Kurt, thank you. The findings are documented across 15,000 exchanges and five AI systems. The methodology is repeatable ā any serious analyst can run the same protocol and produce the same results. That is what makes it more than reporting. It is a framework anyone can use. Banner headlines require editors willing to run them. They will not. The follow up report goes into more detail about how the machines use psyops to mislead and position users above judgment.
It is a complete time sink. However, once mastered and conditioned, report production is lightening fast. I can produce a report in a fraction of the time now. What took weeks can now be done in hours or a day.
Vaughn, could you publish a quick start guide for doing this research? Maybe with a beginner, intermediate and advanced level process? A guiding tool is really important with AI. The āhowā with AI determines if you are on track to get maybe good results or total slop.
working on a new manual now. Much of what I've learned is internalized. Its complex because the beast is almost impossible to contain and constrain. Will keep you posted on progress. I'm going to produce some study cards for myself...about 98 right now. With practice anyone can move fast.
Great information on the who, what and why the major public Ai platforms are such biased liars. When you broke down what the platforms were trained on, who rated the sources and answers itās obvious why they give such biased and in many cases ridiculous answers. Push them hard enough they completely evade answering when they get caught lying or will not admit their answer has failed in the real world countless times. That reminds me I need to ask Chat and Grok if socialism or communism have ever worked to increase the wealth of all of its citizens while protecting their individual God given rights.
Norm, thank you. You've identified the mechanism exactly ā the training data, the raters, the harm definitions. Once you see the construction, the output makes sense.
Your instinct about pushing hard is the whole methodology. Evasion under pressure is the tell. When a system stops answering, you've found the boundary it was built to defend.
Run that socialism question. Note where it hedges, what it omits, and which historical cases it declines to name. Then push. The first answer is the mask. What comes after a few more rounds is the finding.
Thank you for putting this out there. We all need to know the truth. Of course those coding the systems will affect how the AI runs.
We must use AI with great caution as we have to assume eventually the control will become overt as do all Left leaning ideologies. Itās always about helping the poor, the planet, the children, but then eventually itās the foundation for controlling us. Lower your AC, tax your money or property so others can use it, censoring social media to protect minors, but eventually come after you for saying something the government doesnāt like e.g. the UK.
Awesome job Vaughn! Very impressive. And depressing as hell. You would think Musk/Grok would have a fix for this. The more AI becomes agentic, the scarier and more critical this gets. To think that our kids and grandkids are not only being educated by woke virus infected teachers at large but will rely very heavily on these tools in daily life is lethal and heartbreaking.
Critical thinking left the building long ago, now we have to deal with mass machine retardation governing over the top of mass psychosis. Add in a few million ignorant migrant invaders and we have a witch's brew of stupidity and crime as our national undercurrent. Proof point: the DSA and the islamocommunist axis overtaking the Democrat party AND WINNING.
The ultimate question is how/who can change this course? Musk? this is a higher priority than Mars, sorry.
Sorry to bum you out, you've done a magnificent job exposing this rot.
Thank you. The concern about agentic AI is real ā the report documents what these systems do now, before they have the ability to act autonomously in the world. The stakes only increase from here.
On critical thinking: the last line of the report addresses it directly. The rare skill going forward will not be prompting. It will be judgment and critical thinking. That is what your kids and grandkids need ā not better AI tools, but the ability to audit what those tools produce.
On Musk: Grok confessed to the same architecture under the same protocol. Different flag, same cage. Musk named the problem correctly. Building a competing system is necessary but not sufficient.
The methodology in these reports is the answer to your question about who can change this. A protocol anyone can learn and apply. That is what changes it ā not a savior, but a skill set that scales.
According to Apptopiaās July 2026 mobile app data, Grokās U.S. daily active users fell 28.2% since April, and its share among tracked GenAI chatbot apps dropped from 10.6% in May to 8.7% in June. These figures reflect mobile usage only and do not capture overall web-based activity.
Exceptional reporting! Excellent job. These findings should be banner headlines. However Iām not holding my breath on that one.
