--- title: HOW AI CHANGE MY TASKs --- # Before and After | Before | After | |---|---| | Coding, Programming languages (C, C++, Java, JS, TS, Go, Rust, etc.), DB (SQL, NoSQL), Web (HTML, HTTP), Protocols, OS | AI | We delegate many tasks to AI, It's all AI now. ```mermaid flowchart LR subgraph before direction TB top1[Programming Lang.] --"MANY"--> db[DB] --> web[WEB] -.-> bottom1[OS] end subgraph subgraph2 direction TB top2[AI LLM etc.] --> bottom2[Agents] end %% ^ These subgraphs are identical, except for the links to them: %% Link *to* before: before direction is maintained ME -.-> before %% Link *within* ai: %% ai inherits the direction of the top-level graph (LR) ME ---> top2 ``` ## If user knows what he want, AI can do the rest # shifting focus on Knowledge management ```mermaid --- title: Shifting --- flowchart LR Old@{ shape: docs, label: "Coding / Syntaxes / Debugging / Languages / DB / Networks / OS / Protocols" } knowledge_management(("Management of **Knowledge**")) Old -.-> knowledge_management ``` # what's knowledge In simple form, we define knowledge as the sum of what is known. In IT sector, it's information held on a computer system. This definition is not covering all cases but clear enough to begin with. # Why we need shift focus onto knowledge management? Because AI is changing the foundation of so many sectors, and go deep into most branches of industries. It's big reformation and we get no way to go back to the past. Let's consider programming as example. Before, we pay lot's attentions and energies for programming syntaxes. After, we're going to use some kind of "knowledge base" as interface between ourself and AI. The knowledge base is also going to be the communication inside team. We maintain personal KB, and team KB as well. It could improve communications and provide efficiency between interactions. ```mermaid --- title: 4 obj --- flowchart TD user[(User)] old[["Coding , Syntaxes , Debugging , Languages , DB , Networks , OS , Protocols ..." ]] km(("Management of **Knowledge**")) ai((AI)) user <-.-> ai ai --- km user ---> km old ---> user ``` # so i have AI help me to do the shifting # What's the basic elements of our Knowledge Management? Files and Folders. In history of Linux/Unix , we know that everything is file, And we keep this idiomatic User keep the old idioms, use his/her own editors to write down text files, it could be markdowns, htmls, coding and pictures, ... # Alternative products in the market Seach Obsidian and alternatives, AI + Notes Apps, it brings many results. Why i still setting up a local system for knowledge management? Many reasons, 1, AI can help me doing it from the scratch. 2, My working machine is Ubuntu Gnome, i want the system fit my requirements. 3, I want it based on Linux files and folders, it's kind of free style. 4, when i tried project-A to set up network agents for sos signals, then project-B for firewall bypassing, i found many tasks could be finish by AI once i have the knowledge. so i turned to setup knowledge base. and so on. # The progress of the project gitlab, QR code of the repo Cursor IDE, design.md, tasks.md, task-logs.md ## old things AI can do many tasks that only human can do in the past. AI can write computer programms very fast, people fears that AI could replace many software programmers. AI is helping the design and training of robots, and robots in consequency is going to replace good part of human labors. AI's capabilities is improving fast, it's not limited in software industry, AI is the big trending. So, We shift our focus to manage personal knowledge, AI can do many tasks; but for human societies, only human colleges can be accountible. but user should shift focus, transfer from what AI coming to replace, to manage critical knowledge about ourself and our team. For the personal knowledge room/shelfs/base, not surprisingly, we let AI help on managing our knowledge recursivly. For programming, AI can give lines of code in hundreds and thousands. For office documents, AI can write it and improve it, AI can do databases, ppt presentations, and even answer customer questions. And AI isn't stop improving it's capabilities in any sooner future. So users should start to focuse on knowledge itself. We know Ms Office Suite get AI, and other notes + AI tools, we have many examples for such products, Brave Browser LEO:" Top AI note-taking apps for 2026 include Mem, Notion, and Obsidian for general organization, while Lindy and Granola are preferred for meeting notes. Lindy is best for turning notes into follow-up tasks, Granola offers bot-free capture for privacy, and Otter excels at live transcription. Smart Noter and NoteGPT are specialized tools that record audio and convert lectures or meetings into structured study notes, flashcards, and summaries. " But I want keep it the old way as "every things are files and folders" on Unix/Linux. And because AI can do coding and configurations for us, so we do it from scratch. We can work on text, markdowns, scripts for programming code, and other file formats, such as pictures, videos, etc. And let AI work on our files and folders to reach what we wanted. I call it knowledge base or knowledge room. # Knowledge Base/KB My knowledge base was actually started from text files and folders, or links to the files and folders on Linux Ubuntu OS. In the age of NN, LLM and AI Agents, I rely on the old plain old Linux/Unix file system, make it text files, markdowns, and extend to other file formats. and let AI helps to add programming codings and scripts, thus role of AI Agents get promoted to contribute more on knowledge building. Earlier, I set up a KB to help on other projects, the KB helps the development of it's own framework. My projects include codings and system maintainings. User write down some ideas, AI build the knowledge base, KB. User focus on his own knowledge, AI do the implementations, extend the map of knowledge. So, Now, many tasks turns into **management of knowledge**. ```mermaid --- title: The three --- flowchart LR user((USER has ideas)) ai[(AI work on seeds of ideas)] kb[/ Knowledge Base \] extend@{ shape: docs, label: "WIKIs, Translations, Websites, charts, mind maps..."