Where did I read that? Every mention, in one place
DORA turns up in four things you read across three months. You remember two. Pith recognises the named things in everything you save — regulations, institutions, companies, people — and gathers every mention in one place.
In short
Pith extracts the named things from every source you save — regulations like DORA, institutions like BaFin, companies and people — and links them inside the wiki prose. One click shows every mention across articles, wiki pages, briefings and clients, together with the relationships between them.
The team behind Pith Lab
Over three months you read four things that mention DORA. One was a regulatory analysis, one a vendor case study, one a blog post about outsourcing chains, one a note from a client meeting. When the question comes — and it comes — you remember two of them. Maybe.
That is not carelessness. It is how the work goes: you read broadly because you advise broadly, and a head does not index by keyword.
A name is not a word
A full-text search for "DORA" finds the places where those four letters appear. It does not find the article that says Digital Operational Resilience Act throughout. It does not know that BaFin published guidance on it, because that was in a different source. And it cannot tell you which client engagement already had this on the table.
So Pith treats named things as objects rather than strings. As you save, it recognises them: regulations, institutions, companies, people, methods. Each gets its own page, and that page records where the thing has appeared in your reading.
Inside the prose, not in a menu
The connection lives where you read. When a wiki page says "the Sarbanes–Oxley Act", the first mention is a link. One click, and you see:
- what else it goes by — SOX and the spelled-out title are one thing
- how it relates to other named things you have read about
- every mention: which article, which wiki page, which briefing, which client
From there each entry leads back to where it came from. The article that said it. The wiki page that cites it. The engagement where it came up.
A by-product, not a second job
None of this is upkeep. Pith reads your saved sources anyway, to summarise them and write the cited wiki — the named things fall out of that same pass. You keep saving what you read. The mesh forms beside it.
Same stance as the wiki itself: if organising costs effort, you will not do it. So it must not cost any.
Where the limit is
Recognition comes from a language model, and it is sometimes wrong. It will split one thing into two records when one source writes "Sarbanes–Oxley" and the next writes "Sarbanes–Oxley Act". That is why every entity page carries a confidence score, and why you can merge two records into one.
We write that down rather than leaving it out. A knowledge base that acts infallible is exactly the one you cannot lean on in front of a client.
FAQ
Do I have to maintain the entities myself?
No. They appear as you save. Pith reads the source anyway to summarise it, and the named things fall out of that reading. The one thing you can do, if you want to: merge two records when the same thing was recognised under two spellings.
How is this different from full-text search?
Searching for 'DORA' finds the places where that word appears. An entity knows that 'DORA' and 'Digital Operational Resilience Act' are the same thing, records how it relates to others, and shows its mentions across articles, wiki pages and clients — not only in prose.
Can the recognition get it wrong?
Yes. It comes from a language model and carries a confidence score, shown on the entity's page. When it recognises the same thing twice — with and without 'Act', say — you merge the two. That is why the feature exists.