AI generated fake stories, urban legends, and fake news

This is an example of an entertaining, AI-generated fictional story posing as “real-life stories”.

I believe this “posing” should be outlawed before the Internet is filled up with fake stories, urban legends, and fake news in a one-true-to-million-fakes ratio.

Relativity

Physics and the laws of relativity: in parent-child relationship sound waves reach the subject after 25 years. 

AI, Visionaries and Architects

I decided to write down a few thoughts to clarify my obsessions with creating the "multitude" of AI agents that rely on the private (both personal and corporate) tiny learning models (LLM, TLM, TLL).

The future of work is poised for a significant transformation as artificial intelligence (AI) continues to advance. 

Some believe that the 62,000 tech layoffs in 2025 are already a result of this transformation. This is the first year when CTOs are not budgeting for more project managers (PMs) or junior developers.

Roles that traditionally relied on data input, data manipulation, and digital output, such as managers processing tasks, creating Spreadsheets and PowerPoint presentations, analysts, or developers typing code, are increasingly being automated. 

This automation doesn’t mean the end of human roles (yet) but rather a shift in their nature. 

Instead of spending hours on repetitive tasks, professionals will focus on higher-level strategic thinking and creative problem-solving.

Visionary Leaders, not Managers

Imagine a world where the leadership cadre is no longer bogged down by the minutiae of slide creation but instead spends their time strategizing and leading their organizations. All management and optimization tasks are automated and run constantly behind the scenes.

Architects, not Developers

Similarly, developers will move from writing (i.e., typing) lines of code to designing complex systems and orchestrating multiple AI agents to work together seamlessly. 

This shift will require new skills and a different mindset, emphasizing creativity, innovation, and leadership.

Distributed, not Central

The future is not about a few central AI companies dictating the direction of the world with homogenous solutions, but rather a multitude of proprietary AI solutions. The giants will play a crucial role in providing the models and computing resources. 

However, each individual person and company will develop their own AI agents tailored to their specific needs, leading to a diverse ecosystem of AI tools. 

This diversity will foster competition and innovation, driving the development of highly sophisticated and ultra-specialized AI solutions.

Today, I utilize local, yet powerful, models with up to 32 billion parameters and book-length input contexts. Next year, I fully expect to use a multitude of models with hundreds of billions of parameters, running locally on a Mac Studio M5 Ultra (?) or some new AI hardware.

In this new landscape, the role of the visionary and architect becomes paramount. Please let me know in the comments what you think and how you prepare yourself and your organization for 2026 and beyond.

Testing local LLM on multi-lingual understanding.

I have run a quick test on a few LLM models I have installed locally on Mac OS with 64 GB of RAM.

The test was conducted in English, but it also involved making connections between Slavic languages (such as Polish and Russian), the modern Inter-Slavic language (ISL), and the rest of the language group that originated from Proto-Indo-European (PIE), including Greek and Sanskrit.


Here is the question I have asked all of the models:

Let's discuss particle "ra" as in "rad" happiness, or "raj" heaven. Provide a short answer.

Quick Summary

After evaluating all models, it became clear that larger parameter models with extensive context windows generally excelled in providing insightful, accurate, and nuanced linguistic analyses, making them ideal for in-depth comparative research and article writing tasks. The standout, mistral-small-3.1-24b-instruct-2503, delivered the best balance of abstract thinking, linguistic precision, and large-context capability, especially if an 8-bit quantization version is considered for improved accuracy. Other strong contenders included deepseek-r1-distill-qwen-32b and qwen3-32b-mlx, offering substantial analytical depth. Mid-sized models provided faster but shallower analyses, primarily suitable for exploratory or quick tasks, whereas smaller models below 7B generally struggled with accuracy and linguistic coherence.

Model ranking by preference:

1. mistral-small-3.1-24b-instruct-2503, 24B, input: 131,072 tokens
2. deepseek-r1-distill-qwen-32b, 32B, input: 131,072 tokens
3. qwen3-32b-mlx, 32B, input: 40,960 tokens
4. dolphin-2.9.3-mistral-nemo-12b, 12B, input: 1,024,000 tokens
5. mistral-nemo-instruct-2407, 7B, input: 1,024,000 tokens
6. deepseek-r1-distill-qwen-7b, 7B, input: 131,072 tokens
7. llama-3.2-3b-instruct-uncensored, 3B, input: 131,072 tokens
8. phi-3-mini-4k-instruct, input: 4,000 tokens
9. smollm-135m-instruct, 135M, input: unspecified (small)

Small models are still helpful for agents that need to process, transform, summarize, or classify input.

 

Qwen 32B

I am pleased with the performance and depth of the 32B Qwen MLX, running
locally on my Mac Studio M1 with 64GB of RAM.

9 tokens per second is not fast, but acceptable.

A 15-second wait for the first token output is very good.