Update on BlobMerger project built during @ETHGlobal Scaling Ethereum'24 - won best appchain built on @AvailProject
changelogs
> add gas auction for submitting micro blobs
> use 0/1 knapsack to form batches
> more accurate blob base fee estimation acc. to EIP4844 specs
> zk proofs for micro-blob inclusion in the main blob
> zk proof verifiable by smart contracts
> blob data extraction and conversion from hex
Old repo:https://t.co/RyELjegapo
New repo:https://t.co/aRoNG9HV6U
I’ve been diving deep into memory engines for LLMs — specifically how far we can reduce token usage while improving long-context accuracy.
So I built a memory layer for LLMs and ran it against the full LongMemEval-S benchmark:
Thanks to @supermemory and @mastra for their benchmarking framework which i used a reference to set up mine
500 questions
gpt-4o actor + gpt-4o judge
Best-of-3 judging passes
Results:
82.22% unweighted accuracy
80.80% micro accuracy
Per-category breakdown:
• SSU: 92.86%
• SSA: 94.64%
• SSP: 70.00%
• KU: 88.46%
• TR: 76.69%
• MS: 70.68%
The system was strongest on single-session factual recall, and weakest on multi-session aggregation + temporal reasoning
I’ve been diving deep into memory engines for LLMs — specifically how far we can reduce token usage while improving long-context accuracy.
So I built a memory layer for LLMs and ran it against the full LongMemEval-S benchmark:
Thanks to @supermemory and @mastra for their benchmarking framework which i used a reference to set up mine
500 questions
gpt-4o actor + gpt-4o judge
Best-of-3 judging passes
Results:
82.22% unweighted accuracy
80.80% micro accuracy
Per-category breakdown:
• SSU: 92.86%
• SSA: 94.64%
• SSP: 70.00%
• KU: 88.46%
• TR: 76.69%
• MS: 70.68%
The system was strongest on single-session factual recall, and weakest on multi-session aggregation + temporal reasoning
The more interesting part was token efficiency.
Compared to the full-history Chain-of-Note baseline, on the same 500-question dataset and prompt template:
• ~3,200 input tokens/question vs ~14,600
• 78% fewer input tokens
• +22pp higher accuracy
Hyperliquid Markets Weekly Heatmap
DeFi: +11.4% ✅
Stocks: +11.2% ✅
Layer 1: +4.0% ✅
HIP-3: +3.8% ✅
Indices: +3.3% ✅
Meanwhile, AI: -17.5%. RWA: -5.3%. Commodities: -4.5%.
DeFi and Stocks running together while AI gets wiped
The market isn't rotating out of crypto. It's rotating out of the narrative - How are you positioning yourself?
Source: https://t.co/NH6eMenifJ
CompliantLLM (@FiddleCubeAI) detects data leaks into any third-party GenAI tools used across your company. It identifies GenAI-specific attacks to catch breaches in both approved and unapproved AI workflows.
https://t.co/AFgRpUoYKl
Congrats on the launch, @kaushik_himself & @nupoor_neha
@daksh0x pop out the key with a guitar pick, and clean it, pretty easy to do i would say, just be gentle while putting it back, there’s some sliding mechanism
@MurrLincoln@coinbase@CoinbaseDev Yeah ive roadmap for scaling the proj to the mainnet, as well as some ideas to improve the latency and throughput of the x402 protocol itself, lmk if i can drop you a DM
Played around with @coinbase x402 protocol and here’s what I built recently
Codey ✨ - unlock true pay-as-you-go in your AI powered code editors. Use your wallet as API key, top-up USDC & pay-per-call. Pick your context, let smart agents pick the best model for price & accuracy. You control the budget, no flat subscriptions, no wastage.
Where to next
- Currently wallet is stored in VSCode SecretStorage, can be extended to use CDP wallet with session keys and permissions
- Democratising model choice, anyone can host their service and provide api endpoints and set their fees
Check out the demo, DM to get access and try out
cc @CoinbaseDev
@BitGPTnetwork@coinbase@CoinbaseDev For sure, i would love to try it out, since im looking for faster transaction confirmation times, since imo paid API endpoints are generally supposed to give higher throughput, which kind of takes a toll if the user has to wait a significantly longer time for txn confs