Fuzzy.ai API service to end on December 31, 2018

The last day of service for the Fuzzy.ai API at https://api.fuzzy.ai/ will be December 31, 2018. We will release the full stack for the API service under copyleft licenses. Data will be available for download on demand.

Unfortunately, we haven’t made enough money with the Fuzzy.ai API to cover costs and pay the development team, so we can’t justify its continued operation. The team has moved on to greener pastures at other companies; the founders will soon follow.


Developers who want to recover data from the API should send an email to support@fuzzy.ai. We will be able to service these requests through the end of the year, and on an best-efforts basis in Q1 2019.

Open Source

We want our users who have come to depend on the API to be able to set up their own servers for future use. (This should require only a small change in client code.) We continue to believe in the importance of artificial intelligence for all developers. We hope that an Open Source API will be a useful contribution to their toolset.

The software will be available under the following licenses:

  • LGPL 3.0 or later for our core fuzzy logic engine and adaptive fuzzy learning engines
  • AGPL 3.0 or later for the API, development IDE, and microservices needed to make the engine run

We will provide a Helm chart to make it easy to launch a new Fuzzy.ai API instance on your own Kubernetes cluster.

Watch this blog and https://gitlab.com/fuzzy-ai for news as we manage the release process.


Matt and I want to thank everyone who believed in our vision for  artificial intelligence available to all developers.

First and foremost, to the developers who have put our product to the test, made it part of their stack, stretched its envelope, and provided valuable feedback and Open Source contributions. We hope you continue to put your users first and make your software intelligent and adaptive.

Thanks also to our investors: Interaction VC, Real Ventures, iNovia Capital, 500 Startups, Mark Cuban, Julien Smith, Kai Gradert, Julien Genestoux and Stav Prodromou. Your capital and guidance let us put our case before the market. Thank you for providing that opportunity.

Thanks immensely to our startup and tech community in Montreal and San Francisco for the intros, suggestions, and cheerleading. Thanks to our advisers, friends and family for their help, support, and ideas. Thanks to my kids Amita and Stavy for being confident and believing in me. Special thanks to our wives Cassie Sera and Michele Ann Jenkins who have been proud, supportive and forgiving.

Finally, we want to thank our team: Kevin Fox, James Walker and Pablo Boerr. You put in intense effort, creativity and late hours to make a product that mattered. We can’t count the number of times we’ve accepted congratulations for your great visual design and friendly, usable software. We expect to see even more awesome work from you in the future.

We, the founders, will be moving on to other endeavours in 2019. I (Evan) will begin a new position as Product Manager for the API at the Wikimedia Foundation, the folks who make Wikipedia and other great wiki products. Matt will continue to share his experience and expertise as a consultant and mentor for other startups.

Democratization of AI at Mozfest 2017

A quick note that I will be talking about democratization of AI at the Mozfest 2017 event in London this weekend. Fuzzy.ai’s goal has always been to make it easier for developers to include artificial intelligence in more of their software, giving the ubiquitous experience we call casual intelligence. I think this is an important issue for technology and society, and I hope this session will help inspire software makers and thinkers to focus on more widespread and diverse AI development.

What should you do next?

I gave a talk last night at MTLDATA about using Fuzzy.ai for task prioritization. It’s a really interesting subject for me, because tasks are an atomic part of our life and our work. How and when we complete tasks is an important part of any person’s productivity, and artificial intelligence that helps us do that effectively is really valuable.

Most of us use a task management system — whether it’s part of a project management system like Github issues or JIRA, or a personal system like Todoist, todo.txt or Google Tasks. All of them have ways to define important data about our tasks, but very few of them make intelligent recommendations for what to do now.

Intelligent organizations are starting to work on this problem today. After all, having people inside your company work on tasks that are timely, impactful, actionable and personal can make a huge difference. Tying the performance of tasks to metrics within the organization can drive much better relevance tracking.

To test out these ideas, we built a cool task prioritization project using the Todoist API. If you log into the site at https://tasks.fuzzy.ai/ with Todoist authorization, it will automatically score and rank your upcoming tasks, putting the most relevant items first.

Todoist has a relatively simple schema for tasks. We used that data for building a fuzzy.ai agent that uses time, priority, and other features to determine the task’s relevancy.

We think there’s some interesting future developments here. Keyword matching with previously relevant tasks is a big one. Providing other backends, like Github issues, is another. (I use Todo.txt, so I’m especially interested in this.) Most of all, building in different kinds of performance metrics to see if your tasks really matter is going to be the biggest.

Our code is open source. Feel free to fork and try on your own, or send pull requests. We’re always interested in what people figure out to do with Fuzzy.ai.


Security issues in Fuzzy.ai developer environment fixed

We had three security issues in our developer environment reported by security researcher Ron Masas recently, which we’ve identified and repaired. Thanks to Ron for his help in identifying these issues and in suggesting some ways to correct them. (And special thanks to our lead developer James Walker for getting them fixed so fast.)

The first pair of issues was a path that allowed saving an unsanitized email address to our user database. Combined with a way to share a user session across users, it allowed a cross-site scripting attack.

The third issue was a cross-site scripting attack caused by the way we were pulling data into our default React session. Carefully restructuring the request would cause a user’s browser to send their important session data to a third party. We repaired this bug by restructuring how the default session data is injected.

We don’t know of any abuses of these bugs in the wild.

We think it’s important to be transparent about security issues. We especially want to encourage security researchers to share their findings with us and other application developers. Thanks again to Ron for the great work.

We are Montreal AI

One of the great outcomes of the AIFest event in Montréal last weekend was an increased sense of unity among the different parts of the AI ecosystem here. We are all working together to make AI a new industry in this city. 

To that end, we have created an open letter to the applicants for the AI Supercluster program in Montréal. This is a federal program to invest in AI. We want to work with government, industry and academia to make the ecosystem a success. 

The letter is at wearemtlai.org. Fuzzy.ai’s co-founders have signed as well as dozens of leaders in the community. We encourage participants big and small to make their voices heard. We can only do this if we work together.