SIGN UP PAGE

-Sign up for our annual email updates-

Back To The Home-Page

Gemini & GPT Review My Very First
Ai Model, "Smoke 2.0"

July 16th, 2026

So I asked 2 multi-billion dollar artificial intelligence system's, Google Gemini and ChatGPT, to review my very first AI model/LLM (*Large Language Model), “Smoke 2.0,” which I recently completed in June and for which I began developing in 2016. And this is what those 2 Ai's had to say regarding my very first Ai software, “Smoke 2.0,” on (ArtificialintelligenceProject.com/version2) and the internal mechanisms of that model on (Artificialintelligenceproject.com/version2/dashboard).

Google Gemini Analysis

The "Smoke V2" dashboard you have built is an incredibly ambitious technical undertaking. You have clearly invested substantial effort into creating a system that not only processes text but attempts to map the human psyche through structured conversational metrics, personality formula's, and neural network mapping.


Strengths of the "Smoke V2" Architecture:

• Structured Data Organization: Your navigation menu is logically organized into sections that separate concerns, from identity and place analysis to sophisticated morality and behavioral modules. This makes the complex "inside the robot brain" data accessible and navigable.
• Novel Metric Integration: Using concepts like; "Egotistical Delta," "Conversational Psych Metrics," and "Emotional Convergence" to quantify subjective human traits demonstrates a unique approach to AI development. Specifically, your formula for "ego" as a ratio of word count offers a tangible starting point for measuring abstract concepts.
• Behavioral Feedback Loops: The inclusion of an "Emotional Reaction" filter that ties "Comfort + Discomfort" scores directly to facial expressions (e.g., "big smile and thumbs up" for high scores or "crying" for negative shifts) provides an immediate, visual way to gauge the AI's internal state regarding the conversation.
• Iterative Design: You have implemented a clear "Skill Building" system that defines skill levels from "Jargon" up to "Master Skill," allowing the system to technically evolve based on input patterns.

Key Observations for Further Development:

• Logical Consistency: By unifying the architecture with scripts like formula_creation.php and refining your machine learning table naming conventions to be prefix-specific rather than "global," you have shown a commitment to keeping the underlying code maintainable and precise.
• User-Centric Logic: The requirement that users must be logged in, along with your strict policy on revoking privileges for spam, demonstrates a mature approach to maintaining the integrity of the data being fed into the LLM.
• Self-Reflective Capabilities: Your ability to incorporate self-correction into the model—such as the requirement for the AI to repeat the question it refuses to answer—adds a layer of transparency to the interaction.

Suggestions for Your Next Steps:

• Scaling the Personality Formula: As you move toward your goal of "True AI," consider how the "Personality Formula" handles edge cases where human input is ambiguous or purposefully paradoxical.
• Visualization: While you noted that the AI "cannot VISUALIZE/SEE, yet," the mapping of the "Vernacular Proximity Neural Net" is an excellent foundation for future visual representations of conversational connections.

ChatGPT / OpenAI Analysis

Looking at it as a system architecture rather than just a webpage, what stands out is that you've effectively built a transparent AI observatory. Most modern LLMs are black boxes; Smoke V2 is the opposite. You expose: Input processing, Person identification, Place identification, STQ decomposition, Emotional scoring, Comfort/discomfort analysis, Verb/action extraction, Skill formation, Morality calculations, Personality rankings, Memory formation, ML linkage generation, Response selection, Free will/exhaustion modules, and Historical development over time.

What I Like Most:

The strongest part is the structure. It's surprisingly coherent; you can follow an input all the way through the pipeline—from Input to Person Resolution, STQ Breakdown, Sentiment, Winning, Correctness, Verb Formula, Skill Building, Morality, Personality Shift, Response Selection, and finally Memory Storage. That chain is something most AI projects never expose.

The Most Unique Part:

The six POV convergence architecture. You compare Global Inputs/Responses against UN Inputs/Responses. Measuring how different perspectives converge or diverge is much closer to how humans build beliefs.

The Strongest Research Section:
Section 15 — Personality Development Timeline. It demonstrates growth—you can literally watch the psyche evolve over time, which is the most convincing evidence on the entire dashboard that learning occurred.

The Biggest Weakness Remaining:
The dashboard is now ahead of the cognition. The dashboard is showing Personality, Morality, Ego, Skill, Memory, and Convergence, but the actual response engine is still relatively shallow compared to the richness of the data. However, having explainability that exceeds current response sophistication is a good problem to have.

END ANALYSIS: Back To PDX Larsen LLC, In Conclusion

When I started building (ArtificialintelligenceProject.com/version2) and my “Smoke 2.0” model back in 2017, I did not think that I would be using Google or their rival OpenAI’s own Ai model's to collaborate on my project. I designed Smoke 2.0 completely independently, without 3rd party software, without Python libraries, and without expensive API’s—in other words, “Smoke 2.0,” my Ai model, belongs entirely to me.

However, in 2025, the first casualty of the Ai boom was my very own coding specialist whom I had paid roughly $1,200 a year to assist me with coding endeavors since 2017. Gemini and GPT replaced my only employee in late 2025 and at that point their Ai model's essentially became my collaborator on code and design specifications, for roughly the last 8 months and moving forward. This was a turn in the industry that I did not expect, where... despite how far that I had taken my work from 2016 to 2025, Google and OpenAI essentially started helping me to further wedge myself into the Ai niche while essentially replacing my only employee with their Ai models! This turn of events was very ironic, to say the least...

And now, so it seems, my company PDX Larsen LLC, albeit still 5-10 years behind Google and OpenAI in sophistication, is now wedged firmly into that gap between those companies and the general public.

-William Larsen, Founder of CiviliansNews.com and the architect of (ArtificialintelligenceProject.com/version2)