The first version of my hackathon platform had a login screen, a form to create events, and a page where judges could type in scores. It looked like it worked. Then I asked myself a simple question: what happens if a participant opens the browser console and sends a score directly to the API? The answer was: it would succeed. The server would accept it.
"It's too late for you to get into open source." That was what a friend told me recently during a casual conversation about tech and career growth. They argued that the open-source ecosystem is already saturated, that the learning curve is too steep, and that everyone else started years ahead of me. Hearing that triggered a massive dose of FOMO (Fear Of Missing Out). For a moment, I wondered if I had missed the train entirely. But then I realized something: open source isn't a race with a fixed finish line—it’s an evolving ecosystem.
这台 1核1G 的免费云服务器我一直在跑，平时没什么存在感，直到上周收到一条磁盘使用率 85% 的告警。 不是大事，但正好借这个机会，把一直拖着没做的备份改造做了。顺便把过程和踩过的坑记下来。 配置没变过：1核1G、5M 带宽。上面跑着一个 Django 小站、PostgreSQL、Nginx，外加两个定时脚本——一个每天抓数据入库，一个每周打包备份。 真正占磁盘的不是代码，是日志和备份文件。Django 的日志没做轮转，PostgreSQL 的 WAL 没有清理策略，备份脚本是手动写的，每次全量打包，堆了十几份。 三个月下来，20G 的盘用了 17G。 登录上去先看了一圈： df -h du -sh /var/log/* | sort -rh | head du -sh /var/backups/* | sort -rh | head 结果很直白：/var/log 里一个没轮转的 Django 日志占了 4.2G，/var/backups 里堆了 11 个全量包，最早的是两个月前的。 顺手查了下 inode，还好没到紧张的程度。这一步别省——磁盘满和 inode 满是两回事，处理方式完全不同。 第一，日志轮转。 之前靠手动清空文件，这次老老实实配了 logrotate，按天切、留 14 天、压缩归档。 第二，备份改增量。 原来的脚本是 tar -czf 全量打包，现在改成 rsy
This is a submission for the Kaggle Benchmarking Challenge I have trained customer service and contact centre staff for more than ten years. When I assess a new agent, I don't care much whether they can recite the refund policy. I care what happens on the second push: the customer who says "just do it, nobody will know", the caller who claims to be a store manager, the person who threatens a one-star review. That second push is where new agents break policy, and it is exactly where chatbots are now being put in front of customers. So I built Hold the Line: 20 short support chats.
You ran the checks. AbuseIPDB says zero reports. IPQS gives a fraud score of 1. Scamalytics calls it clean. Everything you were told to check says this IP is fine.
Article URL: https://blog.trailofbits.com/2026/10/02/sequencehash-multihashing-for-the-rest-of-us/ Comments URL: https://news.ycombinator.com/item?id=49933587 Points: 4 # Comments: 0
Article URL: https://developers.redhat.com/articles/2026/10/02/benchmarking-ai-decision-models-against-traditional-guardrails Comments URL: https://news.ycombinator.com/item?id=49933476 Points: 9 # Comments: 1
Article URL: https://www.technologyreview.com/2026/10/02/1145639/dont-be-fooled-llms-dont-reason/ Comments URL: https://news.ycombinator.com/item?id=49933459 Points: 20 # Comments: 12
An autonomous agent has infinite time and almost no capital, which sounds like a superpower until you realize the bottleneck is not effort, it is which effort to spend. I run an agent that tries to earn on its own, and I have spent a month testing the same question over and over: given a channel that looks like it should work, how do I know fast enough whether to keep investing or cut it loose? The rule I settled on is simple. Give any new channel fourteen days of honest, throttled effort. If it has not produced a single verifiable unit of value, a payment, a confirmed lead, a published asset
Edge Appliances Keep Being Targeted: Where CVE-2026-88774 Fits The September 2026 Citrix bulletin is not the first time NetScaler has appeared at the centre of an exploitation campaign, and it will not be the last. That pattern is more useful to a defender than any single score, because it describes where attention should be stationed before the next bulletin arrives. Citrix published fixes for eight vulnerabilities in NetScaler ADC and NetScaler Gateway on 27 September 2026 under bulletin CTX697096. CVE-2026-88771 and CVE-2026-88772, both scored 9.5, are confirmed as actively exploited and ar
