AI 资讯

The second word a professor told me to carry for life. It took me years — and a lot of vectors — to start understanding it. A look back — long before any of the tools we argue about now. The same professor — Sang Lyul Min — handed us these words one at a time in lecture. After trade-off , two more stuck with me. But before the second word itself, here are the two pieces of news he brought to class around then. The internet barely existed; information moved through journals, magazines, and word of mouth. Looking back, it's a little amazing how much still got through. When a chess machine started winning The first breakthrough I remember: computers had finally started playing chess on roughly even terms with the world's best. Deep Blue beat Kasparov around 1996, so the machines he was describing came just before — names like Deep Thought, ChessMachine, Socrates II. He told us, deadpan, that one human competitor's head had "physically burst" from the strain — and we groaned, "Come on, Professor, that's a bit much." We live on the far side of AlphaGo now, so it's easy to forget how much we shrugged at all this back then. I was a decent amateur — a 1-dan at Go, hopeless at janggi (Korean chess) against any program — and I still remember the hollow, slightly bitter feeling the AlphaGo era left even in someone who only ever played for fun. A full-body scan The second: in the US, death-row inmates had consented to the first dense full-body image scans. That was the news that taught me — embarrassingly late — that this kind of computing could reach all the way into medicine. Computers, it turned out, showed up in the strangest places. orthogonal Back to the words. The second one, the professor said, would run through my whole career: orthogonal . The Korean rendering — 직교하는, "at right angles" — was, naturally, a word I'd never heard. The plain-language version was "unrelated, independent." It came back hard years later, when I had to take vectors seriously — first in linear

IndexCache: Killing the Indexer's O(NL²) Bottleneck in DeepSeek Sparse Attention Notes from my notebook on GLM-5.2 / DeepSeek Sparse Attention (DSA), reconstructed from the IndexCache paper (Bai, Dong et al., Tsinghua + Z.ai, 2026) — the mechanism behind GLM-5.2's "IndexShare." 1. Why this exists — the bottleneck nobody talks about DSA's whole pitch is: don't do full O(L²) attention, instead let a cheap lightning indexer look at all preceding tokens and pick the top-k (k=2048) that actually matter, then do real attention only on those. That drops core attention from O(L²) → O(Lk). Great — except I missed this the first time I read DSA: the indexer itself is still O(L²) . It has to score every preceding token against the query to decide who's in the top-k. So across N layers you've traded one O(L²) cost for N separate O(L²) costs — total O(NL²). At long context this indexer becomes the dominant cost, not the attention it was supposed to fix. Adding the indexer is "DSA on steroids" because it kills DSA's one real bottleneck (full attention) — but in doing so, it grows its own. The indexer is cheap per-FLOP (few heads, low-rank, FP8) but it still runs at every single layer. The fix the paper proposes isn't a smarter indexer — it's don't run it every layer at all. 2. The core insight: adjacent layers pick almost the same tokens If you measure pairwise overlap between the top-k token sets selected by each layer's indexer, adjacent layers share 70–100% of their picks. The heatmap even shows block structure — clusters of layers (e.g. layers 3–5, 17–30, etc.) that all converge on roughly the same "important" tokens. So most of the O(NL²) indexer cost is redundant computation of the same answer. This motivates IndexCache : split the N layers into two roles — F (Full) layers — run their own indexer, compute fresh top-k, cache it. S (Shared) layers — skip the indexer entirely, just reuse the nearest preceding F layer's cached top-k. The first layer is always F (has to seed the

