The alpha, twice a week
That was 2 min.
Get the best new moments in your inbox every Wednesday and Saturday. It's free.
Check your inbox
We sent a link to your address. Click it to finish signing up. Not there in a minute? Look in spam.
or
Latent Space · 2 min · Nov 2, 2026
Long contracts can lose meaning when split into chunks
Jacob Lauritzen
Chunked searches can miss meaning that only appears across multiple pages, while a huge context window is not a universal fix.
In this episode
Full episode- 8:39Legora tests agents in simulated legal workspacesPro pick 3 min
- 26:13Client confidentiality forces Legora to infer product failures3 min
- 34:25Tabular Review lets lawyers test a workflow on ten rows firstPro pick 4 min
- 14:12Law firms need incentives to turn expertise into durable assets3 min
- 42:33Legal engineers configure workflows while FDEs write code4 min
- 47:22Legora delayed APIs to avoid locking in a fast-changing product3 min
- 50:12Legora grew beyond its plan to stay at 20 engineers3 min
- 58:13Long contracts can lose meaning when split into chunks2 min
“and then each individual chunk is not necessarily relevant to the search term or what they're trying to figure out but together it's extremely relevant”
Jacob Lauritzenat 59:59 · Latent Space
Takeaways
- A relevant point may only emerge across multiple pages.
- Summarizing chunks can lose important nuance.
- Long-document workflows need cross-section reasoning.
About this segment
Chunked searches can miss meaning that only appears across multiple pages, while a huge context window is not a universal fix. Long-contract workflows still need reasoning that preserves cross-section connections.
Jacob Lauritzen is CTO of Legora.
Leads engineering at Legora, building legal AI agents for research, contract review and drafting.
Shawn Wang (swyx), Host · Co-founder, AI Engineer
Wrote “The Rise of the AI Engineer” and founded Smol AI, maker of AI News.
- BC
Basil Chatha, Host · Forward Deployed
From “Building Legal Agents When You Can't See the Data — Jacob Lauritzen, CTO of Legora”, Nov 2, 2026. Watch the full episode
Thanks. We'll look at and fix what's off.
Transcript
Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.
- 58:18
Okay, interesting. Just like if you receive something from a junior lawyer, it's the same thing you do, right? Just like I should double check this and like this probably is fine. Any other elements of let's call it legal reasoning or what's the legal frontier that models are not good enough yet that you want from the next generation of models? Nothing I can think of specifically right now. I think long document reasoning is still difficult.
- 58:41
Long, like million token context. Yeah, exactly. You have a huge contract. And you're making edits across it. It's a difficult thing to do for a model today. And I remember that you changed all these things up here and the definitions changed and the ordering changed and whatever else. So the simplest would be one shot. You feed everything in and then one shot output. You don't do that. you break it up into chunks and then you orchestrate edits. Yeah, exactly.
- 59:05
And this is one of those things that are fun to be part of in this industry, right? So our word at it was like, we've done MapReduce and we've changed and there was one big one and there was one big one in multiple sizes. We keep changing the scaffolding that we build around as models change. I would say I do a lot of news reporting obviously because of what I do and I've recently shifted from MapReduce to throw everything into context.
- 59:30
and it's better but it's also different we had this problem with search where it's like you're searching for certain things in a huge contract and you can map reduce it but then it might, if you're summarizing any sort of thing, then it might lose some important nuance in between the chunks essentially if you just summarize the individual parts up and so this has been a problem where it's like we can do two types of search, we can also do the other one where it's like you need to have reasoning in the different steps there in between, so I'm trying to think of a really good example here there's some documents where they talk about something at length over like let's say two pages or whatever and then each individual chunk is not necessarily relevant to the search term or what they're trying to figure out but together
- 1:00:10
it's extremely relevant and so yeah these are some of the problems where it's hard like you said earlier to actually make the models better at that because you can't even look at the contracts right