DDmytro Nasyrov·10h agoCasino Wallet Recovery: Keep Session, Round and Ledger State SeparateA three-domain ownership map and fictional late-event sequence for assigning recovery evidence.17 min read·pharos-production.hashnode.dev
TTelcoEdge Inc.·23h agoBefore the Bill Exists: How Telecom Mediation Pipelines Turn Raw Usage Into Reliable DataUnderstand how telecom mediation systems ingest, normalize, validate, deduplicate, and process network usage records before they reach billing systems9 min read·telcoedgeinc.hashnode.dev
RRONI DAS·1d agoHow Canada's new Real-Time Rail moves one payment, and what it teaches about backend designPayments Canada switched on a 24/7 system on 9 November 2026 that settles every payment between banks in real time. Here is how one payment travels, why the fraud checks come first, and why a bank asks before it retries.7 min read·systemdesignacademy.hashnode.dev
SSakshi Tyagi·1d agoWhat Should an Agent Remember?Agents That Survive Production · Part 3 of 4 Separate useful context from the durable facts needed to resume work.7 min read·sakshityagi.hashnode.dev
Mmihalykrich·1d agoBuilding KRXY — Part 8: When the Network Slows Down — and the Model Library Grows1–2 November 2026: two-Pi learning through bandwidth limits and a two-minute outage, followed by the first Qwen3.5 2B and 4B installations.13 min read·krxy.hashnode.dev
Mmihalykrich·1d agoBuilding KRXY — Part 7: From Two AI Nodes to Distributed Learning30–31 October 2026: safer node recovery, more reliable serving, and the first measured two-Pi learning and rejoin experiments.12 min read·krxy.hashnode.dev
TTony Dzi (Anton Dziatkovskii)·3d agoCoordination Decay: 103 Days of a Seven-Machine LLM Agent FleetA coordination protocol can keep its checked safety rules and still fall out of use. That is the main thing we learned from running LLM coding agents on seven machines for 103 days, and it is why we a7 min read·tonydzi.hashnode.dev
HHussain Fakhruddin·3d agoRecurring jobs for MENA SaaS: missed runs and catch-upDesign recurring work with durable occurrence identities, explicit missed-run policies and controlled recovery for Saudi and UAE deployments.9 min read·sultanbyte.com
AAmitabh·3d agoFrom One Server to Many: What Each Scaling Step Fixes, and What It BreaksMost system design answers jump straight to the final architecture. Interviewers are more interested in the path: what broke at each stage, which change fixed it, and what that change made harder. Her3 min read·prepd.hashnode.dev
RRONI DAS·3d agoHow GitHub rebuilt Git storage for 5,000 agents on one repoGitHub split Git into immutable packs in Azure Blob Storage and a tiny compare-and-swap log for refs. In testing it took 7,200 pushes a second on one repository. Here is how, in plain words.8 min read·systemdesignacademy.hashnode.dev
