The fenbs blogAll articles
Page 57 of 99
- CLAUDE.md for Obsidian: Give Claude Code Your Vault RulesDescribe your Obsidian vault’s note conventions, source boundaries, and permitted edits in a short CLAUDE.md. Test that Claude Code loads and applies it. · 5 min read
- Claude Projects Disappeared? Checks and Recovery StepsClaude Projects going missing usually means an account plan change, an organization switch, or a sync issue. Here is how to check each cause and recover access. · 5 min read
- Claude Research vs. Web Search: Which Fits Your Question?Claude's research feature synthesizes multiple sources into a single report; a standard web search returns a list of links you evaluate yourself. Each fits a different type of question. · 7 min read
- Codex Context Window: Tokens, Usage, and What Carries OverThe Codex context window is a working budget, not a permanent project memory. Learn how to read usage, prepare for compaction, and continue with a checked handoff. · 5 min read
- Compact a Claude Conversation Without Losing the WorkPrepare a concise continuation before a long Claude session gets difficult to use. Preserve approved decisions, check the summary, and distinguish Claude chat from Claude Code commands. · 5 min read
- Copilot Studio Kit: Testing and Governance Tools ExplainedThe toolkit now documented as Copilot Agent Kit adds testing, analysis, and governance capabilities around Copilot Studio. Choose components by the evidence your team needs. · 6 min read
- Cursor Rules for C#: A Project TemplateA ready-to-copy Cursor rules template for C# projects, covering build commands, naming conventions, nullable types, and review steps, with notes on keeping shared decisions in fenbs. · 6 min read
- How to Make an AI Like Jarvis for Practical TasksThe Jarvis idea — a personal AI that handles tasks, knows your preferences, and acts on your behalf — is a useful framing for building a bounded assistant. The practical version uses selected tools, clear approvals, and a context record that survives across sessions. · 5 min read
- How to Make Claude Less Biased in Work DecisionsYou cannot make Claude completely unbiased. You can make its assumptions visible, compare evidence fairly, and keep a person responsible for the decision. · 5 min read
- Knowledge Management Best Practices for Teams Using AIKeep knowledge useful with small habits: capture the reason, name an owner, check sources, retire stale guidance, and verify that people can reuse it. · 5 min read
- MATLAB MCP Server: Connect an AI Assistant to Analysis and CodeHow to connect an AI assistant to MATLAB through an MCP server, run scripts and queries through verified tools, and keep analysis assumptions and review decisions in a shared record. · 5 min read
- Obsidian QuickAdd With AI: Build a Reusable Note WorkflowQuickAdd is an Obsidian plugin for templated capture. Combining it with an AI step lets you build workflows that draft, classify, or process notes consistently. Here is how to design one that is testable and reliable. · 5 min read
- Qwen AI Memory: Keep Project Context Across SessionsSeparate the Qwen model from the app that stores your chats, instructions, and notes. Test the workflow you actually use, then keep a project record another assistant can understand. · 5 min read
- How to Use Claude Skills on the Mobile AppNot all Claude skills are available on mobile. Here's how to check what runs on your phone, start a skill from the toolbar, and keep the underlying prompts accessible when a skill isn't there. · 5 min read
- Uses of AI in Banking: Workflows, Risks, and Human ReviewA practical look at how banks and financial institutions use AI across customer service, risk management, and staff workflows, with a clear view of where human review is required. · 5 min read
- Zettelkasten in Google Keep: Capture, Label, and Connect IdeasBuild a lightweight Zettelkasten in Google Keep using clear titles, a few labels, and a manual index. Move reviewed project context into a record your assistants can reuse. · 5 min read
- AI-Assisted Writing: From a Clear Brief to a Reviewed DraftUse AI to organize notes, draft in small sections, and revise with a purpose. Keep the facts, editorial choices, and final approval in human hands. · 5 min read
- AI Context Window Size Comparison: Choose for Your TaskAI context windows vary from a few thousand to millions of tokens, but advertised size and practical usability are two different things. Here is how to compare them for real work. · 6 min read
- AI Desktop Assistants: Choose One for Your Daily WorkAI desktop assistants range from browser-based chat tools to code editors with built-in AI and system-level assistants. This guide compares the leading options by task, explains how they handle project context, and shows how to keep work portable across them. · 7 min read
- AI for DesignOps: Keep Briefs, Reviews, and Handoffs ConnectedDesignOps teams manage intake, critique, and handoff across many contributors. AI can speed up each stage, but only if the decisions stay connected across sessions. · 5 min read
- AI for Developers Roadmap: Learn Through Small ProjectsThe most reliable way for developers to learn AI-assisted coding is through a sequence of small real projects, each building on the last. This roadmap gives you a practical path from basic code completion to running multi-session agent work, with lessons that carry forward. · 6 min read
- AI Note Taking for Phone Calls: Capture and Check Follow-UpsAI can transcribe and summarize phone calls, but consent, accuracy, and follow-up tracking all need attention. Here is how to set up call notes and use them without creating more work. · 5 min read
- AI Slideshow Summarizers: Keep the Evidence Behind the SlidesAI can summarize a presentation deck in seconds, but slides without speaker notes lose most of their meaning in a text summary. Here is how to get useful output and save the findings that matter. · 4 min read
- How to Summarize a Chat With AI and Keep the Next StepsTurn a work chat into a short, checked record of decisions, unfinished work, and next steps. Keep the evidence close enough that another person or assistant can continue. · 5 min read
