Very slick and potentially very powerful. After a few minutes playing with it, I have a few recommendations:
- Variables should have more types, like an array of objects
- Prompting should incorporate Jinja2/Nunjucks
- For every prompt, I should be able to create many different test examples, along with an answer key, and measure how well it does across many tests
- It should auto-save. I did a lot of prompting work and then clicked another icon. When I came back, all my work was gone. (In fact, I don't see where to save at all! Maybe I'm just missing it.)
Casetext | Software Engineer, Research Scientist, Machine Learning Engineers, and More | San Francisco, CA | ONSITE
This is an insanely fun time to join us. If you are interested in search, NLP, recommender engines, cutting-edge front-end technologies, or making a difference through your work, you should check us out.
Our startup, Casetext (YCS13) was mentioned a few times here, so I thought I'd stop in.
The crux of the article is that most legal research solutions have ignored the immense power contained in the links between laws:
> Laws frequently reference other laws in order to reuse definitions, introduce exceptions, or make it clear that two concepts are meant to work together. Consequential laws tend to get referenced in other laws as their influence spreads throughout the legal system. Experienced lawyers build up detailed mental maps of these links, allowing them to jump immediately to core issues of complex legal problems.
> However, most laws can only be searched using the dark-age, Lycos strategy—guess at keywords and hope—and it’s often necessary to pay for even that limited functionality.
We at Casetext are taking a very different approach than the "dark-age, Lycos strategy" that you have to pay for:
1. On Casetext, the law is free, as is basic search. Honestly, it's insane that Westlaw and LexisNexis charge as much as they do for basic keyword search over a database that should have been free to begin with.
2. We make money by charging for advanced, data-driven ways that lawyers can research more efficiently. CARA, our premium product, enables a lawyer to drag-and-drop upload a document they're working on, and will recommend the research that the lawyer missed but is very relevant to what they're working on (https://casetext.com/cara). A key ingredient behind this awesome tech is the network of citations that the article mentions.
Whether it's us or other startups, I agree with the article that in the next few years you'll see a trend towards more "Google for Law" -- companies will make legal research free, and their comparative advantage will be on their technology, often driven by ML/AI. As a lawyer/coder, it's a pretty exciting time to be in the space.
A really good book to read on the subject is AnnaLee Saxenian's Regional Advantage: Culture and Competition in Silicon Valley and Route 128 (http://www.amazon.com/Regional-Advantage-Culture-Competition...). It discusses why Route 128 failed while Silicon Valley flourished. A major part is that under California law, non-competes are not enforceable in California. As people moved freely between competing companies, their ideas, information, and best practices traveled with them. The diffusion of good ideas gave the region as a whole a competitive edge.
For what it's worth, the author, Ira Rothken, has been litigating these cases for well over a decade. Notably, he is Kim Dotcom's lawyer for Megaupload-related litigation. It's really cool that he's sharing some creative thoughts about the different safe harbors ISPs can use to fight copyright claims.
Keep in mind that there are a lot of lawsuits that are done for strategic reasons, even where the factual grounding isn't super strong. I'm not saying that's necessarily happening here, but it could be the case.
Although a very "legal" issue, whether people have "standing" (i.e., the ability to sue in court) companies for data breaches will end up having ramifications for how seriously companies take security concerns. Really interesting breakdown of some of the recent cases.
I guess one other way these things are exclusive is that they drive up costs for sharing economy companies. Some of these companies are pricing so low that they are barely making a profit as it is. If their costs are even higher, they will have to drive up prices, and at least part of their appeal immediately vanishes. At least a few of the companies will die if that happens.
Really interesting article. One important newsworthy note from this article that I think has been missed in some of the broader coverage: Washio, Postmates, and Shyp have been sued by their contractors/employees, joining the same club as Uber, Lyft, and Homejoy. This is an industry-wide issue, not isolated to just a handful of companies, and has the potential to severely damage this budding industry, as the author notes.
Whenever I take an Uber or Lyft these days, I always ask what the drivers think about the issue, and the uniform response so far (obviously this is anecdotal and not at all scientific) is that they like the flexibility of being able to choose when to drive. For many, this is their second job they do on weekends, nights, or early mornings. They are worried that they'll lose this flexible source of income that are helping, for example, save for their children's college education or pay their way through night school.
I'm not saying that this represents the majority view, or that the legal issue should come out any differently, but I do worry that some of these workers, happily employed as contractors, will lose out on this opportunity.
Casetext CEO here. First, thank you for writing this. Hearing honest feedback from people who try out our service and improving is precisely how we’re going to get better. And I love that you share both our enthusiasm for the idea and our optimism for the future of legal research.
I wanted to note that we will be getting a lot better. Keep checking in over the next few months, and expect to see a lot of major improvements on the issues you noted as well as many other things as well. Until very recently we were a very small operation, and as it turns out, building a competing product to LexisNexis and Westlaw is exceptionally hard. (I don’t want to make excuses – trust me, nobody wants Casetext to be much better more than I do – but the reality is that a perfect product couldn’t be built overnight.) That said, with the backing of amazing investors [1], we just assembled a killer team [2], and now have the people and resources to get there.
Another commenter (dmix) put it well: “Building hard things takes a long time.” Think of how long it took for a lot of the best sites on the web to become what they are today. We’ll get there – I promise.
Finally, we do have the case you were searching for [3]. Why it didn’t show up in search is something we’ll be looking into this weekend.
We are looking for a full stack engineer with design talent to help us rethink the way that people read, understand, and annotate texts.
Your mission will be to work with Casetext's full stack and take substantial ownership over large parts of the code. (A sneak peek of the new site is at http://beta.casetext.com.)
We aren't looking for experience with a particular set of languages. Instead, you should be comfortable collaborating with other engineers on a diverse software architecture. Our front-end is a single-page AngularJS application written with Jade and Less.js that uses Firebase, a cutting-edge real-time datastore, to make the site incredibly fast and responsive. Our back-end stack includes ElasticSearch, NodeJS, Go, and Python.
ABOUT CASETEXT
Casetext's mission is to make all the world's laws free and understandable. We have amassed an enormous database of legal texts, starting with over two million U.S. judicial opinions. A community of law professors, lawyers, law students, and citizens are adding insight and explanations.
Casetext is disrupting an $8 billion legal research market currently controlled by a duopoly (Westlaw and LexisNexis) that has barricaded quality legal information behind a paywall.
This is an opportunity to be an early employee at a rising start-up, take on a lot of responsibility, and play a substantial role in the future of the company. We are extremely selective with who we hire, but we make sure that our early team-members are well compensated in equity, salary, benefits, and quality of work.
You’ll be working with engineers from Google and IBM, the president of the Stanford Law Review, and former practicing attorneys from Yale’s and Stanford’s law schools. We are a Y Combinator company (Summer 2013) and have raised a seed round of over $1.8 million.
Perks include health/vision/dental, meals and snacks covered, and transit. We're in Palo Alto, blocks from the Cal. Ave. Caltrain.
We at Casetext are interested in this space as well. Our annotations technology will undergo some major improvements in the near future, but even the way they are today may be sufficient to meet your needs. Hit me up (jake at casetext.com) if you want to work together to make this a reality. Here's an example of what docs look like on Casetext now: https://casetext.com/contract/simple-agreement-for-future-eq...
- Variables should have more types, like an array of objects
- Prompting should incorporate Jinja2/Nunjucks
- For every prompt, I should be able to create many different test examples, along with an answer key, and measure how well it does across many tests
- It should auto-save. I did a lot of prompting work and then clicked another icon. When I came back, all my work was gone. (In fact, I don't see where to save at all! Maybe I'm just missing it.)