The Hex team is excited to announce that our Magic AI-assist tools are now available in Public Beta. This brings access to tools for generating, editing, debugging, and documenting SQL and Python to thousands of you, right from where you’re already working:
Starting now, anyone on Pro, Teams, or Enterprise plans can turn on Magic by flipping the switch on their settings page. If you’re not using Hex yet, you can start up a trial here.
Still a beta!
Before we dive into some more details, a big thank you to everyone who participated in our Private Beta these past 10 weeks. Your feedback (and patience!) were essential in getting Magic ready for everyone else.
While we’re excited to expand access, Magic features should still be treated as beta. Like any AI system, it isn’t perfect, and will make significant errors. You should plan to carefully review generated code, and treat Magic as an accelerant – not a replacement – for human judgement.
New features and improvements
Since the launch of our Private Beta in February, our circle of magicians have been hard at work, tackling some major infrastructure improvements and adding new powers.
First, we upgraded the underlying models and prompt strategies. Magic now uses GPT-4 for most tasks, which, along with significant updates to our prompting, is creating better completions, with much lower rates of errors or hallucination (again, though: still not perfect!)
Second, Magic can stream responses. This is a huge improvement for longer-running completions, allowing you to start code review during generation, and ensure they’re useful responses (and cancel if not).
Third, we added a couple small improvements to continue to make Hex feel more magical. Magic CTE Explode takes big nested queries and breaks them up into little, bite-size bits which utilize Chained SQL cells. Magic Cell Title is fun too.
You can see all of these in action below, or visit the documentation for a complete list of tips, tricks, and examples of using Magic:
We’re hard at work making more improvements to Magic, focusing on continuing to make code completions more reliable and accurate. We are also practicing some new, powerful spells (including some from the Restricted Section) that we can’t wait to show you soon!Magic is still free to use while in Beta, so we can maximize access (and have some time to think about how pricing should work). Of course, we’ll be in touch before we make any changes on this front.
Privacy and security are always our top priority
Data security is our top priority, and of course we realize that many people have legitimate concerns about privacy while using AI. A few notes and reminders on that front:
Hex doesn’t – and doesn’t allow any third-party partners – to train models using customer data, reducing risk of IP leakage through passed context. This means a model won’t spit out your code to someone else (or vice-versa).
Hex doesn’t send underlying customer data to the models. This means that values of specific rows of a table aren’t at risk of leakage through Magic.
Hex does use schema information and project code as model context. This means that any sensitive information in your table or column names, or in your code, could be passed to a model.
Magic is built on Hex’s existing, secure data platform, and is protected by the same practices and policies, including SOC 2 Type II, bug bounty programs, and principles of least-access.
We’re also going to hold off on rolling this out to EU or HIPAA stacks until we have a chance to work with those customers on some specific considerations.
We’d encourage folks with any questions to check out our documentation on data privacy for Magic or reach out directly.
We also know some teams might not be there yet in terms of adopting AI for their workflows, so we have added an option for account administrators to opt-out of Magic for members of their organization in the settings page.
Magicians wanted!
We’re growing the Magic team, and are looking for talented AI engineers to come build with us.
Candidates should have deep experience in AI, with an interest in how data analytics workflows can be changed for the better. Our team is small, focused, and fast-paced (Magic has been just 2 engineers to-date).
If this sounds like an opportunity for you to do your best work, please check out the job post and get in touch.
If this is is interesting, click below to get started, or to check out opportunities to join our team.