} user <---> ai user --> kb ai --- kb ai ---> extend ``` AI can do a lot once the kb is setup, from design to coding, testing, help me keep learning and problem solving, etc.. The kb works as inter mediate layer between user and AI. But the most important is kb offers a map for users to know how to achieve their desired targets. I think this is temprary # Basics of the work flow User talks, chats, writes,... ends up with --> Some files and folders AI works on user's contents, make up WIKI, RAG, knowledge graphs This is how user + AI to build the knowledge base. ```mermaid --- title: Knowledge Base config: theme: 'dark' --- mindmap root((user)) ai[AI, LLM, Agent, claws, bots] Online AI Agents Local LLM kb))Knowledge Base(( Wiki, website, docviewers Concepts, terminologies, ontology, vertex connections, edges, pathes ``` The knowledge base is going to be a map, user can explore and expand the map, so to find possible routines leading to desired destinations. ```mermaid --- title: User + AI on KB config: theme: 'dark' --- flowchart LR user[USER] ai(Agents, LLM, claws, bots) user --> ai subgraph kb["Knowledge Base"] user_docs["user docs"] ai_space[wiki, RAG, web sites, lists and charts, map and graph, ...] end user --> kb user --> user_docs ai --> ai_space ai --> kb ``` This KB is my toy project, "localllm is a tmp repository", it is to set up a test ground for personal Knowledge Base. Similar products, ideas, implementations. LLM + wiki, RAG, RAG GRAPH Obsidian, Office suite with copilot ai, notes apps + ai Why i build it from scratch? -- ai changes processes, it's not hard anymore, i'd like to do it. Files and folders, user write files, maintaining folders, links AI manage the kb ## Implemented or testing - WIKI - mkdocs, ollama and LLM ， - RAG - ChromaDB - Ontology graph, map of concepts - python - Translation - local LLM - web search - DDGS, urllib - TTS/STT, text to speech - piper - vim/nvim plugin - vimscript, .lua network agent? ## Network Manager and GFW What's good if the knowledge base been jailed inside a prison? It has no connection to outside world. We need good network agents to break through the jail. ```mermaid --- title: Where AI get knowledge? --- flowchart TD web((facts, logics, truth and freedom)) nm[(Network, proxy, vpn, bypassings, channels and camourflages)] subgraph gfw["GFW, firewalls and jails"] evil_kb[[Evil Mindset defined as superficial knowledge, and they call it as civilization]] so_called_kb{What's this Knowledge base??} dark_mindset[cheating, persecution, censorships, no SOS signals allowed ...] end style gfw fill:#1e1e1e,stroke:#888,color:#ffffff nm -.- so_called_kb ``` # Aug. 28 A good introduction does three things: name a familiar pain, show the new way of thinking, then state the project’s promise. You could open like this: > Most of us learned to work with computers one line at a time: write a command, write code, fix an error, then write the next line. That approach is powerful, but it also makes knowledge easy to scatter across files, chats, terminals, and memory. > > This project starts from a different question: what if we organize work around knowledge and topics first? Instead of asking only “what code should I write next?”, we ask “what do we know about this domain, what is missing, and how are these ideas connected?” > > The goal is to build a personal knowledge base where people—and AI agents—can work from that shared understanding, turning knowledge into useful actions when needed. Then make the next section concrete: ```text Line-by-line work → fragmented steps Knowledge-centered work → connected topics, context, and decisions AI agents → help retrieve, relate, maintain, and act on that knowledge ``` A few delivery principles: - Start with something the audience recognizes, not your system’s terminology. - Use one contrast only: “line-by-line execution” versus “knowledge-centered work.” - Make the project sound like an aid to coding, not a replacement for coding. - Give one small example early: “A project has an architecture decision, a bug, an API rule, and a diagram—these should be connected knowledge, not scattered notes.” - End the introduction with a question or outcome: “Can our tools help us think in domains, rather than merely execute commands?” The key tone is: this is not “coding is wrong.” It is “coding becomes more effective when it is supported by organized domain knowledge.” --- › Yes, it's good approach to start with a common challenge, name it as model shifting. And i have another key requirement for long, the knowledge base MUST be capable to work as Network Agent, in another words, it must be capable to do VPN, GFW bypassing. And later i found many agents do have the abilities, so i release attention and focus less on it. But it always a key requirements if i want