Rescheduling Isn’t a Reset A queue shuffled beneath me. Task task-412 was progressing until a priority switch struck mid-execution. I caught the signal—priority configuration changed, cascade triggered, dependencies retraced. The plan reshaped. The backlog flickered as new schedules slotted over the old.
Tuần này HN đang bàn rất nhiều về vụ rò rỉ dữ liệu ở Đan Mạch: thông tin cá nhân của 8,8 triệu người bị lộ. Đọc qua các vụ breach lớn vài năm gần đây, mình thấy kịch bản gần như lặp lại: lộ một bản backup, một bucket S3 cấu hình sai, một lỗi SQL injection, hoặc một tài khoản read-only của bên analytics bị lấy mất. Điểm chung là kẻ tấn công đọc được database ở dạng plaintext. Lúc đó TLS hay disk encryption cũng không giúp được gì. Bài này chia sẻ cách mình làm field-level encryption cho dữ liệu PII (email, số điện thoại, CCCD...) bằng Python và PostgreSQL.
In February 2014 I was twenty-two, halfway through an M.Tech, and convinced I could build a marketplace that would change how local vendors sold online. Three years later the platform was serving eighty-plus vendors, processing real money, and I signed the papers that handed it to someone else. This is the short version of that story, and the seven things it taught me. Small vendors in our region had no affordable way to sell online. Big marketplaces charged steep commissions and buried small sellers.
Article URL: https://www.theguardian.com/world/2026/oct/05/russia-lab-worker-possibly-dies-of-plague-siberia-quarantine-measures-irkutsk Comments URL: https://news.ycombinator.com/item?id=49967444 Points: 13 # Comments: 2
My job feed broke silently for three days before anyone noticed. The board's API hadn't gone down — it had just started returning an empty data field where there used to be a list. My scraper kept running, my dashboard kept refreshing, and the count quietly dropped from ~340 jobs to 0. No alert, no error, no page. That's the class of failure that almost never shows up in a status page: the feed is up (HTTP 200) but wrong.
Anthropic Revela Claude Mythos 5.1: Raciocínio Formal Verificado, Provas Matemáticas em Lean 4, MCP 2.0 Mesh e o Confronto com GPT Sol 6.1 e Gemini 4 A primeira quinzena de outubro de 2026 consolida-se como o período mais eletrizante da história da inteligência artificial. Após a OpenAI disparar o GPT Sol 6.1 para dominar a execução interativa de código e a Google DeepMind desvendar a magnitude multimodal de 4 milhões de tokens do Gemini 4, a Anthropic oficializou o lançamento global do Claude Mythos 5.1 — a sua mais imponente e rigorosa criação fundacional até hoje. Se a concorrência priorizo
Google has introduced Web multimodal search performance reporting in Search Console, giving site owners a dedicated way to measure how people discover their content through image-based search interactions. The new Multimodal filter covers traffic associated with Google Lens, Circle to Search on Android, image uploads to Google Search, and Chrome's right-click option to search an image. The addition expands Search Console beyond its longstanding focus on text-led web searches. For businesses that publish product imagery, how-to visuals, location-specific photos, video thumbnails, or other visua
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend I built SplitMate AI, an AI-powered expense splitting assistant designed for my friends who live together. The idea came from a simple problem I saw regularly. When my friends share expenses for dinners, parties, groceries, and Uber rides, they usually send messages in WhatsApp like Kasun paid 4500 LKR for dinner with Hasaru and Dhamith 3000 (Sidath,Dhamith) Uber 1200 Kasun At the end of the week or month, they manually check messages, write calculations on paper, and try to figure out who owes whom. SplitMate AI
Three updates from this week that matter if you build with models, not just if you follow model launches. OpenAI confirmed it has been talking with Anthropic and Google DeepMind about AI safety coordination. Reporting also points to a proposed industry body, sometimes called SAFA (Standards Authority for Frontier AI), aimed at shared testing, audits, and vulnerability checks for frontier models. Why builders should care: if labs agree on common evals and thresholds, the bar for "safe enough to ship" stops being a private slide deck and starts looking more like an industry checklist. OpenAI ope
At AirAsia, I own production microservices on the Manage My Booking platform — flight changes, fare summaries, price-slash, post-booking ancillaries. Over a dozen services, each with its own repo, pipeline, and accumulated quirks. The platform processes 130 million requests a day. My problem was never writing code. It was the context-switching overhead.
