Proceeded with the official documentation JSONL conversion using the Markdown → JSONL Converter, increasing the collection from 1,074 to 1,180 items
Worked with an SSAFY peer early in the morning, discussing AI-related philosophy, astronomy, and various hypotheses, which felt like my thinking broadened considerably
While talking with the SSAFY peer about “Is AI not the same as humans?”, I considered AI learning and human learning side by side
The process of a baby learning the world through senses like sight, hearing, and smell seemed like SFT, and a parent correcting repeated words could be seen as DPO
Imagined that if we built an AI simulation, we could include a childhood AI, a trained AI, and descendant/extinction/birth events, as well as natural events like Earth formation, evolution, meteor impacts, and earthquakes
If all AIs shared the same philosophy and safety rules, they might converge to an average or collapse, so an ecosystem of AIs with different philosophies and capabilities might be needed
Age‑related memory decline or dementia can be likened to AI hallucinations, and journaling/reading/context grasp feels like continuous learning and memory correction for humans as well
Briefly checked the reference material shared by the SSAFY peer
Docusaurus Playground: A page for trying out the Docusaurus documentation site directly in the browser
DesignCompass’s Meta Astryx article: An article about Meta’s publicly released React + StyleX‑based design system. It was interesting that the design system was crafted to be understandable and usable by AI coding agents
As the README grew, I judged that Qwen PR Review might miss the actual code‑review context while reading the README body
Modified the Qwen PR Review workflow to exclude README.md from the review input
Recorded entries 23 through 32 with the Qwen Validation Debugger