BYK

Looks like after the OpenAI deal went south, Microsoft wants to "train" their own LLM by distilling the interactions with the other models. This may be a genuine desire for RL maybe but definitely not applicable for individual developers or even small shops. "Learning" can definitely happen in token space.

There are a lot of external LLM/AI memory managers out there and they use the most common denominator to expose the "memories": MCP

A typical setup is: a harness plugin that reads the conversations after they finish, create "memories" from them, then hope the next session uses that MCP tool to "remember" stuff. Where this fails is when you expect your agent to finally "learn" how you work. Even having to say "use your memory to see how we do releases" is friction, AND eventually, you'll forget to say that. Ideal world, iff an agent is going to be your colleague, you'd just say "cut a new release", and it would either know how to do it or would know where/how to get that information and follow that through.

One may argue that these instructions should be codified in `AGENTS.md` files but these are like your developer docs: always out of date (slightly if you are lucky and a lot if you are typical like myself).

With Lore, I'm trying to fix this by automatically injecting relevant preferences, gotchas, and "memories" into the context at multiple different points. Since we are an LLM gateway sitting between the harness and the LLM provider, we have full control over the context, and this is part of active context management.

I've started building Lore because none of the existing "memory" solutions solved my real problem: my agent stopped every 5-10 minutes to "compact" and forget every important thing we covered before these compaction stages.

I hate repeating myself and I was a very strong AI-skeptic so I almost gave up until I came by Mastra's Observational Memory post. Soon after I saw Sanity's Nuum and I knew I had to try porting this to OpenCode.

Lore is the evolved version of this: it is harness-agnostic, works with OpenAI and Anthropic backends and I added Vertex and Bedrock support just recently (hopefully works? :D).

Would love to hear your thoughts about the memory craze of now and using solely MCP-based solutions while ignoring the context management in the active sessions and how we ended up accepting this primitive and terrible solution as our daily driver :D

Lore automatically curates the project knowledge that makes your AI coding agent smarter every session

How we built a secure, auditable, and low-friction release system at Sentry that is resistant to supply-chain attacks like Shai-Hulud

Well I actually did as I was literally hired for that and worked on it for almost 3 years :)

You mean things like these: https://docs.sentry.io/product/relay/#pii-data-scrubbing, https://docs.sentry.io/product/data-management-settings/scru...

This now has been rectified. Apologies for the misrepresentation, wasn't intentional.

Apologies for the issue. There's an issue file on GitHub for this and we are working on a fix: https://github.com/yarnpkg/yarn/issues/4339

See https://github.com/yarnpkg/yarn/issues/4320

We have patched that issue and already tagged 1.0.1 ;)

True that. I reread your comment and realized I misinterpreted it.

Skepticism is good :)

Thanks for calling my honest reply a lie and me a liar, really appreciate it.

I'm @madbyk on Twitter and you can also Google my full name to catch my other lies and bad acting on some of my recorded talks.

I have this for starters: https://gist.github.com/BYK/66dfd985491b483bd140

And that is of course not true, probably just a coincidence.

We may when the service actually starts to cost us =) Thanks ;)

Added, thanks for the suggestion! =)

This is a pet project and we gain nothing but experience. We were not aware of any similar projects when starting. In terms of capitalizing, please keep in mind that there are no ads and we do not intend to add any so it will cost us in the contrary when more people start to use this.

It's already fixed. Thanks for reporting! =)

Javascript version is on the horizon. We used iframe because it was quick&dirty and "good enough" to start with =)