates artificial DNA and replaces the bacteria's DNA with that artificial DNA. 

The DNA stored at the absolute zero point, or 0 Kelvin degrees, remains forever. Maybe. That thing can be the motive for the trip to Pluto. If. We store all our data in the form of DNA in the Pluto or Kuiper Belt. That kind of thing allows researchers to store that data forever. 

That kind of data storage makes it possible. To secure data in new ways. The DNA that carries data can be carried. In. The USB stick. The person puts that USB stick into the system, which reads the DNA. 

This allows ultra-secure data transmissions. In the network, over short distances, the pneumatic mail system can transport those DNA capsules. The DNA will be stored in the carrier, and then the pneumatic system shoots that capsule over the network. In longer distances, the system can store the DNA as a digital code. And then. Nanotechnology puts atoms or amino acids in the right order. This is one of the things that the high-power AI-controlled systems can make. 

And that system can do much more than just store data in the DNA. This kind of technology makes it possible. To transport data straight into human brains. The artificial bacteria. Or. An amoeba can send neurotransmitters or electric impulses into the neurons. 

“ Scientists are exploring how DNA’s physical structure can store vast amounts of data and encode secure information. Since computers first began shaping modern society, scientists have faced two ongoing problems: finding ways to store rapidly growing amounts of digital information and ensuring that this data remains secure from unauthorized access.”(ScitechDaily, Storing the Internet in DNA? Scientists Say It’s Closer Than You Think)

“Researchers at Arizona State University’s Biodesign Institute, working with collaborators, now report an unexpected solution. Two new studies demonstrate that DNA, the same molecule that carries genetic information, can be used to reliably store vast quantities of data while also enabling strong encryption.” (ScitechDaily, Storing the Internet in DNA? Scientists Say It’s Closer Than You Think)

The ability to store data in the DNA makes it possible to create new types of robotics. The biorobot or bacteria can transport information into computers using bioelectric impulses. But that technology also makes it possible to create things that nobody expected. 

The same technology. That makes it possible to store the entire internet. The DNA makes it possible to create customized humans. Human DNA can be involved. Genes that control the growth of the gorilla muscle. That allows researchers to create superstrong humans. 

But this technology makes interstellar flight possible. As. I wrote at the beginning of this text. The idea is that there are some cells in the spacecraft. And when the spacecraft travels to the solar system, there are habitable planets. The system changes the DNA in those cells. This makes it possible to create artificial creatures. That will look anything like the original species that sent them. 

Maybe. That kind of space trip is our last chance in this case. That. Our sun lost its fuel. And turns into a red giant. 

https://scitechdaily.com/storing-the-internet-in-dna-scientists-say-its-closer-than-you-think/

AI drives a Perseverance rover for two days.

“This annotated orbital image shows Perseverance's route during its second day of autonomous driving on Dec. 10th, 2025. The magenta line shows the AI-planned route. And. The orange line shows the actual route. Image Credit: NASA/JPL-Caltech/UofA” (UniverseToday, NASA Let AI Drive The Perseverance Rover For Two Days)

NASA lets the Perseverance rover drive for two days under AI control. This is a crucial step for AI-driven rovers. If humans want to send rovers to Jupiter's icy moons, and beyond that, it means those rovers must have the ability to operate independently. The ability to operate without non-stop communication with ground stations is vital because the signal travels between Earth and Jupiter for 32-57 minutes. This makes it hard to control robots on Jupiter’s icy moons from Earth. 

The ability to operate independently is important for many reasons. If. The rover goes into a cave and loses contact with Earth. That means the rover must know how to reestablish contact. The ability to travel into caves is important. Because. It's possible to find remnants of ancient life in those places. Caves protect DNA or RNA remnants, if they exist, from the UV radiation. If. There were. Some lifeforms. On the red planet. 

The remnants of those cells. And their genetic material could most probably be found in caves. In other places, the UV radiation destroyed those molecules. The independently operating robots can also operate underwater and on the other side of the moon. The ability to operate independently will be made possible. To collect samples from places where the Earth is invisible. 

The Radioisotope thermoelectric generator (RTG). It can give the man-shaped robot an unlimited operational range. Those robots can operate underwater, and those conditions can be simulated. Conditions that those systems can face on distant moons of the giant planets. 

The robot is a good tool. To research underwater caves and shipwrecks. Those robots can operate under heavy pressure. And they must not care about the decompression or any other things. The radioisotope generator gives those robots almost unlimited operational time. Another possibility is to equip those underwater robots with capillary tubes that allow the water to travel through them. 

That water flow can rotate generators. The underwater robot can use the same technology as Perseverance. But the system requires more resistance against pressure. But. The underwater robot. It can look like a man. A human-shaped robot that uses radioisotope batteries can operate for years. Those systems are interesting because AI-driven robots can be used to build the Mars and lunar space stations. 

https://www.universetoday.com/articles/nasa-let-ai-drive-the-perseverance-rover-for-two-days