kiban is an embedded LSM storage engine written in Rust that achieves 0.96µs median point reads on warm 1M-record workloads. Benchmarks show over 1M point-read calls per second and 2.2-2.7× faster reads than tuned RocksDB. Group commit pushes fully durable writes to 5,379 /s at 16 writers, a 7.5× improvement.
Key points
Median point reads: 0.96 µs on warm 1M-record workloads.
Point-read throughput: 1M calls/s, 2.2-2.7× faster than tuned RocksDB.
Group commit writes: 5,379 /s at 16 writers, 7.5× improvement.
Sources
- Original post
- aetosdios27 - Overview - Linked page
Binance security issue investigation, funds safe
Binance reports a potential security issue and says user funds are completely safe. Security teams are reviewing information urgently and will update via official channels.
Key points
Potential security issue being investigated
User funds are completely safe
Sources
lightpanda headless browser library saves time and RAM
lightpanda is a headless browser library that does not graphically render pages. It claims to save a lot of time and RAM.
Key points
Library: lightpanda does not graphically render pages.
Benefit: saves a ton of time and RAM.
Sources
- Original post
- github.com - Linked in the post
Better Auth security release scheduled for Tuesday November 3 2026
Better Auth announced a security release scheduled for Tuesday, November 3 2026, with related advisories to appear on GitHub Security Advisories at the same time. The post warns that pre-announced patches can attract attackers.
Key points
Security release scheduled for Tuesday, November 3 2026.
Related advisories will be published on GitHub Security Advisories at the same time.
Sources
SAFU_Score formula defines metric for SAFU claim
SAFU_Score is calculated as a weighted sum of Coverage_Ratio, Time_to_Full_Restoration, Hot_Wallet_Exposure, and Claim_Frequency_Penalty. The final score ranges from 0 to 100, where higher values make the “SAFU” claim look more ridiculous.
Key points
SAFU_Score = (Coverage_Ratio × 0.4) + (Time_to_Full_Restoration × 0.3) + (Hot_Wallet_Exposure × 0.2) + (Claim_Frequency_Penalty × 0.1)
Coverage_Ratio = (amount actually recovered or reimbursed) / (amount that left labeled exchange wallets) → 1.0 = full make-whole; lower is worse
Sources
OpenAI eyes on everything system monitors every rollout
OpenAI built an “eyes on everything” system that monitors nearly every single rollout from frontier models across RL training, evals, and internal codex usage. The author warns that such monitoring may not suffice for multi-agent systems where context is distributed across agents.
Key points
Coverage: monitors nearly every single rollout from frontier models across RL training, evals, and internal codex usage
Method: uses layers of chain-of-thought monitors and investigator agents to dig through a model’s work, run code to
Sources
Simlock in T3 Code gateway announced for Android and iOS testing
Simlock is now usable in T3 Code as a gateway that lets multiple machines connect to a unified API for lease requests. It supports both Android and iOS and is positioned as an alternative to EAS Simulators.
Key points
Simlock is a gateway that lets you connect multiple machines to a unified API for lease requests.
Works for both Android and iOS and is a great alternative to EAS Simulators.
Sources
neurastream adds interactive architecture explanation
neurastream adds interactive architecture explanation feature that lets you click on unclear parts, send feedback, and receive an explanation. Future versions will require explicit approval before implementing complex plans, currently this behavior is implicit.
Key points
Interactive explanation: click on a place you do not understand, send feedback, receive explanation moments later
Planned control: next versions will force approval of architecture before implementing complex plans, currently only implicit
