This year marks 15 years of making software-based legal experiments. I started down this path with winning a hackathon put on by Yahoo! in my final year of undergrad at the University of Waterloo. My two successful experiments, which became businesses, were in the search space and another in the regulatory management space.
None of the projects below have ever been a particularly great source of income, but they've been valuable in terms of building my knowledge of programming, law, and business. As a tech lawyer advising clients, it helps if you know at least a bit about what your clients do. On a less practical note, each of these legal experiments is an attempt to see whether how people use and access the law can be improved. There's also just a certain joy to building something that's never existed before.
Here are my experiments over the years, starting with hackathons in 2008/2009:
2008: Political Data
I was the Yahoo! hackathon winner at Waterloo for a system that showed the most common words spoken in Canadian Parliament by MP and by party. The idea was the the public ought to know what politicians say in the House of Commons, rather than just what the platforms of their parties are. My thesis was that if the public knows more about what their representatives are up to then democracy will be stronger. This service worked and was public, but it wasn't a sustainable business and it turns out not many people are interested in knowing this sort of information.
In a similar vein of democracy, I developed software in the summer for the 2008 Presidential Campaign for Ralph Nader, notably a system for tracking the petitioning crews as they travelled throughout the US. Because people didn't have Internet on phones at the time, this was done by having them submit their daily counts as text messages which were aggregated through Twitter, which at the time was a platform based significantly on text messaging. This gave the public a near real-time view of how many signatures were collected in the state-by-state petitioning drives. It was a fascinating experience to see how American politics works on the ground and how technology can connect political campaigns with the public.
2009: Financial Data
In my first year of law school I took a few days off to attend the international finals of the Yahoo! hackathon that I had won the previous year, in New York City. I programmed for 24 hours straight through and made a brand new system that looks through SEC filings of insider trading (on EDGAR) to show the public when the executives of public companies buy and sell shares in their own companies. This data was publicly available but not organized, and at the time there was nothing like this available for free. I won the finals with this submission and got to meet one of the judges: the creator of the PHP language, Rasmus Lerdorf. He offered me a free server that I ran this service on for a few years. It was used by various investment professionals at big US banks (I could see their emails) but I never turned it into a business. I learned there's a lot of demand for financial information. I also learned that this kind of service couldn't be made for Canada because the government makes people pay for this data (and they still do).
Yahoo! tried to recruit me and then around the time of finals I had a call with a Facebook recruiter. They said I should quit school and come join them. That would have been a good financial move, but I figured I'd never have a chance to finish my law degree and there'd be more doors to open in a few years.
2010: Law Data
I was sitting in my securities law class and the instructor, a partner at McCarthy Tetrault in the securities law group, noted that they happened to come across a new law that was relevant to a deal they were working on. I was very surprised to hear that lawyers didn't already have tools to alert them to new laws. Thinking about this, and my family's background in government lobbying, I realized there was potential to make a service that would alert people to what's happening with the Ontario government.
The service ended up including bills, Hansard (speeches), regulatory proposals, royal proclamations, and many other features. Lawyers were surprisingly uninterested but OntarioMonitor.ca was used by many of the top government relations firms, several very large multinationals, and, surprisingly, the government itself. It turns out that the different parts of the Ontario government have no idea when laws are being passed or when their areas are going to be discussed in the Legislature, so they ended up becoming customers to monitor the government, while being a part of the government.
Later I expanded to include the federal government (federalmonitor.ca). Unfortunately neither of these two websites is still available because I ended up selling them in 2017.
2011: Environment Data
I've always had a fondness for the environment and spent a fair bit of time around environmental groups growing up. A customer of OntarioMonitor.ca asked me to build a custom version of it for monitoring the Environment Registry, and this tool was used to send millions of alerts to the public about environmental consultations.
2012: Bill Prediction
OntarioMonitor.ca's customers were quite sophisticated people but I learned that few of them knew whether laws were likely to be passed or not. So I built a fairly simple algorithm (and backtested it) that could predict whether a bill was likely to pass with about 80% accuracy. There are several factors that are highly determinative, such as who the person is who is proposing a bill, and whether or not their party is in power. This tool was very powerful but there was never much interest in it. I learned that the core features are what's important and usually no matter how great the add-ons are, they don't really cause new people to subscribe to a product. This was one of my first predictive algorithms.
Most of my software work was on hold this year because I was articling at McCarthy Tetrault, a major law firm in Canada which isn't known for their relaxed attitude towards work.