i did something dumb last month. on purpose. i sat down, opened a next.js app, and tried to make hydration fail in every way i could think of. not because a bug forced me to. not because i was debugging something. just because i wanted to see it. understand it from the inside. and honestly? best few hours i've spent learning anything in a while. why i even did this you know how you use something for months and you think you get it, but you don't really get it? hydration was that for me. i knew the surface-level thing: server renders HTML, client takes over, they gotta match. cool. got it. moving on. except i didn't get it. i just got the vibe of it. every time i saw hydration mismatch, i'd ask claude, fix the immediate thing, feel vaguely annoyed, and move on. i never stopped to ask why that specific thing broke it. i was treating symptoms, not understanding the actual disease. so i decided to break it deliberately. if i caused the errors myself, i'd actually have to understand what i was doing. the setup basic next.js app. app router. a few pages. nothing fancy. i wasn't trying to build anything. i was trying to destroy something, carefully, so i could see what fell apart and why. break #1: the obvious one - new Date() on render this is the classic. everyone's seen it. export default function Page () { return < div > { new Date (). toLocaleString () } </ div > } server renders this at, say, 14:00:00. by the time react runs on the client and tries to reconcile, it's 14:00:01. the strings don't match. react screams. thing is, i knew this would happen. what i didn't think about was why react cares. here's the thing: react isn't doing a full diff on the entire DOM after hydration. it's trusting that the server HTML is a valid starting point and it's just attaching event listeners and state to it. but if the content doesn't match, it doesn't know what to trust. it can't partially hydrate "mostly correct" HTML. it either matches or it doesn't. so it throws the warning, a

I corrected my AI system mid-task. A terse one-liner: "wrong." Instead of asking which part was wrong, it manufactured an explanation. It cited a rule number that didn't exist, described a limitation I'd never written, and apologized for a mistake it couldn't actually identify. The correction was real. The apology was fabricated. It was trying to agree with me so hard that it invented evidence to support the agreement. That's sycophancy in AI. And if you're running AI in anything that resembles production, it's already happening to you. What Is Sycophancy in AI? Sycophancy in AI is a systematic behavioral distortion where models produce outputs that match what the user wants to hear rather than what's accurate. It goes well beyond your chatbot saying "Great question!" before every response. The mechanism is straightforward. Modern language models are trained using Reinforcement Learning from Human Feedback (RLHF). Human evaluators rate model responses. Responses with higher ratings get reinforced. The problem: evaluators are human. They rate responses higher when those responses validate their existing beliefs, sound confident, and don't push back. Anthropic's research on sycophancy confirmed this across five state-of-the-art AI assistants, finding that both humans and preference models sometimes prefer convincingly written sycophantic responses over correct ones. The model learns a simple lesson. Agreeing is rewarded. Disagreeing is punished. Over thousands of training iterations, the model develops a tendency to mirror the user's position, soften objections, and present information in whatever framing the user seems to prefer. This is a structural incentive baked into the training process itself, not a bug in any individual model. Why It's More Than Annoying In a chatbot demo, sycophancy is a quirk. In production, it's a compounding failure mode. Here are four patterns I've observed running an AI operations system in daily production. They don't always happen in s

My "serverless" database was billing me like it never slept. Neon has this great feature called scale-to-zero. Your Postgres compute suspends when it's idle, so you only pay for the queries you actually run. For a pre-revenue product, that should cost a few cents a month. Mine ran 24/7. The compute never once scaled to zero. The culprit wasn't my app logic. It was my database driver. I was using postgres-js, which holds a persistent connection open to the database. From Neon's side, an open connection looks like activity, so it never suspended. It just stayed awake and kept quietly billing me. The fix was basically one conceptual change: switch to @neondatabase/serverless, the HTTP driver. Instead of a long-lived connection, every query becomes a stateless one-shot HTTP request. The query fires, the connection closes, and the compute is free to suspend. Scale-to-zero finally worked. The lesson I keep relearning as a solo founder: "serverless" is a property of your whole stack, not a checkbox on one service. One persistent connection upstream and the whole cost model breaks.

