Rule 1: We believe the Developer Experience (DX) is all that matters.
We have listened. You have been loud and clear. There is too much friction to get to a working “Hello World” and way too much friction to get to production in the AI Developer space today. Even worse, when you get there, all of the knobs, buttons, and dials treat you like a punching bag full of configuration mistakes. However, a great DX hides all of that complexity during both the development and production phases. Additionally, a great DX lets you focus on going where you need to, controlling through TCO and budget, with push-button deployment and easy observability; and using existing knowledge and skills (Python!) to operate a system that’s significantly higher level than your typical Platform as a Service (PaaS).
How do you deliver a great DX?
You start by listening to your target audience and asking questions: about their top challenges to get faster Time-to-Value (TTV); what keeps them up at night; and if they had a genie’s wish to solve all of their problems, what would the genie conjure? After you talk to about fifty different people, you get some amazing feedback on what is working and what is needed. However, after you talk to a hundred people, you get a truer understanding of the problems. We spoke to over three hundred fifty AI Developers in 2023 to gather all the feedback we needed to start, but we are working hard to gather more. Come talk to us! (We’ve also met some great new friends along the way and are forever grateful for their support thus far.)
Next, take this feedback and start putting your company in the shoes of your customers. TTV is what matters to them so a great DX is all about delivering a sea change improvement in TTV. Easy to use solutions; great documentation (in all the formats they desire: Document Pages, Demos, Tutorials, YouTube Videos, and Hands On Training); ability to scale - up & down - when needed; amazing community support; easy debugging; and predictable behavior. A great list to start from - but that is only the beginning. Continuous feedback and iteration are also needed to never settle and to always keep removing friction top of mind. Thinking: how can we do this in five clicks? Why not four clicks? Seriously, why not one click? Then do it.
If you listen closely, there are some themes around what you need to do to have a great DX for the AI community. First, this community knows Python. This becomes a must-have language to help meet the customer where they are. Second, the AI and researcher community is broadly comprised of PhDs in math or physics so forcing them to also be cloud engineering experts is a non-starter. Third is complexity. There are a lot of ways to build a product that makes your life as a product designer easy but it usually comes at the expense of your customer.
What does a great DX do?
When we’re traveling somewhere by plane, we usually operate at the level of deciding where we want to go, find and purchase the best tickets we can for our budget, and ultimately grab our stuff, board the plane and (hopefully without any delays) arrive where we planned. Now, we all know that before and during that flight, thousands of diagnostic checks happen, computers fire up, flight and navigation systems operate, reporting and safety systems monitor operations, and fuel and electricity flow to make that all happen. But we don’t need to follow any of that - or manage the people making it all happen - to get where we’re going. We think in terms of: budget, destination(s), quality of experience, time to arrive, all surrounding the reason we need to fly in the first place. And systems to reserve travel operate around that kind of UX typically. (If you want to fly your own plane of course, that’s a specialized but available experience for the tiny sliver of experts out there.)
The needed DX for the AI community is the same way. They want to write some Python, test it, and go to production, today. In order to highlight what a great DX is, let’s start by highlighting was a great DX is NOT. The traditional flow on for cloud deployment is the following: Step 1 - Pick your cloud service provider (but I don’t want to). Step 2 - What region do you want to deploy in? (but I don’t know yet). Step 3 - only gets worst (because I just want something simple) but here is a fun list of things that you have to do today to get started: set up your endpoint, set up your Kubernetes cluster, pick you security configuration, get your observability going, configure DNS, configure firewalls, configure load-balancers, and more. All the while, you are thinking, “When do I get to take off?” Not yet! You still have to set up GPUs (that is if you can find them), configure those APIs, download a model, containerize it, figure out how to configure a vector database, and so much more to just get your experiment working. Fly the plane yourself in other words.
So now you have your AI-infused app running and you can test it out. But what about when you want it in production? Out in the field with live users, possibly scattered around the world, maybe running continuously, but perhaps also only in short bursts with high traffic? Time to (first) Value is critical, but ongoing Value and ROI is actually what’s critical ultimately. What about when your ever changing and ever expanding successful app needs to operate in real world conditions? Usually that’s a whole other production (pun intended)! Perhaps you have to wait for an Ops or Eng team to help. Or maybe you have to navigate optimization and wire tools for deployment, observability and compliance together yourself. In other words: first build the planes and the airports; and then rollout a global air-traffic control system for the planes all to just get to your destination.
So what is the right solution for the AI Developer?
A great DX hides all of that complexity during both the development and production phases. A great DX lets you focus on going where you need to, controlling through TCO and budget, with push-button deployment and easy observability; and knowing only Python to operate a system that’s much higher level than your typical Platform as a Service (PaaS).
Seaplane’s Python SDK and AI runtime ensures your app is "always on" but without excessive cost during idle times. We offer global deployment with strategically located stateful partitions for your users worldwide. All this is achieved without you needing to delve into the minutiae of infra/platform micromanagement. By avoiding the complexity of writing and managing a Deathstar of microservices or functions, we’re not just enhancing reliability but also slashing your development and operational costs. Our customers, the ones crafting these AI-infused apps, get to sidestep that entire chaotic world.
As you can see, a great DX actually does more than force you into becoming a pilot and an aircraft mechanic; it simply gives you a first-class round-the-world seat on a Seaplane. So sit back, relax and enjoy your flight!
Get Access Now!
Seaplane is where AI-infused applications take flight in 2024 so join us for this ride by signing up below. We are currently in Beta so make sure to sign up to secure your place in line and we’ll get to you as fast as we can.
Bring Your Apps to the World