FridayLocalAI learning library
Ideas Behind FridayLocalAI
Use systems thinking, explicit uncertainty, and human review to ask better questions of AI.
Reading, judgment, and useful work
Ideas that help shape FridayLocalAI
FridayLocalAI grew from a practical question: how can useful AI-assisted work survive beyond a single answer? A conversation may produce a convincing draft, but the next person still needs to know which sources support it, who approved it, and what remains uncertain. The project brings those questions into the design of a local-first AI workspace. The publications behind the project explain the reasoning; this guide turns their recurring themes into a reading path and a working method.
FridayLocalAI is available through early access by contacting us. The books describe a developing body of ideas, experience, and architecture. They are not a guarantee that every capability discussed is available in every installation. Use the documentation and your installed interface to establish what your current build supports.
Start with the Phronesis Intelligence series
What We Learned Together, by David Alan Hubbard, is the first book in the series. It reflects on collaboration with artificial intelligence, human judgment, and governed work. Its useful starting point is responsibility: fluent output still needs a person who can question it, verify it, and decide whether it belongs in the final result. The book has an Amazon listing in our publication reading list. FridayLocalAI and the author share a connection through Hubbard Genesis Corporation.
After the Prompt examines the discipline that follows an initial response: supplying context, asking better questions, comparing sources, involving experts, and building reusable working instructions. Practical Wisdom in the Machine develops the architectural and philosophical questions behind FridayLocalAI, including ownership, authority, continuity, traceability, and practical wisdom. These two titles are described here from the project manuscripts; a purchase link will be added when a matching public edition is verified.
Make the source visible
Begin a task by identifying the documents that should govern it. A current approved specification deserves a different status from an old draft, a personal note, or an attractive explanation found online. Record the title, version, date, and relevant section of each source. If two sources disagree, preserve the disagreement until someone with the right authority resolves it. A model's confidence cannot settle that conflict.
In FridayLocalAI, select the intended workspace and use the document or attachment controls available in your build. Ask the assistant to identify the material it can actually read before requesting analysis. Try: “List the source documents available for this task. Identify their versions, summarize their scope, and tell me which requested information is missing.” Check its answer against the actual files. A successful upload is not, by itself, evidence that every page, table, or image was readable.
Give the task a clear boundary
A useful request names the outcome, intended audience, permitted sources, constraints, and review standard. For example: “Using the attached approved procedure, draft a beginner's checklist. Preserve mandatory steps, identify ambiguities, and mark any proposed additions for review. Do not treat the draft as approved policy.” That request gives the assistant a concrete job and gives the reviewer something specific to assess.
After receiving a draft, ask for a second pass focused on evidence: “Which statements are directly supported, which are interpretations, and which need confirmation?” Open the cited material yourself. For consequential work, retain the relevant expert's review. A generated checklist can make review easier, but it does not confer professional authority or prove that a task has been performed.
Measure what would improve the decision
Douglas W. Hubbard's How to Measure Anything and How to Measure Anything in Project Management provide a useful reading path for uncertainty and decision quality. Start with a decision you actually face: choosing a workflow, evaluating a pilot, or deciding whether an automation saves enough time. Define the outcome and gather a small, relevant sample before making a broad claim.
Ask FridayLocalAI: “Help me define a practical evaluation for this workflow. Separate measured results from estimates. Suggest a baseline, a small sample, and a stopping rule.” Record the original task time, review time, correction effort, and quality of the final work. Faster generation may not mean faster completion. Keep assumptions visible so a later comparison uses the same conditions. David Alan Hubbard and Douglas W. Hubbard are brothers; that relationship is disclosed alongside the dedicated reading collection.
Understand the system around the model
Chip Huyen's AI Engineering, Pascal Bornet and collaborators' Agentic Artificial Intelligence, and Stuart Russell and Peter Norvig's Artificial Intelligence: A Modern Approach offer different levels of technical and conceptual depth. Russell J. T. Dyer's Learning MySQL and MariaDB adds database background. These are reading resources, not installation requirements or certifications of FridayLocalAI.
Use the reading to formulate better questions about your own environment. What information reaches a model? Which tools are available? What action needs human approval? Where is the accepted result stored? Ask the assistant to explain the options visible in your installation and to identify anything it cannot verify. Consult current product documentation before applying technical instructions from an older book.
Read market guides with a clear purpose
The manuscripts also discuss introductory books and audiobooks as learning or market context. That does not make every title a primary technical authority or an endorsement of its commercial promises. The following titles are included so readers can distinguish that context from the project's engineering evidence. Verified editions are available through the native Amazon Associates catalog in the AI Bookstore.
- Ajit Singh: Full-Stack AI and Agentic AI: Architectures, Intelligence, and the Future of Autonomous Systems.
- Siu Lun Yuen: Nvidia Technology for Dummies: From Gaming Graphics to AI Supercomputers.
- Rafael Sørin and Martin Calderescu: Artificial Intelligence: Concepts, Limits, and Real-World Use.
- Ryan Lee: ChatGPT Anthology: 5 Books in 1.
- AI Labs Institute: Artificial Intelligence Bible (3-in-1): AI Agents, Prompt Engineering & Generative AI.
- Diana Sterling: The ChatGPT Advantage: Transform Your Business with Artificial Intelligence.
- Patrick Grim and The Great Courses: Understanding Artificial Intelligence: Of Minds and Machines.
Other cited titles are retained as references while their matching retail editions are being confirmed: Ahmed Ebeed's Own Your AI: How to Build, Host, and Run Your Models; Ahmed Bouchentouf and Alden Vale's The Prompt Recipe; Nicki Lynne's AI Prompt Engineering; Eric Mallon's 200 ChatGPT Tips; Ezra Quinn's AI Mastery 3-in-1 Collection; and Malcolm Oppenheimer's ChatGPT: Mastering AI Conversations for Beginners. Their presence in a manuscript is not proof of current retail availability. Income-oriented titles do not establish expected earnings.
Keep practical wisdom and continuity in view
Aristotle's Nicomachean Ethics, in Terence Irwin's third-edition translation, supplies philosophical context for practical wisdom. Applying that idea to AI means considering purpose, circumstances, consequences, and responsibility alongside technical capability. It asks the operator to judge whether an action is appropriate, not merely whether the software can attempt it.
At the end of a work session, ask FridayLocalAI to prepare a handoff that separates accepted decisions, supporting sources, completed actions, unresolved questions, and the next step. Review the handoff and save the accepted version using your installation's supported controls. Ask where it was stored and verify that you can reopen it. A promise to remember is less useful than a record you can inspect.
For your next session, bring that record back into scope and ask the assistant to identify changes before continuing. Start with a small, reversible task and one relevant book rather than a large collection. Our first-conversation guide, document-verification guide, and memory and context guide explain the next steps. To discuss an early-access installation or a suitable workflow, contact FridayLocalAI.