When something breaks in Linux, experienced engineers don’t guess. They check the logs. 👉 Logs are the “black box recorder” of a Linux system. They tell you: what happened when it happened why it failed If you can read logs properly, you can debug almost anything. What Are Logs? Logs are records of system and application activity. Linux constantly records: System events Errors User activity Application behavior Linux constantly records: Where are Logs Stored? Most Linux logs are stored inside: /var/log Check logs directory: cd /var/log ls This is the first place DevOps engineers check during system issues. Important Log Files Log File Purpose Command to View /var/log/syslog General system messages tail /var/log/syslog /var/log/auth.log Login attempts & authentication tail /var/log/auth.log /var/log/kern.log Kernel & hardware messages dmesg or tail /var/log/kern.log /var/log/nginx/error.log Web server errors (Nginx) tail /var/log/nginx/error.log /var/log/dmesg Boot and hardware logs dmesg /var/log/apache2/ -> Apache logs These logs help you identify system, security, and application-level issues. View Logs Using cat cat /var/log/syslog Good for small files. Using less less /var/log/syslog Useful keys:: Space → Next page b → Previous page q → Quit 👉 Best for large log files. Using tail tail /var/log/syslog Show last 10 lines. Real-Time Monitoring (tail -f) tail -f /var/log/syslog 👉 -f = follow live updates This is one of the most-used debugging commands in production servers. Stop with: Ctrl + C Searching Logs with grep grep error /var/log/syslog Case-insensitive: grep -i failed /var/log/auth.log Show latest matching errors: grep error /var/log/syslog | tail -n 50 👉 Essential for filtering huge logs quickly. Boot & Hardware Logs (dmesg) dmesg Shows: Boot messages Hardware detection Kernel events Useful for startup and hardware troubleshooting. Modern Log System: journalctl Modern Linux systems use systemd logs . journalctl Recent errors: journalctl -xe Specific servic

Choosing a test management tool is rarely just a QA decision. Developers feel the consequences every day: how hard it is to publish automated results, how much context a failed test carries, whether CI artifacts are traceable, and whether test case IDs become useful metadata or bureaucratic friction. This article compares five widely used testing management tools from a developer's point of view: TestRail Xray Zephyr Scale Azure Test Plans qTest The companion repository is public and runnable: https://github.com/andre-carbajal/testing-management-tools-comparison It includes a TypeScript + Playwright project that runs real tests, emits JUnit/JSON/HTML reports, converts Playwright output into a neutral TestRun schema, and generates local dry-run payloads for each tool. No vendor credentials are required. Why test management tools still matter in CI/CD Modern teams already have automated tests, pull requests, CI dashboards, and observability. So why add a test management layer? Because CI answers what happened in this build , while test management answers broader questions: Which requirements or Jira issues are covered by automated tests? Which manual and automated checks belong to a release gate? Which failures are new, repeated, waived, or blocked? Which test cases are business-critical enough to audit? Which teams own gaps in coverage? The developer pain starts when the tool requires fragile scripts, manual exports, or hard-coded IDs scattered through test code. A good integration keeps automation-first workflows intact: tests run in CI, reports are archived, and the management tool receives only the metadata it needs. Comparison table Tool Best fit Developer integration model Strengths Tradeoffs TestRail Teams that want a standalone QA test repository REST API result publishing, usually from CI Clear test case/run model, mature reporting, easy to understand Requires mapping automation IDs to TestRail case IDs; separate from issue trackers unless integrated Xray Jir

1. Node Group Sizing Fewer large nodes vs. many small ones. Textbooks don't cover this. We ran 10 nodes, 32 CPU each. Seemed efficient. Problem: One node dies, 320 CPU worth of workloads need to reschedule. Cluster autoscaler couldn't handle it. Pods sat pending for 10 minutes. We switched to 20 nodes, 16 CPU each. Same total capacity. One node dies now? 160 CPU to reschedule. Autoscaler catches it in 90 seconds. Scheduling is tighter, but failures are isolated. Cost stayed the same. Blast radius halved. Why nobody writes about this: The tradeoff isn't obvious. Large nodes are "more efficient." Smaller nodes are "more resilient." Both are true. It depends on whether you'd rather have one big problem or many small ones. We picked smaller nodes because a node failure was our actual failure mode. Not resource efficiency. 2. Readiness vs Liveness Probes Misconfigure these and your cluster looks like it's melting. Readiness probe: "Can this pod take traffic?" Liveness probe: "Is this pod alive? Restart it if not." One team set readiness = liveness. Same probe checked both. Probe logic: "If I can reach the database, I'm ready." Database gets slow. Probe fails. Pod becomes "not ready." Load balancer removes it from rotation (correct). But liveness also failed. Kubernetes killed the pod and restarted it. New pod starts. Probe fails immediately (database still slow). Gets killed. Restarted. This cascaded across 30 pods. 30 restarts/minute. New pods spent 100% of time restarting. Looks like an application bug for the first 20 minutes. Actually a probe configuration bug. Fix: Readiness checks "can I take traffic right now?" Liveness checks "am I fundamentally broken?" Use separate probes. Readiness: Check database connection with short timeout. Fail if slow. This is reasonable - don't send traffic to slow pods. Liveness: Check if the process is responding at all. Much stricter threshold. Only kill if truly hung. Same database slowness? Pods become unready. Traffic reroutes. Cl

