Article URL: https://beza1e1.tuxen.de/ai_feelings.html Comments URL: https://news.ycombinator.com/item?id=49587128 Points: 12 # Comments: 0
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
A long-standing race condition in database applications allows concurrent transactions to read the same value simultaneously, causing lost updates while the audit log falsely appears consistent. In tests simulating ten workers each withdrawing from a shared account, between four and six transactions were silently lost per run, yet the audit table showed no anomalies. PostgreSQL 18 introduces old and new row references directly within a RETURNING clause, allowing the read, write, and audit insert to occur as a single atomic statement. This eliminates the gap between reading and writing a value that competing transactions could previously exploit. The change means audit logs now reflect what a query actually overwrote, not what the application last read, making silent data corruption detectable by design.
A fine-tuned LoRA adapter requires a deliberate deployment strategy before it becomes a usable product. FineTune Studio supports three inference paths — local inference, a vLLM server, and a Hugging Face Space — each suited to different use cases ranging from demos to production traffic. Merging adapter weights into the base model simplifies serving but is optional when using vLLM, which supports LoRA adapters natively. Before declaring a model deployed, engineers should verify that the chat template, stop tokens, precision settings, and throughput under realistic concurrency all match training conditions. The article also stresses versioning the adapter, dataset, and evaluation results together to track changes over time.
Article URL: https://kaeruct.github.io/posts/2026/09/06/conquering-entropy-cultivating-trust/ Comments URL: https://news.ycombinator.com/item?id=49587833 Points: 3 # Comments: 1
A developer tutorial published on DEV Community outlines a two-sided authentication flow for Telegram Mini Apps built with React and a PHP backend. The React app reads initData from the Telegram Web App SDK and sends it as a custom header with every API request. The PHP server then validates the header using HMAC-SHA-256 with the Bot Token as the key, rejecting forged or expired data before issuing a short-lived JWT tied to the verified Telegram user ID. Protected API endpoints subsequently verify the JWT and confirm that the telegram_id claim matches the expected user for each resource. The approach is designed to defend against three main threats: replayed initData, forged user IDs, and JWT reuse across different users.
Discussion (0)
Log in to join the discussion and vote.
Log in