- Apple Notes AI: Summarize Notes and Keep Next StepsCheck which AI features your device supports, summarize a selected note, and turn the result into verified decisions and a useful next step. · 5 min read
- Azure Copilot: Plan and Review Cloud Operations WorkUse Azure Copilot to investigate resources and develop a proposed change, then check the evidence, approval and outcome before calling the work complete. · 5 min read
- B2B AI: Practical Uses for Sales, Marketing, and OperationsB2B AI is most useful when it helps a team finish a defined business job. Start with reviewable outputs, measure the whole task, and give assistants a shared record of the rules behind the work. · 5 min read
- Best AI Available to the Public: Choose by Work TaskThe best publicly available AI depends on what you are trying to do. This guide compares the main options by common work tasks so you can choose on fit, not hype. · 5 min read
- Buying a Prompt Library: How to Check Quality Before You PayPrompt libraries vary widely. Some contain tested, specific instructions; most contain generic text that ChatGPT produces for free. Here is what to check before buying, and how to get real value from the prompts that pass. · 5 min read
- ChatGPT AI Models: Choose One for the Work You NeedChatGPT gives you access to several models and modes. This guide explains how they differ and how to choose based on your actual tasks rather than the highest tier available. · 6 min read
- ChatGPT Connectors API: How the Different Connection Routes WorkChatGPT offers three distinct routes for connecting to outside tools and data: interface connectors, MCP connections, and direct API calls. Here is how they differ, which to use for which job, and how fenbs illustrates shared project context through a supported connection. · 5 min read
- ChatGPT Memory Getting Worse? Check the Record FirstCheck stale facts, conflicting instructions, and missing context before you reset memory. A small repeatable test helps identify what is actually wrong. · 5 min read
- ChatGPT vs Claude Hallucinations: How to Compare ReliabilityBoth ChatGPT and Claude produce incorrect information. Here is how to test unsupported claims on your own tasks, compare their reliability fairly and keep verified sources where the next session can find them. · 7 min read
- Claude Code With Qwen: Connection Options and What to VerifyQwen Code and Claude Code are separate tools from different companies. Here is what each does, how to point Claude Code at a Qwen-compatible endpoint, and what to check before trusting the result. · 5 min read
- Claude Projects: What Are You Trying to Achieve?The project goal in Claude Projects says what you are trying to produce. Here is how to write one clearly, with examples, and how to keep it separate from standing instructions. · 5 min read
- Copilot Memory and Personalization: What Carries OverCopilot can tailor replies using remembered details and preferences, but personalization is not a complete work record. Learn what to review and what to save deliberately. · 5 min read
- Copilot Memory Prompt: Save the Context You Need NextA prompt for reviewing a Copilot session and saving the facts, decisions, and next steps worth keeping — rather than the whole conversation — so the next session starts usefully. · 6 min read
- Cursor Rules for Ruby on Rails: Models, Jobs, and TestsPractical Cursor rules for Rails projects: application boundaries, migrations, background jobs, and testing, with fenbs recording the domain decisions and recurring lessons behind those rules. · 4 min read
- Custom GPT Instructions Character Limit: What to KeepCustom GPT instructions have a character limit. Here is how to find the boundary, decide what stays in the instructions, and where to put everything else. · 5 min read
- Explainable AI in Finance: What Reviewers Need to KnowAn explanation should help a reviewer understand a financial model’s output and challenge its use. Learn what explanations establish, what they leave unanswered, and which human decisions need a lasting record. · 5 min read
- Export Your Entire ChatGPT HistoryOpenAI lets you request a full export of your ChatGPT conversation history from your account settings. This guide explains how to request it, what the archive contains, and how to find what you need inside it. · 5 min read
- Gemini for Teachers: Planning, Feedback, and ReviewGemini can help teachers prepare lessons, generate examples and draft feedback, but checking accuracy and staying within school rules are steps that belong to the teacher, not the model. · 7 min read
- GitHub Copilot Commands: Choose the Right Action for the TaskGitHub Copilot offers chat commands, slash commands, and agent actions that vary by editor and interface. Here is how to find what is available to you, when to use each type, and how to build a library of reusable work prompts. · 4 min read
- How to Give ChatGPT Access to Google DriveChatGPT can connect to Google Drive to browse and read your files in plans with the Drive connector. Here is how to set it up, what it can access, and how to keep document context beyond a single chat. · 5 min read
- How to Give Claude Access to LinkedIn: Options and LimitsClaude cannot log into LinkedIn for you, but you can share profiles, posts, and research notes in several supported ways. Here is how to use Claude for LinkedIn work. · 5 min read
- Multilingual GPT: Draft Across Languages With Clear ContextUse GPT for multilingual drafting and translation with a defined audience, approved terminology, and checks for meaning. Keep language decisions consistent when you change assistants. · 5 min read
- NotebookLM for Books: Read, Ask, and Check Your NotesUse a book you can access as a research source, ask focused questions, and check each useful answer against the text. Keep the reading plan and references ready for your next session. · 5 min read
- Prompt Engineering vs Harness Engineering: What Each ChangesPrompt engineering changes what you say to a model. Harness engineering changes the tools, execution loop, checks, and context the model works inside. Here is what each controls and when you need both. · 7 min read