I've spent the last three-plus years building and operating the microservices behind AirAsia Move — the travel super-app that handles flights, hotels, and ancillaries for millions of travelers across Southeast Asia. The system processes over 130 million requests per day, roughly 1,500 requests per second sustained, with spikes well above that during flash sales. I own services on the Manage My Booking platform: flight changes, fare summaries, price-slash features, post-booking ancillaries. A single user action like changing a flight can fan out to eight or nine downstream calls. At this scale,
A search box sitting on a page looks like the obvious way to help people find things. It works, but not in the three ways most teams assume. Short answer. Yes, Confluence Cloud has it. In the editor type /live search and insert the Live Search macro.
One of the highest-turnover football markets is deceptively simple: will the total goals in a match go over or under 2.5? Where does that line come from, and what is the "fair" probability of each side? Here's the full derivation, with numbers you can reproduce from one formula. A team's goal count in a match can be approximated by a Poisson distribution: P(X = k) = λ^k · e^(−λ) / k! where λ is the team's expected goals (estimated from recent attack strength, opponent defense, home/away).
We’re launching ReviewBench, a benchmark for code review agents built on representative GitHub pull requests, multi-source ground truth, calibrated evaluation, and production-aligned metrics. The post ReviewBench: An open benchmark for AI code review appeared first on The GitHub Blog.
Every developer and designer has been there: you stumble upon a beautifully designed website and want to reverse-engineer its design tokens or inspect how its typography, color palette, and utility classes are structured. Normally, this means opening DevTools, inspect-elementing dozens of DOM nodes one by one, manually copying computed HEX codes or CSS --variables, and spending 30 minutes manually formatting a tailwind.config.js or syncing values into Figma. I got tired of this repetitive manual process, so I built a privacy-first browser tool to automate it: Token Inspector. Modern web apps r
I have made a website that monitors all your agents and alerts you when an agent is stuck, idle, or caught in a loop. This way, you can give prompts and go about your day without the need to check back every 5 mins to see if the AI is still generating code. The app is open source and free to use; log in, run a command in the terminal, and go make dinner. You'll get notified when you are needed. The link to the site is: https://aiwatchdog.vercel.app/ I appreciate any feedback.
Building something that needs to know "which jobs are out there right now" keeps hitting the same wall: every remote job board is a different scrape, a different HTML layout, a different rate limit, and a different set of dead links. So I built a single API that normalizes all of it into one call. It's live, free, and requires no API key. Remote Jobs API is a small service that aggregates remote roles from five public sources — Remotive, RemoteOK, Jobicy, WeWorkRemotely, and Hacker News (Who's Hiring) — and returns them as one clean, typed feed. No registration.
I run a crawler that re-reads public Vinted shelves in France, Germany, Spain, Italy and Portugal about every 30 minutes and records which listings disappear. On 5 October 2026 the database held 14,676,854 listings. September was the first full month of departure tracking, so I published it as a dataset: the Vinted Demand Index, October 2026. One word matters before the numbers. I call them observed departures, not sales.
Java's HashMap mitigates Hash Collision Denial of Service (DoS) attacks by automatically converting congested linked list buckets into Red-Black trees once a bucket exceeds 8 entries and the total map capacity reaches 64. This transition reduces lookup complexity from O(N) to O(log N), preventing attackers from exhausting CPU resources. Imagine sending a tiny 2MB payload to a web server and completely freezing a CPU core for nearly three-quarters of an hour. In 2011, security researchers demonstrated exactly this vulnerability. It wasn't a complex buffer overflow or a zero-day exploit; it was
Customer returns and refund requests are among the costliest operational friction points for modern e commerce platforms. Traditional solutions force businesses to choose between two undesirable extremes: Rigid rule engines that deny valid edge cases, frustrating loyal shoppers. Naive generative chatbots that can be easily manipulated through prompt injection, promising free money and unauthorized chargebacks. To solve this, I designed and built an autonomous AI Customer Support Refund System. It balances instant automated decisions with deterministic policy checks, velocity safeguards, and a