Why We Started Early
It’s been 15 years since I wrote my last line of production code, and then, in the past 7 months, I generated over 30,000 unit tests and 300,000 lines of working test code, approximately.
The Power of Craftsmen’s Tools
Three weeks ago, I had to quickly learn Python. My daughter, a graduating high schooler, needed help with submitting her final cybersecurity project. I showed her the tool I’d been using for the past year to solve problems - effective ChatGPT prompting. We spent time on how to ask the right questions, get quality answers and validate them. This resulted in days of work saved, with better outcomes. When her friends asked for help, I couldn’t say no, even though this is one of the busiest times of my life. One of them had been stuck on a bug for a couple of days. I showed her how to ask ChatGPT the right question. A three-day blocker bug got resolved in 30 seconds. More importantly they acquired a new skill. These bright young ladies don’t realize how much I learnt from them too.
Why am I telling you this story? Because it directly relates to what we do here at EarlyAI. We are here to deliver a very powerful tool for developers; one that will save them days and nights of fighting bugs and will help them deliver better software, faster.
We are in a new Era where every developer can become 10x better if they use the right tools, effectively. Imagine what it would do to the developers that are already 10x more productive than others. More importantly, what new, sophisticated products can be built with such capabilities.
EarlyAI is on a mission to build the perfect coding environment that will deliver bugless code and a leap in quality and development speed.
Keep reading.
The Stars Have Aligned
Founding a startup has been a dream of mine. I’ve always been passionate about taking an idea from zero-to-one - creating something from scratch, building it up and seeing it thrive as a business.
And I always knew that a startup is a team effort. Lior Froimovich, my co-founder, and I worked together for 5 years at EY Global, and the intensity and fun since we started this venture is worth at least 5 more years. We both share a long career of engineering and product leadership, spanning from very early stage startups to some of the largest corporations in the world. Those who know us see how we complement each other. I really enjoy our brainstorming sessions and the amazing ideas that come out of them.
Lior is a technological prodigy. He is the 10x engineer that is becoming 100x with the new GenAI tool he uses. Personally, I’ve had the privilege of working in Silicon Valley, the epicenter of technological innovation, for 17 years. My experience includes building AI solutions at Yahoo when it required a cluster of 1000 machines… and serving as the first VP of engineering at Couchbase, which later on went public. I was also the CTO of Bina technologies, a successful life science startup acquired by Roche, where we built cancer genomic products, and more.
Now, the stars have aligned for us. Our professional experience, the team we’ve assembled, our network and our personal lives have given us the capabilities, freedom and support to pursue our startup goal. Just as importantly, the technological landscape has evolved.
LLMs, The 21st Century’s “Combustion Engine”
Generative AI has exploded, offering a brand new way to solve problems. There are now numerous opportunities for development on the horizon. Leveraging this new technology makes me feel as excited as the early inventors of the Industrial Revolution in the 18th century probably were. They must have asked themselves, “What can we do with this magical machine called a combustion engine?” GenAI is our 21st century combustion engine, and we can leverage it to build products that were unimaginable just a year ago.
Our vision is to expand GenAI’s intelligence and autonomy and redefine software development, and its quality and possibilities.
Liberating Developers From Bugs: Once and for All
Throughout our careers, Lior and I faced the problem of software quality. Or rather, lack thereof. While software is the foundation of our future, its quality determines the type of future this will be. High-quality software ensures that products and services perform as intended and meet users’ needs efficiently. Bug-free code helps avoid potential system failures, security vulnerabilities and costly downtime, which can make or break a company’s reputation and longevity in a competitive market.
We are not the first ones addressing this problem. It is a real pain and, as such, many organizations have tried to support developers and liberate them from bugs. However, most companies’ attempts focus on process optimization, delivering a better “dashboard”, not fixing the engine itself. This means - developers will keep introducing bugs.
Our approach is different. Instead of assisting developers to fix bugs, we let developers focus on what they love most: coding. We’ll take care of the rest, meaning the test. Our technology takes care of writing unit tests (to begin with) on behalf of the developers. It’s the first step towards taking the task (or at least most of it) of developing unit tests off of developers’ plates, freeing them up to focus on the application code, the one that brings value to the business.
This new way is almost magical. It enables us to tap into the essence of building quality software. Devs will write the code, we will generate the tests for them, from A to Z.
This is a journey of course, it won’t happen overnight. Today, LLMs are mostly assistants for performing tasks. In some cases the results are breathtaking. In others, not so much. But in any case the user (developer) is responsible for the results and has to improve and validate them. We aim to become an AI agent for test generation and code quality, relieving developers from bug fixing altogether, and our starting point is far off from any solution that exists today.
So What Does Early Do?
We’re here to address the task (and pain) of writing unit tests and take it away from developers.
Each one of us experiences this pain slightly differently. Some care about the code quality, others are instructed to move faster but the bugs will hunt them later on, yet a third group just want 70% code coverage because it’s their organization’s standard and they can’t commit otherwise.
EarlyAI knows how to generate complex unit tests, working tests (not code snippets), tests that pass, and also tests that fail (red tests), to find bugs in your code so you can fix them faster. As such, EarlyAI enhances visibility and reduces the heavy lifting of writing tests and looking for bugs.
We are also working hard to measure and increase the quality of the tests themselves, which is just as essential for building and maintaining quality software. For this, we’ve developed the EQS (Early Quality Score) which will be discussed further along our journey, and takes into account code coverage, the mutation score and the ability to generate working unit tests.
The proof is in the pudding: We are already using our product to develop our product. This has allowed us to find different ways to become more productive. 100 times more productive. And improve the product and usability themselves.
Plus, during our tests with EarlyAI, we were able to generate 930 unit tests in 30 minutes, covering over 50% of ts-morph, a fairly large scale OSS project. EarlyAI covered 12,000 lines of code, worth about 1 year of development, in less than 30 min. [read more here on our benchmark blog]. I probably generated over 500,000 lines of code in the past 6 months testing and experimenting with our product. More code than a lifetime career.
In fact, why don’t you try for yourself? In our sample todo app you can generate over 100 unit tests for 24 public methods in 2 minutes. Clone our sample todo-app, install EarlyAI on your VSCode IDE, search for where coverage is low on the tree, and generate tests for the remaining 5 methods.
Where Do You Fit In?
The future of developers is to lead, develop and generate more sophisticated solutions, faster than ever. To do so, they need tools that will empower them. We aspire to be that test tool that every engineer will search for.
Now it's on us to prove to you the value of these lines of code we generate, the quality of our tests, and ultimately, help you release more sophisticated software, faster than ever.
We invite you to take part in this journey, allowing you to leverage our product to produce higher-quality code and release higher-quality products, faster, without the overload. You can connect EarlyAI straight to your IDE and use it to generate verified and validated unit tests Start here.
Thank you
I would like to thank my wife and daughters for their support, my partner Lior who's wisdom and friendship has no limit, and for the very long list of family, friends, users, founders, CEOs, executives, and investors who are part of our journey.
Yours
Sharon Barr, Co-founder and CEO