2013: Legal Marketplace
When I started my own tech law practice I was surprised to find out that there wasn't a place for lawyers to offer their services. A few lawyers were advertising on Kijiji. There were Google Ads to try to buy customers. But there wasn't a website to visit that shows people what lawyers are available to do what work. So I made my own: flatlaw.ca. This was the first flat rate legal marketplace in Canada, but it was far from the first in the world. There were a number of great US platforms at the time. After making this, I met many international founders of somewhat similar platforms, including a very interesting one in Germany for questions and answers where the lawyers get paid a small fee to publicly answer questions.
FlatLaw.ca was covered by Canadian Lawyer Magazine and the Globe and Mail. Its legacy lives on in the form of lawyers regularly emailing me to ask about what happened to it (I sold it five years ago and the acquirer sold it on to someone else).
2014: Global Laws
My OntarioMonitor.ca service led a PhD student to reach out to me about the prospect of building a global version of OntarioMonitor. He had previously worked on a government legal technology project and had worked as a lawyer. He was now studying regulation in Canada in order to become an even better lawyer. I liked his idea but it wouldn't have been feasible. With a bit of brainstorming we came up with the idea of using the newly available translation systems (Google Translate) to gather the world's laws and make them available in English through a search engine. This was a bit like Google's original idea of making the web available, but in our case we focused on making what isn't on Google available. I ended up writing a framework for writing indexing and parsing systems for law websites which made it very easy to develop 150+ different sub-systems that gathered up around 1.5 million laws from almost 100 countries. This service has been used by major governments, KPMG, most of the top research universities in the world, and many others. It's still available online and is made available in the United States through LexisNexis.
My co-founder, Nachshon Goltz, got his PhD, and still runs Global-Regulation.com today. He's the author of a couple interesting books about tech and is a Senior Lecturer at ECU in Australia.
There were so many lessons learned from making this service, such as:
-No one ever believed us that Google didn't already have the laws. The reason why it doesn't is that most of these laws aren't accessible by general web crawlers.
-Hardly anyone in policy or similar jobs in government has access to LexisNexis/WestLaw or other legal research tools.
-Everyone assumed this service already existed! We'd ask them to name the service they were thinking of and they couldn't.
-Machine translation of laws was good and continually getting better. The global language divide became a solved problem around this time.
-Very few people care about the laws of other countries.
2015: Court Statistics
In 2014 I began a few experiments based around studying all of the ONCA and SCC cases in Canada to see if there were statistical conclusions that could be drawn. I wrote up my results on my blog (ONCA word frequency, ONCA cost awards, more on costs, and a look at who the top ONCA litigators are). I learned there's a surprising lack of data in the legal world. One lawyer who worked for insurers said that my statistics were shocking to see because their insurance companies regularly demanded data but the lawyers could only say that each case stands on its own merits. Although true, at a bigger level, there really are conclusions that can be drawn from patterns of legal decisions.
In 2015 I finished up my experiments in this area, concluding that there wasn't much of commercial value because I think court decisions are so random that there aren't really steps that lawyers can take that will improve outcomes (that they're not already doing). I published a calculator for the odds of winning at the Supreme Court, a list of the most prolific litigators, and the top SCC firms by filings. I met some interesting lawyers by publishing this data but there's no business in doing this.
2016: Global-Regulation.com Growth
Global-Regulation.com was starting to make money by this point and we'd struck a deal with Microsoft to provide more translation. The key to this business was my partner who managed to get $250k of support from Google, Microsoft, and Amazon. At one point we were getting subsidies from two overlapping tech giants and Microsoft published a case study on us for using multiple cloud services. On a shoestring budget we found and translated millions and millions of pages of laws. I learned how hard it is to sell a service in this space, mostly because the customers are hard to reach but also because of the assumption that someone else was already doing this. In fact, the big law giants weren't doing this, because they focus on having excellent coverage of single jurisdictions. But the time was right for this service because even developing countries were posting laws online by this point.
2017-2018: The Crypto Boom
In 2017-2018 there was a big cryptocurrency boom. My law practice was overflowing with clients and my best client asked me to join as their Chief Legal Officer. Not long after that, I became the President and oversaw the research, development, and launch of a brand new cryptocurrency wallet. I didn't have any notable legal experiments from this period, other than perhaps my own life, since I took a leap by closing my successful law practice to join a business I hoped would be even more successful (and it was).
2019: Better Blogging
2019 was a big year in cryptocurrency law and I published a number of long form blog pieces. I'd been making legal information available for years before on my blog but I resolved to write larger pieces and take on stronger positions.
