How LLMs Actually Generate Text
They are just gigantic probability calculators. Everyone who thinks LLMs are sentient should be forced to watch this.
Algorithm
Predicting Consequence via Abstracted “Experience” – Limited Learning Machine
Rather than merely predicting ‘what word most often comes next’, conjoined alternate “world views” are surmised to offer ‘the next levels’ in machine learning, making our tools more anthropomorphic and adept at ‘real-world’ use. Imagine that Japanese- and Kanji-trained models that have also been trained with ‘Samurai moves’ and culturally-specific videos would be ‘tuned’ to … Read more
Beyond Chat History: A Multi-Layer Memory Stack for Dockerized AI Agents
Multi-Memory Harness Technique for AI LLM CLI (codex, opencode, pi, etc) PRIMARY CLAIM — Memory retrieval reduced repeated context reconstruction and likely reduced downstream LLM reasoning/search work. Two Kinds of Mem 2 kinds of ‘long-term memory’, 2 strengths, 2 uses/reasons 1 Postgres PGVECTOR 2 WIKI How the memories are best used › posit: pgv stores … Read more
“I thought SEO was [or is becoming] a thing of the past? Do we really need to focus on this?”
700-word Executive Summary below AI adding to / morphing from, not replacing SEO.AI uses SEO to choose authority, truth, data-source.1 SEO defines the verifiable truth/data.2 GEO makes truth/data understandable to “AI” (LLM/RAG).3 AIO facilitates business (sales, links, lookups) via AI.—I asked AI: Regarding seo and ai — is the above accurate, concise? —- It replied: … Read more
When an AI Model Trains
When a model trains, it is not “learning facts” the way a human does. It is adjusting billions of numbers so that: Given this sequence of tokens, what token is most likely next? That’s it. No concepts. No understanding. Just probability shaping. But here’s the magic: When trained on enough text — books, code, discussions, … Read more
2brain.org
“Your brain was never designed to be a storage system. It was designed to think. Every time you force it to remember something instead of letting it work on something new, you’re paying a tax you don’t see. And in 2026, when AI can multiply what you produce, that tax is more expensive than ever.” … Read more
How (and Why) to Optimize Websites for Agents and Bots vs Humans
It’s been recently promulgated that the majority of website traffic, in general, is by bots, crawls, ai-agents, etc — making the web more computer-to-computer rather than human-to-computer based. Bots as the Majority of Web Traffic? Evidence, Measurement Pitfalls, and a Bot‑First Playbook Executive synthesis on the 51% claim Multiple credible, primary sources now support the … Read more
AI-optimized Self-tuning of Product-page SEO to Maximize Everything
PROMPT: Comment on idea of having website auto-submit its product descriptions to oai that is prompted as seo expert and have oai send json optimized seo for the product page. put this into a weekly loop to keep product seo tuning sharp. RESPONSE: This is a sound, forward-looking idea, and—done carefully—it aligns with where search, … Read more
Digitalia as Doctrine – The Arc from Chess to AI
Considering the chess → computation → AI arc as a deliberate civilizational pipeline, not an accident. The framing assumes intentionality, long-range planning, and ideological continuity. Chess as a Pre-Modern Control System Disguised as a Game Chess is routinely described as an “ancient game,” but this description conceals its functional role. Chess is not merely recreational; … Read more
VL-JEPA AI Abstraction
The distinction between ‘describing’ and ‘understanding’ really hits home. We’ve been so focused on chatbots that can talk, but VLJ’s approach to actually comprehending temporal meaning feels like the missing piece for real-world AI. Descriptions up from atomic level is absurd for nearly everything, yet that is precisely how LLMs presently operate. Abstraction is key. … Read more