Palantir DevCon2 Recap
Consumer Applications on the Horizon!
Palantir’s developer conference, DevCon2, took place in Palo Alto on February 26th and 27th and showcased many new capabilities that are driving Foundry application to the edge. Palantir is clearly expanding the horizon of enterprise AI and eyeing consumer applications. My DevCon team was tasked with building a consumer product and chose to create a personal assistant, Vicki, who can send emails and Slack messages, assign tasks, and schedule meetings (video demo at the end of this article). Below are the most significant announcements from the conference and my thoughts on the 2025 outlook for Foundry developers.
OSDK Media Set Support
The OSDK now supports uploading media to Media Sets. This allows developers to not only upload videos, audio files, images, and documents in the applications but also chain AIP Automate and Logics together to create dynamic workflows incorporating media and LLMs. At the last DevCon, my team achieved this by reverse engineering the API calls made to Foundry when uploading files to Media Sets. This was not fun. It’s nice to see first-class support for Media Sets in the OSDK. Using Media Sets with the Ontology can unlock transcription workflows and other unstructured inputs, feeding AI applications hosted in Foundry.
OSDK Intelligent Object Set Watchers
This was my favorite feature announced at DevCon. The OSDK team gets an A+ again for their continued focus on the features developers must have to build best-in-class applications on Foundry. With Intelligent Object Set Watchers, you can create server authoritative applications in Foundry that can synchronize many occasionally connected clients. For developers familiar with pub/sub, this will immediately be an obvious value unluck. Applications that I am building involve state machines constructed dynamically by LLMs. These state machines power everything from insurance claims denials to personal assistants. These workflows require multiple actors and systems, which can now be kept in sync with state changes thanks to Object Set Watchers. This feature is in a private beta, which I hope to opt into in the coming days.
OSDK Derived Properties
When fetching data from the Ontology using the OSDK, you traditionally have had an over-fetching problem that degrades application performance, especially in consumer applications that demand sub-second page load times. Derived properties allow you to push aggregations to the server, avoiding over-fetching on the client.
Embedded Ontologies
Working offline has previously not been possible in Foundry without custom solutions. Palantir announced Embedded Ontologies at DevCon2, allowing devices such as mobile phones, controllers, and other devices to work without internet connectivity using a local copy of the Ontology. This can unlock monopoly use cases (big share of a small market) like those in mining and home construction where networks are unreliable or non-existent. More importantly, this could power consumer use cases that require features common in Progressive Web Applications (PWA) that allow users to continue using applications when internet connections are lost or to precache critical assets to improve performance. This feature is also in private beta.
Eventing
Palantir announced a new data connection type to support event-based workflows and automation for the Ontology. This new connector supports AWS SNS, GCP Pub/Sub, Microsoft Bot Framework, and other event sources. Combined with AIP Automate, you can trigger automation such as AIP Logics and functions to respond to incoming events.
Open Data Formats and Push Down Compute
Palantir announced support for using push-down compute operations in supported data warehouses such as BigQuery, Redshift, and Snowflake. This is a huge step forward to answering the age-old question of “Where does Palantir fit in my stack?”. Customers can fully leverage their existing data warehouses to model their data assets and ontology in Foundry without data replication or additional computing.
2025 Developer Outlook
Palantir is quickly absorbing feedback from its enterprise customers and early adopters, which is translated into valuable features for developers. I expect this pace to accelerate as the teams are pushing changes at an obscene pace. I expect developers to feel the firehouse effect, meaning that thing you are drinking from, as this cycle continues to play out. Developers need to focus on the parts of Foundry that deliver the most value creation for their respective organizations while continuing to dedicate more time to Foundry education.
Get Dorian Smiley’s stories in your inbox
Join Medium for free to get updates from this writer.
There is a risk of adopting and investing in features that will be deprecated. Palantir values product evolution over backward compatibility. Many features I have adopted, like AIP Orchestrators, become deprecated without a replacement as the product teams work to evolve the offerings. Be sure you subscribe to product updates and read the in-platform announcements when you launch applications. Document and build roadmaps around deprecated features as they are announced.
The reference architectures for rebuilding enterprise and consumer products with AI at the center do not exist yet. This is a very exciting time to leverage AIP/Foundry to invent them. Developers have a unique window of opportunity to be the first to create breakthrough use cases and killer apps that will become the reference architectures of tomorrow. Palantir is unique in its ability to keep pace with innovation and, in large part, drive the adoption of the most advanced models and techniques. Becoming a Foundry developer will reward the most risk-on entrepreneurs, and enterprise thought leaders looking to build what is next in software. If that is, you need to get on this train.
