wooders

Letta Code - It’s a much more coworker-like experience because it can learn, but also performs very well for coding, and the harness can be extended like pi

(disclaimer: I work on Letta Code)

I think the problem with ChatGPT / other RAG-based memory solutions is that it's not possible to collaborate with the agent on what it's memory should look like - so it makes sense that its much easier to just have a stateless system and message queue, to avoid mysterious pollution. But Letta's memory management is primarily text/files based so very transparent and controllable.

An example of how this kind of memory can help is learned skills https://www.letta.com/blog/skill-learning - if your agent takes the time to reflect/learn from experience and create a skill, that skills is much more effective at making it better next time than just putting the raw trajectory into context.

FYI the LOCOMO benchmarking done by Mem0 was very sus so I wouldn't recommend relying on those numbers for anything

https://www.reddit.com/r/LocalLLaMA/comments/1mon8it/woah_le...

https://www.reddit.com/r/LangChain/comments/1kg5qas/lies_dam...

I think the "memory blocks" are essentially what you are describing - to have an infinite session (which systems like Letta is designed for) you have to have a mechanism for organizing the important information and persisting it for future interactions. This organization can be done via tool calls (which was what MemGPT did) or done by other agents in the background. While the message buffer is continues to grow / old messages get evicted, the memory blocks are fixed size and always in context.

Agents are already usually deployed as an API service. You can have "agent-to-agent" communication by having agents call each others APIs. I don't understand what this protocol is for.

MCP actually fills a gap since people don't normally expose things like writing to their local filesystem as a callable API.

Stealth AI company (based on MemGPT - memory for LLMs) | Founding Staff SWE | San Francisco | ONSITE | Full-time | $220k–$300k + Equity

We’re a team of UC Berkeley PhD grads starting a company (currently in stealth) around the MemGPT project (10k+ GitHub stars). The founding team comes from the same research lab and PhD advisors at Berkeley that produced Spark (→ Databricks) and Ray (→ Anyscale). We have deep expertise in both AI and systems, are currently hiring a founding team of exceptional engineers to join us in building the next generation of LLM agent technology.

We are seeking a Staff or Senior level SWE for development of both our OSS and AI agent platform (mostly Python). Also open to strong, less senior candidates.

See full job descriptions: https://jobs.ashbyhq.com/memgpt

Highly recommend https://www.amazon.com/Triangle-Selling-Sales-Fundamentals-G... - as an engineer and first-time founder(Glisten AI - YC W20) it gave me a great framework to approach sales calls with.

We're a company coming out of the YC W20 batch working on the product attribution problem http://glisten.ai/.

There's too many products nowadays to be manually attributed (e.g. pattern=stripes), making it hard return good results even with entity resolution for queries. We train classifiers to categorize products, including what something is not, using their images and descriptions.

I don't think it's simple to deploy scaleable predictions - that's why model hosting solutions like SageMaker's and GCP's AI Platform exist, and there's no need for people to be re-implementing model deployment/monitering.

Thank you so much! I feel the exact same way and have always been so frustrated by how broken product search on online marketplaces is. Hope Allparel can help you, and let me know if you have any feedback :)

Yes, sorry! Only having womens' clothing make the image classification a lot easier.

Yep! Image classifiers are used to generate tags from the product image to help match products to searches. Right now there's about 30 tags that are being labelled, but in the future there'll be around a couple hundred tags being identified.