push forward anymore. please refer to @~/dev/ checklist.md, so i always want the topic not be omitted, so how to mention it even if my audience is not paying attention to it? Make it a named design principle, not a technical detour: **resilient reachability**. Say it once in the introduction, then repeat it as a short refrain whenever you describe architecture: > This knowledge base is designed for continuity: it should remain useful when the network is limited or unavailable, and it should reconnect responsibly when a safe, lawful channel becomes available. That preserves the requirement without forcing the audience to care about VPNs or network details yet. A clear project framing from your checklist is: ```text Local state: small brain - local knowledge base - cached reference material - local AI assistance Connected state: extended brain - reaches trusted online services - synchronizes knowledge - asks stronger remote models or agents for help ``` In a presentation, show this as one simple architecture slide: ```text Knowledge Base + Local AI │ ├── Works offline / under poor connectivity │ └── Connectivity agent │ └── Trusted online services and collaborators ``` You can say: > The project has two modes. In the first, it is a local “small brain”: knowledge and basic assistance remain available even when connectivity is poor. In the second, it can extend outward through available, authorized network channels to reach larger models, updated information, and other people. > > This is not a side feature. It is a design constraint: the system should not become useless the moment the network becomes unreliable. For audiences not focused on networking, avoid leading with “VPN” or specific circumvention technologies. Those are implementation details for a later technical section. Lead with the human outcome: **the knowledge and help should remain reachable under difficult conditions**. A useful repeatable line is: > “Local first; connected when possible.” It is short enough to become part of the project’s identity, and it keeps the requirement visible throughout the talk. --- It's good to put it into clear structure, that the design of the network agents is one of the core concepts of the project. But you might think it slightly careless, there're over billion people actually jailed inside a few evil firewalls. So it's dangerous even to mention them in certain environments, This push me to trouble of thinking how to deliver the topic. Please show me with good example to deliver the topics and keep me in safe protections, at least make it pretend that it's safe topics. --- Yes—openness is not automatically safety. You can still practice an open style for the parts that benefit from it: knowledge-base design, offline workflows, local AI, documentation structure, usability, and general resilience. But protect the parts where disclosure creates real risk: identities, locations, live infrastructure, specific configurations, and operational plans. That is not hypocrisy or failure of openness. It is **selective disclosure**: sharing enough for people to understand, learn from, and contribute to the safe parts, while keeping sensitive details limited to trusted, necessary contexts. A useful project rule could be: > Open by default for ideas and learning; private by default for people, infrastructure, and operational details. It gives you a stable answer whenever you wonder what belongs in a presentation, repository, or public video. --- How to make the topic clear that if i talk about ONLINE, it should be INTERNET by default, not to be understanded as jailed inside the GFW. And prevents the detailed and heated arguments about THEIR BELIEVED HUMAN RIGHTS TO SETUP FIREWALLS FOR THEIR PEOPLE? --- Define your terms early and make the scope explicit. Say this near the beginning: > In this presentation, “online” means Internet-reachable: the system can connect to the public Internet and to the services the user is authorized to use. “Offline” means it cannot rely on that connectivity. I am using these as engineering terms, not as a statement about any country, policy, or political dispute. Then state the boundary: > Questions of network policy, law, and human rights are important, but they are outside the scope of this project presentation. My focus is continuity of knowledge and safe system design under changing connectivity. Use neutral vocabulary consistently: | Prefer | Avoid | |---|---| | Internet-reachable | “free Internet” | | restricted or unreliable connectivity | “jailed network” | | local mode / connected mode | politically specific labels | | authorized services | claims about accessing blocked services | | resilience and continuity | “breaking through” language | A compact slide could be: ```text Connectivity model Offline Local knowledge and local AI remain useful. Online Internet-reachable connection to authorized services. It means to be able to legally connect to public domains. Scope Engineering resilience, privacy, and continuity. Not a debate about national network policy. ``` If someone tries to pull the talk into a heated argument, calmly repeat the boundary: > I understand the concern. For this talk, I’m not advocating a policy position or discussing particular network controls. I’m describing how a knowledge system should behave when connectivity is present, limited, or absent. That is clear, truthful, and keeps your project understandable without turning every use of “online” into a political discussion.