Arguments that survive contact.
A working knowledge base for technical arguments, architecture notes, mathematical models, research fragments, and longer-form essays.
AI diagrams are supposed to simplify complexity.
But arrows, ladders, nested circles, and “evolution” timelines do more than organize terms —
they make claims about how those terms relate.. .
The question is not only whether the labels are correct.
It is what false mental model a competent novice is likely to retain.
It's mathematically incoherent. "Best for creative content" has no task, no H, no L, no version, no date. A model that scores 9 for coding, scores 2 for image gen. (same logo, opposite out come.)
Claim: Tools can be totally ordered by marketing label.
Result: Information loss, apophenia, decision paralysis.
This page kills that chart.
We tested 8 publicly available GenAI tools: ChatGPT, Grok, Gemini, Claude, Perplexity, Meta AI, Mistral, DeepSeek...
... on 6 tasks with the same source set. Scores are task-based, not brand rankings.
TOn the page link you use the row relevant to your work (not the column). Click on the link below to find the task-based Scores (and the scores are not brand rankings.)>
A single 1-10 leaderboard feels definitive.
It flattens context, hides tradeoffs, and rewards brand memory over task fit.
A model that scores 9 for coding can score 3 for image gen — same logo, opposite outcome. .
This sheet replaces brand ranking with task evidence.
Each cell is scored on observed behavior for that task, not general "intelligence."
No aggregate winner — use the row relevant to your work.
The common graphic implies: AI ⊃ ML ⊃ DL → GenAI → RAG → Agents, a s if they were the same type.
They are not.
AI is the research program (Dartmouth, 1956). ML / DL / GenAI are method families within it.
RAG and Agents are not types of AI — they are architectural patterns: Retriever (1940s: file cabinets, Hollerith cards) + Generator (2017) + Controller.
Informatics vs File Cabinets vs Typewriters vs Email. You cannot leave out the file cabinet — the vector DB is the file cabinet.
To kill the whole “Types / Evolution / Landscape” bubble. The familiar ladder — Symbolic → ML → Deep Learning → GenAI → Agentic → AGI — confuses substrate, capability, and system architecture.
A real system is not a step on a ladder. It is a coordinate in S × C × A.
“Agentic is the new” collapses under engineering inspection. The agent is a wrapper around the model, not a new foundation-model class.
The engineering pattern is still perceive → reason → act inside a control loop.
Data quality cannot be defined independently of the hypothesis, architecture, and loss function. Data is a derived constraint from the decision problem, not automatically the starting point.
An inner cognitive or emotional sense that distinguishes right from wrong, guiding moral conduct and prompting remorse or satisfaction relative to an individual value system.
Read more ↗Mental processes involved in gaining knowledge and comprehension: thinking, knowing, remembering, judging, perception, language, and problem-solving.
Read more ↗Cognitive consciousness concerns the processing, interpretation, and use of information through mechanisms such as attention, memory, and executive function, with or without subjective awareness.
Read more ↗