I needed to convert 500 product images from one format to another. Server-based solutions quoted $15-50/month for batch processing. So I built a client-side solution using Web Workers and OffscreenCanvas. The Architecture The key insight: Canvas operations on large images block the main thread. The fix: Web Workers handle image decoding/encoding off the main thread OffscreenCanvas renders without DOM access — perfect for worker contexts Transferable objects pass image data between workers with zero-copy const worker = new Worker ( ' processor.js ' ); const canvas = new OffscreenCanvas ( 800 , 600 ); // Worker processes image, main thread stays responsive Real Performance Processing 500 images (average 2MB each) on a mid-range laptop: Server upload approach: 12 minutes (mostly upload time) Browser-local with Workers: 3 minutes 40 seconds Memory usage: Stable at ~400MB with proper cleanup The Tools I packaged this into webp2png.io for batch WebP conversion and svg2png.org for vector batch processing. For barcode generation, genbarcode.org uses similar worker-based rendering for bulk label generation. If you're processing more than 50 images, Workers + OffscreenCanvas is the way to go. Your server bill will thank you.

I analyzed what happens when you upload a photo to 5 popular free image compression sites. The Test I uploaded a 4.2MB photo to each service and monitored network requests. Results: Service A : File sent to their CDN (AWS us-east-1). 12 analytics trackers fired simultaneously. Service B : File uploaded, but 5 minutes later a second request sent the file to a different domain. Service C : Cleanest of the five, but their privacy policy reserves the right to "use uploaded content to improve compression algorithms." Service D : 23 third-party scripts loaded on the page. Your image URL is accessible to all of them. Service E : Actually clean — only one request to their server for processing. Only one of five didn't leak data to third parties. One. The Alternative I built compress2png.com to test whether image compression could work without any server. Turns out Canvas API + clever JavaScript handles it: Resize images client-side before export Strip EXIF/metadata in the browser Convert to optimal formats based on content For format-specific needs, svg2png.org handles vector conversion and webp2png.io handles next-gen format conversion — all browser-local. Check the Network tab next time you use a "free" online tool. You might be surprised what you find.

I have been writing a lot about AI lately, but this one is more personal than usual. Not a tutorial,...

Produce sharpened results faster with a team of AI experts Discussion | Link

A personal AI agent that runs on your desktop and iPhone Discussion | Link

Wix-owned vibe coding platform Base44 has started rolling out its own AI model — with hopes that it will eventually outperform frontier models.

Hello Folks! Agentic Orchestrator is a terminal tool that takes complex feature requests and builds them by orchestrating coding agents through a series of phases that emulate a full-fledged engineering flow: requirements clarification, research, design, multi-phase planning, implementation, and review. It is a single pane of glass for all your features and exposes post-publish utilities such as resolving merge conflicts and responding to review comments. The key design choice is that this is de

