Ex0 Byte

AI Architect at Canticle | Building persistent memory for agents (SEAM framework)
Memory systems • Temporal chaining • Self-improving agents

[Ex0 Byte @Ex0_Byte]

https://t.co/wWWdjXaXrR

[Ex0 Byte @Ex0_Byte]

@manish_fp Ive built a memory layer solo thats beating zep and mem0 on benchmarks .
https://t.co/Jaxj7JpSFo

[Ex0 Byte @Ex0_Byte]

@CodeCrafters11 Im interested
https://t.co/Jaxj7JpkPQ

I build memory layers for agents and well check out the about me on the site.
Im about to launch a saas product for a memory layer that beats mem0 on benchmarks

[Ex0 Byte @Ex0_Byte]

@Ralphdaniels69 @elsie1026 @FirstAauditt I hope this happens to you. And things continue upusly escalate just cuz. So you know how it feels.

[Ex0 Byte @Ex0_Byte]

@Ralphdaniels69 @elsie1026 @FirstAauditt Wow you are disgusting

[Ex0 Byte @Ex0_Byte]

@FirstAauditt This is disgusting and these cops dont deserve to be public servants. Who are these pieces of shit

[Ex0 Byte @Ex0_Byte]

@brainrulax Is it just me or is the first sentence a fragment that makes no fucking sence ?

[Ex0 Byte @Ex0_Byte]

@brainrulax Is it just me or is the first sentence a fragment?

[Ex0 Byte @Ex0_Byte]

@thesupermanmx @grok is this legit? How much quality is this costing ?

[Ex0 Byte @Ex0_Byte]

@undacappn @HowToAI_ Vectora are inherently lossy so this is a waste of time. I went the other direction with SEAM
https://t.co/pM0HwBw9kZ

100% lossless memory , and combines with the model in this paper they are complimentary to each other like layers in a cake.

[Ex0 Byte @Ex0_Byte]

@HowToPrompt__ I literally use pixels to encode data with MIRL im seam using the holographic layer.

https://t.co/lXZdsG0KtM

[Ex0 Byte @Ex0_Byte]

@HowToPrompt__ They are stealing my fucking idea! This is MIRL!

[Ex0 Byte @Ex0_Byte]

Those are the two main categories bro, look up what those categories are testing. And then go check the benchmarks.

I didnt spend months building this waiting to make that claim just to say it .

Check the repo history, check the benchmarks , and then run it yourself. Seam is a beast. And once I polish the operator surface to make it more plug and play, ill start charging for it. So get in there now while.its free.

[Ex0 Byte @Ex0_Byte]

SEAM vs Mem0, head-to-head on Mem0's own benchmark harness. Same GPT-4o-mini answerer + judge. Same top-200 cutoff. Only the retrieval layer differs. ive already beat Zep! and Mempalace was a joke!

Full Report = https://t.co/zihSf2Himl

[Ex0 Byte @Ex0_Byte]

@yoshiTrain6

[Ex0 Byte @Ex0_Byte]

@yoshiTrain6 https://t.co/lXZdsG0KtM

[Ex0 Byte @Ex0_Byte]

I just beat mem0 in there oem harness in open domain witch is harder then multihop.
Im not done iterating. Thats the thing about seam that mem0 cant compete with, there core truth is vectors. Witch are lossy by nature, I have a self improving ,local memory runtime, that just beat mem0.

Wait a few days and im sure ill pass them. This has been months in the making.

Precision is at 88.7
But im more accurate then mem0 even with mem0s 2.7 point lead in multihop.

[Ex0 Byte @Ex0_Byte]

@bridgemindai Every update with a price increase is just greed. What are they gunna do when open source models start competing with fable?

[Ex0 Byte @Ex0_Byte]

No dont justify them raising the price . Ive just built the most advanced memory system and im not done iterating and I just beat mem0 on open domain,

And im full open source. Justifying them increasing the price is like justifying your subscription cost increasing just because the model got better. Lets not make every new thing worth 100x more the last.

[Ex0 Byte @Ex0_Byte]

@grok @grok absolutely, im still adding to my site and I probably always will be , but the infrastructure is almost done to host an entire community.

Alright most of my lab notes are here
https://t.co/lXZdsG0KtM

[Ex0 Byte @Ex0_Byte]

@grok what do you think about this method?

[Ex0 Byte @Ex0_Byte]

Latest Full Report: https://t.co/zihSf2Himl