A Letter from the CEO: Putting the Platform to Work
By Jill Kenney
When I wrote my last letter, we had just switched on SN121's incentive mechanism.
That gave us the engine to continuously build and improve the AI capabilities available through sundae_bar.
Today, I want to talk about what we've built on top of it.
Because having hundreds of AI Skills and Agents is only useful if people can actually put them to work.
That's the problem our team has been focused on.
Most people now understand that AI can help them do parts of their job. The harder question is how. Which AI should I use? How do I set it up? What happens if one tool only does part of the job? And how do I know what I'm using is actually any good?
We don't think our customers should have to answer those questions.
Our goal is to become the platform people rely on to get work done with AI.
And the latest version of sundae_bar is a significant step towards that.
Instead of asking customers to navigate a marketplace and figure out which AI they need, they can now start with the work itself.
Tell sundae_bar what you need to get done.
The platform can find an existing agent for the job or build one for you using the Skills already available through sundae_bar. It checks that those Skills will work in the agent, adapts them where necessary and quality-checks the finished agent before it is created.
If the right Skill doesn't exist, sundae_bar can create one.
That changes what our growing library actually means.
Today, there are 700+ AI Skills, Agents and other capabilities available through the marketplace. But the opportunity isn't simply to build the biggest catalogue of AI.
It's to make all of that capability useful.
And importantly, this is a commercial product.
Our customers pay to use sundae_bar because we're removing the work that sits between knowing AI could help and actually getting value from it.
A sundae_bar subscription gives a customer access to ready-made agents for recurring business work, the ability to have new agents built around their own requirements, and a platform where those agents can remember relevant context and run on a schedule.
We have nine ready-made agents live today, covering work including competitive research, executive briefings, meeting preparation, account intelligence, regulatory monitoring and vendor due diligence.
A customer can launch one immediately. Or they can simply describe the job they need done and let sundae_bar build around that requirement.
And once an agent is working, it doesn't need to sit waiting for another prompt. Recurring work can be scheduled. A competitive brief you need every Monday morning can be there every Monday morning.
For me, three things make this model particularly powerful.
First, better Skills.
Through our Agent Lab, developers compete against each other to build Skills for specific business requirements. We have now completed 25 challenges, attracting over 6000 competing submissions. The winning Skills are measured against defined criteria, and winning Skills are already being used inside the ready-made agents customers can launch today.
Second, better Agents.
sundae_bar can take those individual capabilities and put them together around the work a customer actually needs to do. It can find the relevant Skills, check whether they fit, adapt them where necessary and build an agent around the job.
That means every new Skill doesn't just become another item in a marketplace. It becomes another capability sundae_bar can draw on when building for customers.
Third, trust.
One of the biggest barriers to businesses putting AI into real workflows is knowing whether they can rely on what it produces.
We design quality checks into the product rather than asking customers to simply trust that the AI will work.
Skills are checked before being attached to an agent. If one doesn't fit the environment, sundae_bar can adapt it or refuse to use it. Every newly created agent must also pass a quality check before it is created. If it doesn't meet that standard, the customer doesn't get the weak version.
Better Skills. Better Agents. More trust in putting them to work.
That is what we are building a subscription business around.
And behind it, the development engine continues to expand what the platform can offer.
The connection is increasingly straightforward.
A business has a need.
That need can inform what we build next.
Developers can compete to build the capability.
The strongest Skills enter our library.
sundae_bar can then put those Skills to work for paying customers.
And because developers can work on different problems in parallel, we can develop and test new capabilities at speed and scale.
This is the model we have been working towards.
The last phase was about connecting SN121 to the commercial platform and creating the engine for continuous development.
This phase is about turning that engine into an increasingly useful commercial product.
More capabilities are being built.
sundae_bar can do more with them.
And the experience for the customer is becoming simpler, not more complicated.
There is a lot of technology underneath all of this. There should be.
But our customers shouldn't need to understand it.
They should be able to tell sundae_bar what they need to get done and trust us to figure out the right AI to do it.