标签：#ai

How I let non-engineers ship AI tools to production — and the boring infrastructure that made it safe. A product manager described a workflow in plain English — "every morning, pull yesterday's failed payments, group them by error code, and post a summary to our channel." Twenty minutes later it was running in production. She never opened an editor. She never saw a line of TypeScript. She talked to an agent, the agent wrote the code, and — once a human had reviewed the pull request — it shipped. That sentence should make you nervous. It made me nervous, and I'm the one who built the thing. The demo is "look, it wrote the code." The operation is "a marketer's tool now has a path to the payments database and nobody reviewed it." The interesting engineering isn't the part where an LLM writes code — that's the easy, demo-able part. It's the guardrails that decide whether the code it writes is allowed to exist. Here's the platform, and the five problems I had to solve to make it safe to hand to people who can't read the code that runs. The shape of the thing The platform is a place where anyone — engineers, PMs, designers, QA — can publish a reusable AI tool, and everyone else can use it. Write once, available to all. A few terms up front, because the whole design leans on them: MCP (Model Context Protocol) is a standard way for an AI client to discover and call your functions. The key detail: there's a step where the client asks the server "what tools do you have?" and the server answers with a list. Hold onto that — half the design hangs off that one list. Cloudflare Workers is code that runs on Cloudflare's servers at the network edge instead of your own. Durable Objects is per-session server-side storage that lives outside the model's context — the finite, token-costing window of everything the model can currently see. None of this is exotic; what matters is where each piece of state lives. Under the hood it's three small Workers speaking MCP: a gateway (auth, routin

Traditional content pages are often designed for a linear reader. The introduction sets context, the middle develops the idea, and the conclusion ties everything together. AI retrieval does not always work that way. A system may identify smaller content units, pull the most relevant section, compare it with other sources, and use that fragment to support an answer. The full page still matters, but the retrievable blocks inside the page matter just as much. A useful Tumblr post explains the idea in simple terms: https://www.tumblr.com/digitalisedsoul/820825642809573376/ai-does-not-read-your-content-like-a-human?source=share For Dev Community readers, the pattern is familiar. Poorly structured inputs lead to weaker outputs. If content is dense, vague, or dependent on surrounding paragraphs, it becomes harder to extract and reuse. If content is modular, clear, and properly scoped, retrieval becomes easier. Marketing teams can learn a lot from this. A strong content page should behave like a set of well labelled components. Each section should answer a specific question. Headings should be descriptive, not decorative. Paragraphs should avoid vague references such as the above point or this approach when the section may be read independently. Definitions should appear close to the terms they explain. Examples should include enough context to stand alone. Proof should be written as text, not only displayed as graphics. Internal links should connect related concepts in a way that helps both readers and systems understand the topic map. A page about AI search visibility, for example, should not only include one broad explanation. It should break the topic into useful blocks: what AI visibility means, why AI systems retrieve passages, how source trust works, what makes content reusable, and how brands should measure answer presence. Each block becomes a possible answer unit. That structure does not weaken the reader experience. It improves it. Developers, marketers, and busi

In 2007, humans took roughly 85 billion photos a year. Photography was a specialized craft that...

👋 Hey there, Tech Enthusiasts! I'm Sarvar, a Cloud Architect who loves turning complex tech problems...

A new report finds "high-intensity AI adopters” saw headcount increase 10.2%. Among those companies, entry-level headcount rose by 12%, countering the rhetoric that AI kills junior jobs.

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 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

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

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

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 —

Let’s be honest: as developers, our relationship with our office chairs is... complicated. We start the day sitting upright like productivity gurus, but four hours into a debugging session, we’ve morphed into a human pretzel. This "gamer lean" isn't just a meme; it leads to chronic back pain and decreased focus. In this tutorial, we are going to build a real-time posture tracking system using MediaPipe Pose and Computer Vision to save your spine. By leveraging AI productivity tools and the power of cross-platform Electron desktop apps , we will create a silent guardian that watches your form and pings you the moment you start slouching. If you've been looking for a practical way to dive into MediaPipe and Node.js integration, you're in the right place. For those looking for more production-ready patterns and advanced AI implementations, I highly recommend checking out the deep dives at WellAlly Blog . 🏗 The Architecture The system works by capturing frames from your webcam, processing them through a pre-trained neural network to identify body landmarks, and then applying some basic trigonometry to determine if your posture is healthy. graph TD A[Webcam Stream] --> B[MediaPipe Pose Engine] B --> C[Extract 33 Keypoints] C --> D{Geometry Engine} D -->|Angle > Threshold| E[Slouch Detected] D -->|Angle < Threshold| F[Good Posture] E --> G[Electron Main Process] G --> H[System Notification 🔔] F --> I[Wait 5s] I --> A 🛠 Prerequisites To follow along, you'll need: Node.js (v16+) MediaPipe (The pose solution) OpenCV.js (For frame manipulation) Electron (For the desktop shell) 🚀 Step 1: Setting Up the Pose Engine MediaPipe provides a "Pose" model that gives us 33 landmarks in 3D space. For posture correction, we specifically care about the Ears (7, 8) , Shoulders (11, 12) , and Hips (23, 24) . The Math: Calculating the "Slouch" We measure the angle between the Ear, the Shoulder, and a vertical axis. If your ear moves too far forward relative to your shoulder, that's "Forward

