Jesse Han says personal AI software engineers will learn a developer's preferences and codebase, then support work from specification through deployment.
2
Morph Code Index combines static analysis, vector embeddings, graph algorithms, and a neurosymbolic query language to give coding assistants broad codebase context.
3
Morph Labs uses the index to generate training data for Rift Coder 7B, an open-source code-editing model that runs on a laptop and in an IDE.
Summary
Jesse Han presents Morph Labs' plan for personal AI software engineers that understand a developer's habits, preferred idioms, and codebase. He expects these systems to improve with the code and support the full software development life cycle, from ideation and specification through deployment. The first infrastructure product is Morph Code Index, a neurosymbolic code database that supports natural-language search, precise queries, code history, and retrieval across an entire codebase. Morph Labs builds the index with static analysis, vector embeddings, and graph algorithms, then uses it to turn codebases into training data. Han says the same pipeline trained Rift Coder 7B, an open-source code-editing model that runs locally on a laptop or inside an IDE. Morph Labs plans to make its index open-source, transparent, and free, and says it will announce a platform for building and deploying personal AI software engineers.
Personal AI software engineers will learn each developer's code and habits
Jesse Han describes a personal AI software engineer that understands its user's preferred practices and coding idioms. It should know the user's codebase well, keep its knowledge current, and change as both the developer and the code change. Han frames this as a system that works alongside a person throughout software development rather than only generating code in an editor. Its planned scope covers ideation, specification, implementation, testing, and deployment. The product goal at Morph Labs is to make this kind of personal assistant available to everyone.
Morph Code Index gives coding assistants a structured view of a codebase
Han introduces Morph Code Index as a neurosymbolic code database for developers and their coding assistants. It supports semantic searches over a codebase using natural language, while also making code and its history visible to the assistant. The index combines static analysis, vector embeddings, and graph algorithms to find relevant code. Morph Labs also includes a neurosymbolic query language for cases where the user needs precise results. The product is intended to provide codebase-wide context through retrieval, rather than limiting an assistant to the files currently open.
Python bindings are meant to connect the index to different assistants
Morph Code Index includes Python bindings that let developers attach an index to a coding assistant. Han says this gives the assistant context from across the codebase through retrieval. The design makes the index an infrastructure layer that can sit underneath different coding tools, instead of tying the code search system to one assistant. Alongside natural-language semantic search, the neurosymbolic query language gives developers a way to request exact results when broad similarity search is insufficient. Han presents both capabilities as parts of the same code understanding system.
The indexed codebase can become training data for a personal model
Han says Morph Code Index can traverse a codebase and generate large amounts of training data. Morph Labs has used that pipeline to train Rift Coder 7B, which he describes as an open-source model for code editing. The model can run on a laptop and in an IDE. His proposed workflow is direct: index the code, run the generation script, and begin training a personal AI software engineer. The index therefore supports two different uses in the talk, giving an existing assistant more context and helping create a model adapted to a particular codebase.
Morph Labs wants the infrastructure to remain open and free
Morph Labs says the Morph Code Index will be open-source, transparent, and free. Han links that choice to the idea that software's future should belong to everyone. He also describes a planned platform for building, managing, and deploying personal AI software engineers. The platform had not yet been announced at the time of the talk, but attendees were invited to join a waitlist for early access. The broader product direction combines an open code index with tools for creating and operating assistants that learn from a developer's own code.
"Finally the Morph Code Index can traverse your codebase to generate vast quantities of training data."02:31
Who should watch
You are building coding assistants and need codebase-wide retrieval rather than context from a few open files.
You want to train a code model on a private codebase and are interested in an indexing pipeline for generating training data.
You are evaluating open-source infrastructure for personal software agents that can support work beyond code completion.
