Journal

No master prompt can replace a solid data foundation

Why the quality of an AI application is often decided before the first prompt is written.

Data foundationData qualityData governance

AI begins where code meets data

Why the value of an AI application comes from data, software, rules and control working together — not from the model alone.

System architectureData contractAgents

3/40 · ARTICLEFREE

From documents to usable project knowledge

Storage keeps files. Usable project knowledge preserves meaning, relations, validity and decisions.

Project knowledgeRAGRetrieval

4/40 · ARTICLEFREE

Data sovereignty is more than local hosting

A server location tells you where processing occurs. Data sovereignty tells you who can determine, inspect, restrict, transfer and terminate that processing.

Data sovereigntyData governanceZero Trust

5/40 · ARTICLEFREE

Data quality is always purpose-dependent

Data is not abstractly good or bad. Its quality depends on whether it can support a concrete task at an acceptable level of risk.

Data qualityEvaluationBias

6/40 · ARTICLEFREE

Reliable AI requires provenance

An answer does not become reliable because it sounds precise. It becomes reliable when its path from valid source to output remains auditable.

ProvenanceData lineageAudit

7/40 · ARTICLEMEMBERS

Metadata makes knowledge machine-readable

People infer meaning from context. Machines need meaning expressed as explicit, consistent and testable structure.

MetadataSchemaVocabulary

8/40 · ARTICLEMEMBERS

Data needs a lifecycle

Data does not become dangerous merely because it is old. It becomes dangerous when nobody knows why it exists, whether it still applies, or when its purpose ends.

LifecycleRetentionDeletion

9/40 · ARTICLEMEMBERS

RAG is not a truth machine

A language model does not become truthful merely because documents are placed in its context. It only gains a better opportunity to ground its answer in inspectable evidence.

RAGRetrievalEvaluation

10/40 · ARTICLEMEMBERS

Human-in-the-loop needs decision boundaries

"A person checks it at the end" is not a control concept. Oversight becomes meaningful only when a person has enough information, time and authority to alter or stop a machine-supported outcome.

Human oversightAutomation biasGovernance

11/40 · ARTICLEMEMBERS

A skill is not a long prompt

A prompt describes one request. A skill describes how a class of requests is handled repeatedly, verifiably and safely under defined conditions.

SkillAgentsEvaluation

12/40 · ARTICLEMEMBERS

An agent needs a workspace, not just a persona

A persona shapes the voice. A workspace shapes behaviour — and makes visible what an agent knows, may do, changes and hands over.

AgentsAgent harnessHandoff

13/40 · ARTICLEMEMBERS

More agents do not automatically make a team

A team is created not by adding voices, but by combining complementary capabilities, clear ownership and verifiable integration.

AgentsOrchestrationEvaluation

14/40 · ARTICLEMEMBERS

What an agent did must remain reconstructable

An agent is not controllable merely because it writes a convincing final report. Control requires its actual path through sources, decisions, tools and state changes to remain verifiable.

AgentsObservabilityAudit

15/40 · ARTICLEMEMBERS

AI risk lies in the use, not the model

The same model can draft an internal text, rank job applicants or control a safety-relevant process. Technical similarity does not make these uses equally risky in law or operations.

RiskAI ActGovernance

16/40 · ARTICLEMEMBERS

Good AI consulting starts with a defensible system assessment

Consultants who present solutions too early often advise an imagined organisation. A defensible assessment first reveals how work, data, systems, decisions and responsibility actually connect.

ConsultingGovernancePlanning

17/40 · ARTICLEMEMBERS

AI literacy is not a one-off training session

An attendance record proves that someone was present. It does not prove that they can detect a false AI output, protect sensitive data, override a decision effectively or stop at the right moment.

AI literacyAI ActGovernance

18/40 · ARTICLEMEMBERS

An AI notice alone does not create transparency

"This content was created with AI" is a signal. Transparency begins when people understand what that signal means in the situation and what they can do next.

TransparencyAI ActProvenance

19/40 · ARTICLEMEMBERS

A regulatory sandbox is not a seal of approval

A sandbox is a supervised learning environment. Treating it as a certification machine confuses the process of discovery with the outcome of conformity assessment.

AI ActGovernanceRisk

20/40 · ARTICLEMEMBERS

A pilot deployment is not yet controlled real-world testing

Once real people, decisions or consequences are involved, the word "pilot" is no longer a safety architecture.

AI ActGovernanceRisk

21/40 · ARTICLEMEMBERS

Evidence work begins after go-live

A production AI system is not a completed project. It is a continuing claim that purpose, performance and controls still hold under changing conditions.