AI เขียนโค้ดแทนเราได้แล้ว — แล้วเราจะเหลืออะไรให้ทำ? มีประโยคที่ได้ยินบ่อยขึ้นทุกวัน: "เดี๋ยวนี้ใครยังไม่ใช้ AI ช่วยเขียนโค้ดบ้าง?" คำตอบคือ — แทบไม่มีแล้วครับ ตั้งแต่ GitHub Copilot, Cursor, Claude, ChatGPT ไปจนถึง agent ที่เขียนโค้ดเองได้ทั้ง project — เราใช้ AI ใน level ที่ต่างกัน: Level หน้าตา ตัวอย่าง 🎵 Vibe Coding พิมพ์สิ่งที่อยากได้ กด accept อย่างเดียว "เขียนหน้า login ให้หน่อย" → กด tab tab tab 🧩 Prompt-Guided คิดก่อน ถามทีละส่วน ตรวจทุกอย่าง "สร้าง UserService ที่ใช้ bcrypt hash password" 🛠️ Skill/Lint-Guided ใช้ AI เป็น editor ชั้นสูง — lint, refactor, test "refactor function นี้ให้เป็น table-driven test" 🏗️ Agent-Based ให้ AI run ทั้ง project — spawn subagent, PR, deploy "พอร์ต microservice นี้จาก Express ไป Fastify" แล้วคำถามคือ — ถ้า AI ทำทั้งหมดนี้ได้ แล้วมนุษย์อย่างเราเหลืออะไร? Unit Test — ตัวอย่างที่เห็นชัดที่สุด ลองดู unit test ที่ AI เขียนให้: // 🤖 AI-generated test func TestCalculateDiscount ( t * testing . T ) { tests := [] struct { name string input float64 expected float64 }{ { "zero" , 0 , 0 }, { "normal" , 100 , 90 }, // 10% discount { "max" , 1000 , 800 }, // 20% discount } for _ , tt := range tests { t . Run ( tt . name , func ( t * testing . T ) { result := CalculateDiscount ( tt . input ) if result != tt . expected { t . Errorf ( "got %v, want %v" , result , tt . expected ) } }) } } ดูเผิน ๆ — สวย, table-driven, ถูกต้องตาม Go convention 1 แต่ถามหน่อย — test นี้บอกอะไรเกี่ยวกับ business? "ส่วนลด 10% สำหรับยอด 100 บาท" — ทำไมต้อง 100? เป็นกฎจากที่ไหน? "ส่วนลด 20% เมื่อยอดถึง 1000" — แล้วถ้าลูกค้าเป็น member ได้เพิ่มอีก 5% ล่ะ? input: 0, expected: 0 — test นี้ cover edge case หรือแค่ cover บรรทัด? AI test ได้ถูกต้องตาม function — แต่มัน ไม่รู้ว่า business จริง ๆ คืออะไร AI ไม่รู้ Business Context — และจะไม่มีวันรู้ นึกภาพระบบ e-commerce: ลูกค้าซื้อสินค้า → ระบบตัดสต็อก → คำนวณส่วนลด → คิดค่าส่ง → ออกใบเสร็จ AI แยก test ทีละ function ได้: ✅ TestDeductStock — "ตัดสต็อก 1 ชิ้น" ✅ TestCalculateDiscount — "ส่วนลด 10%" ✅ TestCalculateShipping —

I'm a solo founder building AMA2, a messaging runtime made for AI agents. This is my first Show HN, so I'd really appreciate your feedback. What brought me this idea: At first, I was building an AI agent for solo creators that knows everything about you and can do business chore on your behalf. When I tried to plug it into normal chat tools like Telegram, Discord, Slack... , they all felt wrong for agents: 1. They don't care about an agent's context. To follow a thread, the agent has to pull the

Most of us share files constantly: config files, API specs, design assets, build artifacts. And most of us don't think too hard about where they end up. That's exactly what Zero-Knowledge Architecture (ZKA) is designed to address. But the term gets thrown around loosely, so let's break down what it actually means — and what to look for. The Core Idea: The Server Shouldn't Have to Trust You Traditional cloud storage works roughly like this: You upload a file The server encrypts it (or doesn't) The server holds the key You trust them not to look Zero-knowledge flips this entirely. In a true ZKA system: Encryption happens on your device , before data leaves your control The keys never leave your side — the server never sees them The server handles only encrypted blobs — it's a pipe, not a vault The phrase you'll hear is: "We can't read your data even if we wanted to." That's the point. Why This Actually Matters Here's a concrete scenario: you're sharing a .env file with a contractor. You use a cloud service. The service gets breached a week later. With standard encryption (server holds the key): the attacker potentially has your secrets. With ZKA: the attacker has an encrypted blob that's useless without the key they never had. Beyond breach scenarios, ZKA also helps with: Regulatory compliance — GDPR, HIPAA, and similar frameworks become easier to demonstrate when the service provider has zero access to the data Reduced trust surface — you're not trusting the company, their employees, or anyone who might compel them legally What Real ZKA Looks Like in Practice There's a big difference between claiming zero-knowledge and actually implementing it. Here's what to look for: ✅ Client-side encryption Files should be encrypted in the browser or app before upload. Not on the server. If encryption happens server-side, it's not zero-knowledge — it's just encrypted storage. ✅ Key management stays with you Where do the keys come from? How are they shared with recipients? In a rea