Agent memory is still an unsolved problem.
Most work on memory falls into one of three buckets:
1. RAG against a static corpus
2. Agentic RAG - a tool call to retrieve from dynamic memory
3. Better indexing for either of the above
But none of these are how human memory actually works. Human memory isn't a slow-moving corpus. It's fast-changing. And critically, we don't *fetch* memories by posing carefully curated queries. Subconscious association surfaces the right context at the right time, automatically.
Imagine a coworker who acknowledges something you said, then forgets it 30 seconds later. It'd feel broken. Yet that's exactly how most agents behave today.
The hard problems few people are talking about: fast-changing memory that stays coherent, and associative retrieval that doesn't require the agent to "know what to look for."
This is what we're working on at Run. If these problems interest you, we're hiring.
we're working across a handful of industries already at @runcomputing, but one of my favorites so far has been legacy manufacturing firms.
there's nothing like deploying 24/7 ai coworkers into the most traditional environments and watching the "oh shit" moment when they delegate real work to a background agent for the first time.
we’ve been heads-down building a new platform for delegating repetitive work at @runcomputing.
not another workflow builder. no infinite canvas. no nodes. just describe your process in natural language.
we’re starting to roll out to more companies now.
if you have a process that is
- highly variable with lots of edge cases
- multiplayer and needs humans involved
- long running across days or months
- high enough volume where you wish it were automated
please dm. we’d love to see if we can help.