AI ActGovernanceObservability

22/40 · ARTICLEMEMBERS

A CE marking is not a promise of AI quality

The CE marking says that conformity is declared for a specific system against the applicable requirements. It does not say that the system is the best, error-free or endorsed by an authority.

AI ActGovernanceEvaluation

23/40 · ARTICLEMEMBERS

RAG is not a chatbot with attached PDFs

Uploading files creates access. A RAG system creates a controlled path from a question through selected evidence to a verifiable answer.

RAGRetrievalEvaluation

24/40 · ARTICLEMEMBERS

Chunking, parsing and the lost connections

A chunk is not merely a piece of text. It is a unit of meaning with provenance, neighbourhood and a role in the document.

RAGRetrievalProvenance

25/40 · ARTICLEMEMBERS

A RAG system needs more than a vector database

A vector database can find similar passages. It does not know which source is authoritative, which version applies or which result actually answers the question.

RAGRetrievalObservability

26/40 · ARTICLEMEMBERS

Do not find the needle — find its most probable location

Good search need not hit one perfect sentence immediately. It can first identify the region of knowledge where a defensible answer is likely to exist.

RAGRetrievalProvenance

27/40 · ARTICLEMEMBERS

A gap in the graph is not yet knowledge

A strong idea may exist between two dense topics. The same space may also contain a data error, poor segmentation or simply no defensible relationship.

RAGRetrievalProvenance

28/40 · ARTICLEMEMBERS

GraphRAG must lead every answer back to the source

An answer is the end of a processing chain. Evidence exists only when that chain can be inspected in reverse all the way to a valid source.

RAGRetrievalProvenance

29/40 · ARTICLEMEMBERS

When AI does not just read but changes files

An answer can be wrong. A file change can also alter the next process run, other people's work and the authoritative knowledge state. Write access therefore begins a new risk class.

Prompt injectionRiskGovernance

30/40 · ARTICLEMEMBERS

Skills as executable knowledge

Documented knowledge explains what to do. Executable knowledge also defines when, with which tools, inside which limits and with what quality evidence it may be done.

GovernanceEvaluationPrompt injection

31/40 · ARTICLEMEMBERS

Separate personas, skills and project data cleanly

A persona controls how a system communicates. A skill controls how it performs a capability. Policy controls what it may do. Project data defines what it works on. Mixing these layers creates opaque bundles of power.

GovernanceRiskProvenance

32/40 · ARTICLEMEMBERS

Automation increases the need for control

Automation removes work steps, not accountability. The faster a system can act, the farther it can reach and the more often it can repeat decisions, the more precisely its boundaries, signals and intervention points must be designed.

GovernanceRiskObservability

33/40 · ARTICLEMEMBERS

Worktrees, checkpoints and reports as data safeguards

An isolated workspace prevents collisions. A checkpoint makes a state addressable. A report explains the path. A backup survives failure. Recoverability needs all four.

ProvenanceGovernanceObservability

34/40 · ARTICLEMEMBERS

AI may prepare – humans must decide

Human oversight is not a confirmation window. It is a decision architecture in which a capable person understands in time, can object and accepts responsibility for a precisely bounded effect.

AI ActHuman oversightAutomation bias

35/40 · ARTICLEMEMBERS

Who owns the data spaces of AI?

A data space rarely belongs to one party. It becomes governable when every data product has explicit decision rights, access conditions, permitted purposes, evidence and a tested route for every participant to leave.

Data sovereigntyGovernanceProvenance

36/40 · ARTICLEMEMBERS

Research and implementation need different AI workspaces

A system that may read everything and change everything at the same time is convenient. That is precisely why it is difficult to govern.

OrchestrationGovernanceHuman oversight

37/40 · ARTICLEMEMBERS

An AI system needs an executable rulebook

A rule that merely sounds good does not govern a system. It must be discoverable, applicable and testable.

GovernanceEvaluationOrchestration

38/40 · ARTICLEMEMBERS

Data sovereignty is more than local hosting

Data can remain inside your own building and still belong to a logic that nobody there controls.

Data sovereigntyGovernanceProvenance

39/40 · ESSAYFREE

The order before the answer

Data as the operating system of artificial intelligence. A long-form essay: the quality of AI is not created first in the answer, but in the order that makes an answer possible, reviewable, and accountable.

Data foundationProvenanceRetrieval

40/40 · AGENTIC PACKAGEFREE

The Agentic Operations Package

A portable operating system for traceable AI data work — canonical core, Obsidian vault, auditable workflows, skills, harness and adapters.

HarnessObsidian VaultGovernance