This is a lightly redacted and shortened version of the update that was sent to our investors. Our investors will always receive these updates meaningfully ahead of any publication.
To the investors of Mailpass Pte. Ltd,
I will be switching up the format from here on, to try to give more insights into our approach and direction, alongside notable developments that are affecting or influencing the path of the company and the product(s).
To get the basics out of the way, our PassGuard product saw confidential data paid seats, Wyntk saw confidential data user growth. Our runway stands at confidential data / confidential timeline with our current burn rate of confidential amount and room for future headcount. We also anticipate modest increases to OpEx after July and when most of our cloud-inference credits expire.
–
Two themes have influenced how May played out - the sheer leaps and bounds virtually weekly that AI-first development/coding is making, as well as, the dramatic rise of the company Palantir. They have been applying a ‘forward-deployed’ approach successfully for the past decade - and are now accelerating beyond all expectations with the help of LLMs.
AI-first development/coding
When I first started using Cursor to help develop Wyntk back in December, it was mostly helping to add a few tedious functions or UI elements here and there.
Fast forward to today, and Cursor + Claude 4 can reliably write about 95% of the code used within our applications. This dramatically changes the software landscape where a feature can be developed and pushed out almost daily, based upon a set of tenets that I set out in the beginning - such as how our auth was going to operate, where our data would be stored, and which AI models we use within the app itself.
However, users are not even close to being able to keep up with a platform where this happens. Similarly, the rise of ‘MCP servers’ especially since April that enable a user to connect virtually any type of data to each other in a stream controlled by LLMs has huge potential, but actual adoption by the average non-technical user is extremely low. The quality of outputs and what is delivered to the user is still in its primitive stage. Just being able to join up data is nothing unless - in simple terms - it drives better decision making, whether that is for personal or business use.
Which brings us back to Wyntk and how we’re approaching that product. Email still requires a lot of manual labour from the user to keep it in check and derive any meaningful data from it, to then drive any kind of decision. We don’t believe in just adding a prompt box, connecting your email account, and letting the LLM rip.
I fundamentally believe ‘chatbot’ interfaces are/were just the first step to our regular interactions with LLMs. How we can produce, use, and meaningfully benefit from interfaces backed by LLMs is where we are headed, and this underpins everything that Wyntk is looking to achieve with users, primarily with their email initially.
In layman terms, this means taking the grunt work out of labelling, sorting, sifting, prioritizing emails every day, without the repeated need for prompting, so that the user has what they need when they need it to make the decisions they need to make.
I encourage you to give it a try, and especially our ‘Rules’ feature that enables you to set up a rule in natural language for a specific sender or domain. Set, and forget.
Palantir and the forward-deployed approach
Having worked in the past with ex-Palantir employees, this has always stood out to be an enigmatic company, known for better or worse for its deep integration with the US military. However, what people perhaps are less quick to understand, is that the money it generates largely comes from the private sector, like helping Airbus put its A350 programme back on track back in 2015.
The key to Palantir’s success has been developing re-usable products out of problems garnered in the field. They do not consult for a company, create a bespoke solution, and then leave it with them. The learnings are continually rolled into products that are deployed throughout the military and private/public entities alike. The key to making this work for both sides, however, has been the speed at which they go into the field and look at the problems and iterate on solutions there and then (literally, coding overnight and bringing back solutions the next day).
If you combine the speed of development gains mentioned above with my past experience - sitting inside companies as an external entity to build solutions for them - and apply a productized approach, the opportunity becomes clear. We are well positioned for an exciting era where mid-sized or even large Palantir-like companies will emerge.
–
So I am excited to keep pushing the envelope with Wyntk into users’ hands - and I hope - organisations over the next few months, and deeply learn from their experience with the product and pain points with email. Equally, we already have made inroads with companies in the confidential data industry where we believe we are able to help and productise their problem sets. Email will almost certainly come into the picture and we expect to leverage the approaches and features being developed into Wyntk as part of this.
–
If you have not already read the - 362 slide - report by Bond / Mary Meeker and her team that just dropped, I highly recommend it to also help frame at just what pace things are moving for developers, infrastructure, costs, and of course the end users/organizations/governments.
It is fair to say we are well placed to take advantage of increasingly falling inference and LLM costs, as well as a golden age of AI-backed development where we have the power of at least 2-3 developers via one pair of hands + Cursor / Claude.
The pace highlighted in the report is also throwing many common metrics out of the window - such as ARR. Whilst we have seen unprecedented growth of companies (like Cursor) reaching $100m ARR, we are also witnessing an explosion in ‘tourist users’ who use and pay for a product for 1-2 months, then switch it out for another one. Extrapolating MRR to ARR makes less sense when months have become days in terms of development speed.
It should also be noted (in the report and through my observations) that developer tools have driven most of the unprecedented revenue metrics of LLM/AI products seen of late (e.g. Cursor, Bolt.new, Windsurf). The challenge we are solving is taking these huge efficiencies and turning those into the same advantage for non-developers.
–
I will be looking to give these updates around every 2 months from now on. Our finances do not change dramatically month to month as you have hopefully witnessed, but the landscape does. If you ever have any questions about our fundamentals or the business, do not hesitate to email me.
Thank you,
Gregor Vand