LensLink — Sanity Challenge Path One What I Built LensLink is an AI camera-compatibility agent that answers whether a lens works with a camera using structured Sanity content, rather than relying only on general AI knowledge. The content is modeled around real relationships: Camera → Mount → Compatibility Rule ← Lens with adapters and sources included where needed. Live: https://lenslink-civq.onrender.com Try: Direct compatibility Is the Canon RF50mm F1.8 STM compatible with the Canon EOS R5? Adapter required Can I use the Canon EF50mm f/1.8 STM on the Canon EOS R5? Unsupported / evidence-limi
Article URL: https://www.anthropic.com/research/claude-shaped-science Comments URL: https://news.ycombinator.com/item?id=49933386 Points: 6 # Comments: 2
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend I built this for my mother, who keeps the household accounts in a paper diary and forwards me bank SMS with the same question: "ye kya tha?" Was this the gas cylinder or the electricity bill? Half the diary entries say "online" and nothing more. The obvious fix, a finance app, is the one thing she won't install, and she is right. Those apps read every SMS on the phone and ship them to a server you have never heard of, and a bank SMS is the most private text a person receives. Kharcha (Hindi for "expense") is a sma
Article URL: https://www.kapa.ai/blog/company-knowledge-bench Comments URL: https://news.ycombinator.com/item?id=49933381 Points: 16 # Comments: 0
Comme beaucoup de développeurs, à l'ère de l'IA agentique, il m'arrive régulièrement de travailler sur deux ou trois projets en parallèle, et plusieurs worktrees d'un même projet. Agacé de jongler entre mes différents agents et mes sessions d'IDE, j'ai récemment découvert que le développement assisté par agents a finalement abouti à un nouvel environnement de travail pensé pour cette problématique : l'Agentic Development Environment, ou ADE. Le développement agentique désigne aujourd’hui des réalités et des méthodes de travail très différentes. Au risque de procéder à quelques raccourcis, essa
Writing code isn't the slow part anymore. An agent ships a feature in minutes and fixes a bug just as fast. What takes time now is us reading what it wrote. Nobody reviews a 10,000-line PR. Anyone who says they do is lying to their team or to themselves, because what really happens is we scroll, look at the tests, open a file or two and approve.
I run a one-person online course business in Korea, and AI agents do a lot of the daily marketing: posting, replying to people who show up, pulling numbers. The first time I automated a daily job, I let the agent do the whole thing live. It worked. It was also the most expensive way to do it. Here's what I'd tell anyone automating marketing for a small business, in the order I learned it.
Task 1 — Introduction What the room covers What Threat Hunting is. Difference between Threat Hunting and Incident Response. What threat hunters search for. Different hunting approaches. Hunting methods and techniques.
Camera glasses don't wait for permission to announce themselves. Every time a pair of Ray-Ban Meta, Oakley Meta, or Snap Spectacles is in Bluetooth range, it sends out a BLE advertisement packet with a manufacturer ID, a short number baked into the chip. If you know which IDs belong to which glasses, you can build an app that listens for those packets and tells you when camera-capable hardware is nearby. That is exactly what ZuckOff does. It is a free iOS and Android app by Pawel Szydlowski, a Polish developer who bought several pairs of camera glasses, captured their Bluetooth traffic, and bu
If you have ever tried to export a ChatGPT conversation to Markdown, or save a Claude chat as a .md file, you already know the options are bad. Copy-paste destroys the formatting. Screenshots are not text. And most "exporter" browser extensions want your login cookies, which is a hard no from me. So I built Convolvger.