If you've spent any time building or working with APIs, you've probably realized that the terminal is...

The interesting part of OpenAI and Codex on AWS is not that another cloud menu got more model names. That part is useful. Enterprises want strong models. Developers want Codex closer to their infrastructure, data, and deployment machinery. The interesting part is that frontier AI is being pulled into the same boring machinery that already governs everything else companies run: procurement, IAM, billing commitments, region policy, audit logs, support contracts, data boundaries, and security review. That sounds like paperwork. It is also how enterprise software becomes real. model access was the easy problem For a while, AI adoption was framed as an access problem. Can we call the model? Can we get enough rate limit? Can we wire the SDK into our product? Can the coding assistant see enough of the repo to be useful? Those are real questions. They are not the end of the story. The next set is much more familiar to anyone who has operated software inside a company: which account owns this usage, which data can cross the boundary, who can create agents, which region runs inference, how the bill is allocated, and what evidence exists when an incident involves model output. That is the part where the demo becomes a platform. OpenAI on AWS matters because many companies already have that platform muscle in AWS. They have IAM, billing, private networking, audit trails, procurement paths, compliance evidence, cost allocation tags, and teams whose job is to make all of this survivable. Putting a frontier model behind that machinery does not make the hard parts disappear. It makes them legible. bedrock is a procurement surface Amazon Bedrock is usually described as a managed model service, which is true and also undersells the point. For enterprises, Bedrock is a procurement and control surface. If OpenAI models and Codex are available through Bedrock, a company can route adoption through an existing cloud relationship instead of creating a new vendor path for every team that wa

This week's Java roundup for June 22nd, 2026, features news highlighting: the GA releases of Hardwood 1.0 and Endive 1.0; the June 2026 edition of Azul Payara; point releases of Quarkus, LangChain4j; the first beta release of WildFly 41; and introducing Eliya JDK and the Open Source Sustainability Initiative (OSSI), the latter of which was founded by HeroDevs and Commonhaus Foundation. By Michael Redlich

you probably didn't think about this when you signed up. you entered your card details, verified...

Smartphones are welcoming the agentic AI overlords.

Starting in IT I started in IT at a local school in my small town in December 2024. It was my first job out of college after earning my B.S. in Cybersecurity. None of the infrastructure was updated. Everything was failing, and luckily, I had one other IT person there: my director. I honestly think he knew less than I did, and he would get frustrated at almost every ticket. Even then, I knew I wanted a role where I could code. Fast forward to more recently, he had a freak out and quit. Now we have a two-person team, and everything is finally up to date and functioning. Even with things improving, I still knew I wanted to move toward a SWE-type role. Learning to Code I first decided to learn C# for Unity. I was really into game dev while I was in college, so it felt like a natural place to start. I began with the Microsoft/freeCodeCamp C# certification, and I surprisingly really enjoyed it. I made a few small games on itch.io that no one cared about, but I had fun building them. After that, I went on a bit of a language-hopping spree. I jumped from C# to C++, then into full-stack web development. I actually stuck with web dev for a while and really enjoyed it. But this cycle went on for awhile of just constant swapping. Wannabe Founder Then something switched overnight. I went from writing maybe 0-5% AI-generated code to using AI for nearly everything. I started spam-building startup ideas that did not really go anywhere. I may have made around $2-3k from them, but most of the time I was just chasing money and building whatever I thought had the quickest path to making some. I got seriously addicted to vibe coding. I tried Codex, Cursor, Claude, and basically anything with AI in it. I did like Codex the most, though. Eventually, I realized I had almost completely stopped coding by hand. I was not passionate about the startup ideas I was building. I loved coding, and I knew I had to step back. Back to Coding Now I am back to coding without AI assistance. I will eventua

