This may be of interest to you, a few years before Kant with “Syādvāda” going beyond the binary implied by contradiction alone: https://en.wikipedia.org/wiki/Anekantavada
Yes! We used our friends at Reducto (https://reducto.ai/) for all document extraction and parsing (one of the best companies I've ever referred to YC ;) )
We did an initial parsing pass of all four DOJ document batches on Friday. This takes a raw PDF and returns chunks containing typed blocks—each with a type (Title, Text, Figure, etc.), bounding boxes, content, and confidence scores. For PDFs that were just scans of photographs (which was like 90% of new content in Friday's release), it gave in depth descriptions of those! You can type search terms like "door" at https://www.jmail.world/photos to see what I mean.
For apps like Jmail and JFlights we use their structured extraction endpoint instead—you define a schema (e.g. {from, to, subject, date, body} for emails or {departure_airport, arrival_airport, passengers[], date} for flights) and it pulls those fields directly into JSON.
The JFlights example served as the best ad for Reducto and how doc parsing technology can speed up hours of journalistic investigations like this.
Personally I am waiting for the day I can realistically buy a second hand three year old datacentre GPU so I can run Kimi K2 in my shed. Given enough time, not a pipe dream. But 10 years at least.
For graph/tree document representations, it’s common in RAG to use summaries and aggregation. For example, the search yields a match on a chunk, but you want to include context from adjacent chunks — either laterally, in the same document section, or vertically, going up a level to include the title and summary of the parent node.
How you integrate and aggregate the surrounding context is up to you. Different RAG systems handle it differently, each with its own trade offs. The point is that the system is static and hardcoded.
The agentic approach is: instead of trying to synthesize and rank/re-rank your search results into a single deliverable, why not leave that to the LLM, which can dynamically traverse your data. For a document tree, I would try exposing the tree structure to the LLM. Return the result with pointers to relevant neighbor nodes, each with a short description. Then the LLM can decide, based on what it finds, to run a new search or explore local nodes.
My experience here in the UK, despite getting the highest "tier" of survey carried out on my (current) home when buying it, was that within the 74 page report they produced, there were at least a dozen occurrences of the surveyors recommending a "specialist".
They avoid any liability by saying, "we couldn't survey under the floor", we recommend getting in a specialist. "we can't assess the roof structure", we recommend getting a specialist.
By the time all was said and done, we were looking at tens of thousands of pounds in further "specialist" surveys, which nobody realistically is going to do only to decide after that you won't buy the house.
I can imagine once you're looking at houses priced in the millions it might make sense, but blowing the equivalent of your deposit just isn't tenable.
My home inspection report was two pages of useful information (here’s where the water shutoff is, the breaker panel is here), a page with two actual real issues (garage door opener didn’t work, kitchen foundation was settling) and then ten to twenty pages of “we don’t look at shit” legalese.
They specifically disclaim being experts in damn near everything.
If you want a real inspection you hire two or three building contractors to do it. I’d go with a general, a roofer, and an electrician. If I cared.
Without more information I'm very skeptical that you had e.g. Claude Code create a whole app (so more than a simple script) with 20 cents. Unless it was able to one-shot it, but at that point you don't need an agent anyway.
The key element for a bona fide sale at common law is the buyer’s absence of knowledge of the defective title of the seller.
Not sure how US courts have interpreted this requirement but that’s the onus and I believe it rests on the third party buyer (to show absence of knowledge through evidence).
The icc warrant claims it is an international armed conflict.
This is important, because palestine did not ratify the amendment to the rome statue criminalizing starvation in non-international armed conflict, so that charge goes away if it is just an internal thing as opposed to an international war.
reply