~/blog · filtered by category

knowledge.

9 entries filed under this signal. ← back to all entries

[blog/category/knowledge] · reverse chronological$ ls -la --filter=knowledge

2026-07-20knowledgeTribal knowledge is a single point of failureThe one person who knows why that module exists is a risk you priced at zero. Now half your new hires are agents who never met them.#paelladoc#knowledge-graph#tribal-knowledge#onboarding#ai-coding8 min./read →2026-07-20knowledgeReading a repo into product artifacts: stories, criteria, decisionsEveryone teaches the forward path: spec first, then build. For software that already exists you need the reverse, and almost nobody explains it.#paelladoc#knowledge-graph#reverse-intake#acceptance-criteria#ai-coding9 min./read →2026-07-20knowledgeProduct memory: what your team forgets and your system shouldn'tTeams don't forget the code. They forget the decisions, the reasons, and the dead ends. When nothing remembers those, you keep re-litigating settled questions and agents keep re-walking paths you already abandoned.#paelladoc#product-memory#decision-logs#knowledge-graph#ai-coding9 min./read →2026-07-20knowledgeOnboard against the graph, not tribal knowledgeA new hire ramps by asking the person next to them. So does a new agent, except it can't. Both should onboard against a graph of the system, not a map trapped in someone's head.#paelladoc#onboarding#knowledge-graph#tribal-knowledge#agents8 min./read →2026-07-20knowledgeLiving documentation: docs that stop rottingDocs rot because they're a second copy of the truth that nobody keeps in sync. The only version that stays current is the one derived from the system, not written beside it.#paelladoc#living-documentation#docs-as-code#knowledge-graph#ai-coding8 min./read →2026-07-20knowledgeKnowledge graph vs vector search for code: when each one winsEmbedding your whole codebase and hoping the right chunks come back is the default. It works, until the question is about relationships instead of resemblance.#paelladoc#knowledge-graph#rag#retrieval#ai-coding9 min./read →2026-07-20knowledgeDocs-as-code in the agent eraDocs-as-code was already good practice. Once agents read your docs to decide what to build, the docs pipeline stops being a nice-to-have and becomes part of the build.#paelladoc#knowledge-graph#docs-as-code#living-documentation#ai-coding9 min./read →2026-07-20knowledgeCodebase maps, not bigger promptsWhen an agent keeps getting it wrong, the reflex is to write a longer prompt. The agent doesn't need more words about the codebase. It needs a map of it.#paelladoc#codebase-maps#ai-coding#agents#context-engineering8 min./read →2026-07-20knowledgeCode knowledge graphs: what they are and why agents need themAn agent can read any file in your repo on demand. What it can't do is see how the files relate. That map is a code knowledge graph, and it's the difference between a change that fits and one that breaks three files away.#paelladoc#knowledge-graph#ai-coding#codebase-maps#agents11 min./read →

— end of log · 9 entries in knowledge —