Trajectory is building a continual learning platform to close AI agents' 'experience gap'
“These models are smarter and smarter but always feels like when you're talking to them, it's their first day on the job.”
continual-learning
11 tracked signals on continual-learning.
LangSmith Engine and Context Hub turn agent traces into durable, versioned long-term memory for continual learning.
“Most agents don't learn, they just leave traces.”
Engram, a new lab from Dan Biderman and Jessy Lin, treats memory and continual learning as baking new context directly into model weights.
“We don't see the world through the lens of pre-training or post-training. Our models are always training.”
Rich Sutton argues continual learning is just learning — the field's framing is the anomaly
“I'm not weird. The field is weird.”
LangChain frames agent improvement as a data mining problem over production trace data
“there's a very tight coupling between what observability is and what continual learning is”
AI agents excel at coding but fail broadly due to lacking expertise, not intelligence
“why we are so successful at the coding agents, but they're so terrible at anything else”
Startup Engram frames 'scaling compute on context' as AI's path to deep domain expertise
“the thing that we're after like this idea of scaling compute on context is the pursuit of depth in AI”
Continual learning lets AI agents update their own prompts, skills, and tools over time from feedback.
“continual learning is when you give the agent the ability to update itself over time, say it's prompts, subagents or skills”
No good standards exist for agent development; real-world continual learning is largely absent.
“there aren't really good standards for these things, right? Like there there's not some one-size-fits-all uh solution”
Applied Compute is bringing continual learning to enterprises via a distillation spectrum from offline to online
“this is sort of the holy grail of continual learning where I have a model that's serving production traffic, it does a rollout, it creates a trace, we figure out how to learn from that trace, we update the model”
AI research frontier access requires an 'unreasonably narrow path' that excludes most talent
“the unreasonably narrow path”