Continue work
Stop rebuilding the project story.
A new task can read the current state, identify the unresolved decision, and continue the next action.
AI knowledge base / field guide 001
Stop starting over with AI.
Give AI a portable workspace for your projects, decisions, instructions, trusted sources, and current priorities, so a new conversation can understand the work and continue from the current state.
Pay what you want. $0 minimum, $20 suggested.
- Plain files you own
- No coding required
- Not model training
The problem
Your AI is not starting from your work. It is starting from the chat.
The useful context is usually scattered across messages, documents, tools, and memory. When the source of truth is unclear, AI can repeat settled questions, use stale assumptions, or produce convincing work from the wrong premise.
01Project in your headGoals, history, constraints
→
02Scattered contextChats, notes, documents
→
03No source authorityOld and current blur together
→
04Repeated explanationSearch, correction, rework
The continuity layer
A knowledge base is external context your AI can read.
It is an organized set of ordinary files that records what matters, what is current, what has been decided, which sources are authoritative, and what should happen next.
The five files are the smallest useful continuity layer. The starter can expand into projects, reusable knowledge, raw source material, privacy rules, and finished outputs as the work grows.
01START-HERE.mdHow a new task should orient itself
02CONTEXT.mdThe durable background it needs
03INSTRUCTIONS.mdHow to work with you and the project
04PROJECT-STATE.mdCurrent state, unresolved work, next action
05SOURCES.mdWhere facts came from and what to trust
my-ai-workspace/
Fig. 02 The folder is not the proof. Continuity is.
InputRaw sourcesCalls, files, notes, links
→
StructureSystem + project stateContext, decisions, authority
→
WorkAI taskFind, reason, continue
→
GateHuman reviewCorrect, decide, update
→
OutputUseful artifactBrief, plan, lesson, handoff
The real walkthrough
Watch a blank folder become a working knowledge base.
A rough voice answer about a Hong Kong trip and a future venture studio becomes linked projects, explicit gaps, source trails, privacy rules, a current-state file, and a recommended first research brief.
- 01DownloadChoose the version that matches your AI tool.
- 02OpenGive Codex or another AI access to the workspace.
- 03AnswerUse normal language. Rough voice notes are enough.
- 04ContinueTest the base in a fresh task, not the setup chat.
Read this workspace and tell me where the project stands.
Recovered from the base
Mission: plan the Hong Kong trip and use it to build meaningful venture-studio relationships.
Open gaps: travel dates, venture-studio thesis, and the value you can offer potential partners.
Best next action: confirm the dates, then build the ecosystem and meeting brief from traceable sources.
What becomes possible
The same context can support the next decision, the next brief, and the next output.
Research with authority
Keep facts, assumptions, and questions distinct.
Build briefs that show what is verified, what is inferred, which gaps remain, and where each claim came from.
Preserve decisions
Keep the reason, not just the result.
Record the decision, rationale, owner, date, and conditions for revisiting it so future work does not silently reverse it.
Compound outputs
Turn one body of work into many useful forms.
Use the same grounded context for a client brief, meeting prep, research report, launch plan, article, lesson, or handoff.
Persona field guide
Useful wherever the quality of the work depends on accumulated context.
Founder or solo operator
Decisions, customer evidence, experiments, and priorities split across tools.
What the base holds
Company context, decision log, customer sources, experiment state, and current priorities.
Useful result
Prepare the next experiment or decision brief without rebuilding the company story.
Consultant or agency lead
Repeated onboarding, scattered feedback, and unclear source authority.
What the base holds
One project space per client with goals, constraints, decisions, source material, and delivery state.
Useful result
Faster meeting prep and fewer revisions caused by missing context.
Researcher, analyst, or student
Sources and interpretations blur together, so reading does not compound.
What the base holds
Research questions, evidence notes, citations, competing interpretations, and unresolved gaps.
Useful result
Source-traceable literature maps, briefing documents, and research next steps.
Creator, educator, or writer
Ideas, audience questions, drafts, examples, and published work live apart.
What the base holds
Voice rules, content pillars, source library, draft state, audience signals, and reusable examples.
Useful result
Turn one real workflow into a video, article, lesson, and lead magnet without starting from zero.
Job seeker or career builder
Achievements, role research, resume variants, and application decisions become fragmented.
What the base holds
Verified career evidence, target-role research, application state, resumes, and interview stories.
Useful result
More specific applications and interview prep grounded in real proof.
Calculate your context tax
How much time do you spend getting AI back up to speed?
This is an estimate, not a savings claim. It makes the repeated reconstruction visible so you can decide whether the problem is worth fixing.
TimeFewer reintroductions, repeated searches, and handoff explanations.
MoneyLess paid time lost to reconstruction and avoidable rework.
GrowthMore experiments, follow-ups, and reusable outputs from the same work.
What it is not
Useful because it is inspectable, not magical.
- It does not train a model on everything about you.
- It does not create perfect or automatic memory.
- It does not guarantee that an AI output is correct.
- It does not remove source checking or human judgment.
- It does not need a proprietary database to be useful.
The files make context editable, portable, and easier to verify. That is the advantage.
Choose the right path
Build one yourself, or find the highest-value place to use AI first.
DIY starter
Pay what you want$0 minimum / $20 suggested
Stop Starting Over: AI Knowledge Base Starter
Use this when you already know the project you want AI to understand.
- File-aware version for Codex, Claude Code, and similar tools
- Chat-only version for browser-based AI tools
- Voice-friendly guided onboarding
- Plain Markdown files you own
Guided diagnosis
$497AI Assessment
Find the three places where AI could create the most useful leverage.
Choose this when you need help deciding where AI is worth using before you build anything.
- 60-minute recorded working session
- Work and slow-point map
- Three credible AI opportunities
- Practical 30-day action plan and review call
If the Assessment cannot identify three credible opportunities and a clear first path, the fee is refunded. Continue into an AI Workspace Foundation within 14 days and the full fee is credited toward that work.
Questions before you start
The practical details.
Do I need to know how to code?
No. The file-aware version uses ordinary folders and Markdown. The chat-only version is designed for browser-based AI tools.
Which tools does it work with?
The download includes a file-aware version for tools such as Codex and Claude Code, plus a separate version for ChatGPT, Claude in the browser, Gemini, and similar tools.
Is this the same as training an AI on my data?
No. It is external context the AI reads when you provide it or connect the relevant workspace.
Will the AI remember everything automatically?
No. The base has to be maintained, and the AI still needs access to the relevant files or pasted context.
What should I build first?
Start with one real, ongoing project where repeated explanation or scattered sources already cause friction.
Can I use it for sensitive information?
Only after deciding what the AI tool is allowed to access. The starter includes privacy boundaries, but you remain responsible for the data you provide to a model or service.
The fresh-task test
Give one real project a place to continue.
Download the starter, answer the onboarding questions, then open a fresh AI conversation. The proof is whether it understands the project, finds the right sources, identifies what remains unresolved, and completes a useful next action.