AI makes it easier to build the wrong thing with confidence. That is the part I think a lot of beginner builders and freelancers miss. The obvious story is that AI makes execution faster. That is true. I can ask an AI coding tool to explain an error, compare implementation options, inspect a project, write code, refactor a screen, generate a QA checklist, or help me pick up where I left off. That is a huge change. But speed is not the whole story. When the tool gets faster, your judgment becomes more important, not less. You have to decide what the project is allowed to become. You have to decide which tradeoffs are acceptable. You have to decide whether the output actually matches the user's job. You have to decide when the AI is solving the real problem and when it is decorating the wrong one. In my freelance work, AI changed the job from searching and stitching to directing, reviewing, and verifying. That sounds cleaner than it feels. Directing means you need to know what outcome you want. Reviewing means you need to notice when the answer is plausible but wrong. Verifying means you cannot treat a green checkmark, a pretty screen, or a confident explanation as proof that the app actually works. The beginner mistake is believing AI removes the need to think clearly. The better rule is this: AI removes some friction from execution, then hands you more responsibility for scope. The Faster Tool Still Needs A Smaller Job When I started using AI heavily for software work, the old research loop changed immediately. Before modern AI tools, a lot of software work meant digging through documentation, old forum posts, Stack Overflow answers, YouTube videos, outdated examples, and half-related blog posts until something clicked. You stitched pieces together and hoped the tutorial you found still matched the version of the framework you were using. Now you can ask the tool directly. That is better. It is also dangerous if you confuse a fast answer with a good product decision

I recently built a multi-agent customer support system on Azure AI Foundry and NVIDIA NIM. First time doing anything like this. Made four predictions upfront about what would happen. Three of them were wrong. Here is what I actually learned. 1. "Tokens" is not a unit of cost It is a unit of work. The price per unit of work varies by 5-10x depending on which model did the work. I was tracking total token count across both the small 9B model and the large 49B model as if they cost the same. They do not. Total tokens went up in the optimized version. Cost in dollars probably went down. I was measuring the wrong thing the whole time. 2. A verbatim hash cache on natural language traffic deflects ~0% of queries I predicted 25-40% cache deflection. The actual number was 0%. Every query in my test set was a unique string, so the hash-based cache never had a single chance to fire. A verbatim cache is not a simpler version of a semantic cache. It is a different thing entirely. If your workload is natural language, build semantic similarity caching from day one, not as an upgrade later. 3. configure_azure_monitor() does not capture OpenAI SDK calls by default You need to install and initialize opentelemetry-instrumentation-httpx explicitly: pip install opentelemetry-instrumentation-httpx==0.61b0 from opentelemetry.instrumentation.httpx import HTTPXClientInstrumentor HTTPXClientInstrumentor().instrument() Without this, your App Insights Logs will show customMetric and performanceCounter entries (CPU, memory) but nothing about what your agent actually did. 4. Pin your OpenTelemetry versions or everything breaks Installing opentelemetry-instrumentation-httpx without version pinning pulled in opentelemetry-api 1.42.1. But azure-monitor-opentelemetry-exporter needs opentelemetry-api==1.40. The conflict is silent until things start misbehaving. Pin everything to the 0.61b0 / 1.40.0 line: pip install \ "opentelemetry-api==1.40.0" \ "opentelemetry-instrumentation==0.61b0" \ "opentelem

Token burn and business outcomes are not correlated. More burn means more inefficiency, not more value. The electricity problem Imagine you walk into a dark room. Turning on a light helps you see. Turning on every light in the building does not help you see better. It's still the same room. Now every surface is equally lit, the contrast is gone, and you're paying for power you didn't use. Tokens work the same way. A focused prompt with clear scope is the single overhead light over your desk. A sprawling prompt with unlimited exploration is every light in the building — you're burning power, not producing insight. Tokens are electricity, not output. More throughput doesn't mean more value. I've had weeks where I burned through my allocation and looked back at the end to find nothing concrete. Code that worked but went unused. Exploratory branches that dead-ended. Agents that generated plausible-looking output that didn't survive first review. A lot of motion. Not much progress. The ceiling stops you from doing that indefinitely. It forces a moment of reflection: did this burn produce anything real? If the answer is no, more capacity isn't the fix. More discipline is. Three patterns I now use instead I started paying attention to what actually ships versus what just burns context. I gave the patterns names so I could catch myself faster: RTK — Read The Knowledgebase. A focused 15-minute read of the codebase, identifying the exact files and exact changes, saves 200K+ tokens of exploratory waste. The agent doesn't discover the shape of the task — it executes against a known one. Caveman — compress before you prompt. Strip greetings, filler words ("I think", "basically", "Let me know if that makes sense"), and closing courtesies. Every word in your prompt multiplies across every response token. Less fluff in means less fluff out. Ponytail — spec the minimum viable solution. "Robust", "scalable", "enterprise-grade", "comprehensive" — these words invite scope creep. Specif