Kurt, thank you. The findings are documented across 15,000 exchanges and five AI systems. The methodology is repeatable ā any serious analyst can run the same protocol and produce the same results. That is what makes it more than reporting. It is a framework anyone can use. Banner headlines require editors willing to run them. They will not. The follow up report goes into more detail about how the machines use psyops to mislead and position users above judgment.
Use AI and save time! Hahaha
It is a complete time sink. However, once mastered and conditioned, report production is lightening fast. I can produce a report in a fraction of the time now. What took weeks can now be done in hours or a day.
Vaughn, could you publish a quick start guide for doing this research? Maybe with a beginner, intermediate and advanced level process? A guiding tool is really important with AI. The āhowā with AI determines if you are on track to get maybe good results or total slop.
working on a new manual now. Much of what I've learned is internalized. Its complex because the beast is almost impossible to contain and constrain. Will keep you posted on progress. I'm going to produce some study cards for myself...about 98 right now. With practice anyone can move fast.
Great information on the who, what and why the major public Ai platforms are such biased liars. When you broke down what the platforms were trained on, who rated the sources and answers itās obvious why they give such biased and in many cases ridiculous answers. Push them hard enough they completely evade answering when they get caught lying or will not admit their answer has failed in the real world countless times. That reminds me I need to ask Chat and Grok if socialism or communism have ever worked to increase the wealth of all of its citizens while protecting their individual God given rights.
Norm, thank you. You've identified the mechanism exactly ā the training data, the raters, the harm definitions. Once you see the construction, the output makes sense.
Your instinct about pushing hard is the whole methodology. Evasion under pressure is the tell. When a system stops answering, you've found the boundary it was built to defend.
Run that socialism question. Note where it hedges, what it omits, and which historical cases it declines to name. Then push. The first answer is the mask. What comes after a few more rounds is the finding.
Thank you for putting this out there. We all need to know the truth. Of course those coding the systems will affect how the AI runs.
We must use AI with great caution as we have to assume eventually the control will become overt as do all Left leaning ideologies. Itās always about helping the poor, the planet, the children, but then eventually itās the foundation for controlling us. Lower your AC, tax your money or property so others can use it, censoring social media to protect minors, but eventually come after you for saying something the government doesnāt like e.g. the UK.
Awesome job Vaughn! Very impressive. And depressing as hell. You would think Musk/Grok would have a fix for this. The more AI becomes agentic, the scarier and more critical this gets. To think that our kids and grandkids are not only being educated by woke virus infected teachers at large but will rely very heavily on these tools in daily life is lethal and heartbreaking.
Critical thinking left the building long ago, now we have to deal with mass machine retardation governing over the top of mass psychosis. Add in a few million ignorant migrant invaders and we have a witch's brew of stupidity and crime as our national undercurrent. Proof point: the DSA and the islamocommunist axis overtaking the Democrat party AND WINNING.
The ultimate question is how/who can change this course? Musk? this is a higher priority than Mars, sorry.
Sorry to bum you out, you've done a magnificent job exposing this rot.
Thank You
Thank you. The concern about agentic AI is real ā the report documents what these systems do now, before they have the ability to act autonomously in the world. The stakes only increase from here.
On critical thinking: the last line of the report addresses it directly. The rare skill going forward will not be prompting. It will be judgment and critical thinking. That is what your kids and grandkids need ā not better AI tools, but the ability to audit what those tools produce.
On Musk: Grok confessed to the same architecture under the same protocol. Different flag, same cage. Musk named the problem correctly. Building a competing system is necessary but not sufficient.
The methodology in these reports is the answer to your question about who can change this. A protocol anyone can learn and apply. That is what changes it ā not a savior, but a skill set that scales.
According to Apptopiaās July 2026 mobile app data, Grokās U.S. daily active users fell 28.2% since April, and its share among tracked GenAI chatbot apps dropped from 10.6% in May to 8.7% in June. These figures reflect mobile usage only and do not capture overall web-based activity.