2020: Crawling The Law
Although Global-Regulation.com was a success, I always dreamed of making a Google-like service focused on the law. I saw through my law practice that legal information was increasingly in the form of documents posted on regulator's websites, rather than published through official channels. Global-Regulation.com only ever tried to capture primary and secondary legislation (acts and regulations) but the daily work of many lawyers, compliance, and other professionals is sifting through government documents. To do that effectively, you need a special type of search engine. I published an early attempt at this and wrote up what I'd learned.
A key lesson from my search engine work is that there's a number of regulators who are using advanced techniques to block any search engine that isn't Google or Microsoft. This is an anti-bot measure, but it's also a fundamental challenge for anyone who wants to build a service in this area because there will inevitably be gaps caused by these regulators blocking public access.
2021: Confidential Conversations
The tools for lawyers to speak with their clients are the same as for anyone else. I tried my hand in 2021 at providing a better way of communicating using encryption because of the newly available WebCrypto API that became standardized in browsers around this time. My experiments with this weren't particularly successful because they'd require a lot more engineering time than I think is justified for me. I published some of my findings about doing this and a few entrepreneurs reached out to me about their own efforts to build these sorts of services. Even in 2023, I think there's a big hole in the market for better management of encrypted conversations with clients for lawyers. Lawyers still don't take confidentiality seriously enough. But it's a very tough sell to offer lawyers a tool even more complicated than what they use right now - I was on a lot of Zoom calls at the time and it shocked me how many lawyers couldn't figure out how to use that. Coincidentally, it was in 2021 that Zoom's fraudulent claims of end-to-end encryption were made public. Security is very important, difficult to make available, and poorly appreciated commercially. It's probably not a great market.
2022: Law Firm Content
Law firms have a treasure trove of public articles that examine and explain the law. I've always thought that a good search engine for law would surface this kind of content, or integrate it into Ai tools. I wrote up an attempt at doing this using the search engine I'd been tinkering with over the last few years. I think there's room for an offering like this but it would be an uphill challenge to get people to pay for this. Ultimately I decided that a search engine on its own wasn't going to be a viable business, however interesting it is to build, because general purpose search engines like Google and Bing are so good. Any gaps between what lawyers need and what these offer probably won't be a great market on its own.
2023: The Rise Of AI
ChatGPT changed the game for AI. I wrote about some of my experiments such as evaluating contract clause generation, problems with legal reasoning, and
I also published the results of my experiment to take my encrypted chat ideas from earlier and make it into a contract signing platform. I like this idea a lot but it's too much engineering effort to take it from experiment to useful prototype. When it comes to contract management, even a prototype needs to be reliable and long-term, otherwise it's not really something people can use other than as a toy.
I reused my search engine code from earlier and mixed it together with ChatGPT to extract structured data from job postings. This process works surprisingly well, and I was amazed at how good the ChatGPT API (GPT-3.5) could be for doing this sort of work which would have previously required writing many different dedicated parsers (like how global-regulation.com worked). I'm still making this one available at https://carrentis.com, and I've had some interesting feedback. I think AI is the future of finding jobs, but probably I won't be the one to make that future into a popular service. In a few years people will finally achieve the 1990s-era dream of having an AI agent that can go out and identify suitable jobs, and perhaps even apply for them.
A few months after the job search engine prototype, I published a similar effort for extracting anti-money laundering rules from the FINTRAC website in order to answer anti-money laundering questions. Although not perfect and far from ready for a commercial product, it's amazing how good a fairly basic AI system can be at answering anti-money laundering questions. This approach is extensible to many other regulators.
I expect to publish more experiments about AI and law in 2024. The AI services are getting better very quickly so I expect we're on the cusp of brand new possibilities for automating the process of turning complicated laws into easy answers.
Conclusion
It's been an amazing 15 years of experimenting with law and technology. Although most of what I've done hasn't changed the world, I have met a lot of people along the way, and hopefully inspired some of them to try to tackle these big problems. Legal advice might actually be even less accessible in 2023 than it was in 2008, despite huge advances in technology. I am most hopeful about how ChatGPT-like approaches will help to change this, but there's a lot of other ways to tackle the problem.
Taking the rules of society and transforming them into practical everyday advice is one of the great challenges of our time. For thousands of years, people have done this through painstaking and often error-prone analysis. But because of the massive increase in the volume of laws in most countries of the world, it's becoming increasingly difficult to do this through manual means. AI should be perfectly suited for combing through this vast amount of information, but the tools just aren't there yet, and the data is even less available.
I hope you've enjoyed reading about my 15 years of trying out ideas. Hopefully it gives you some inspiration to go out and do your own experiments. There's never been a better time!