标签：#EV

I remember the exact moment the excitement turned to dread. I had just integrated GPT-4 into a side project—a small document summarization tool. The pricing page said $0.03 per 1K input tokens and $0.06 per 1K output tokens. Clean, simple, two numbers. I calculated roughly $0.01 per summary and smiled. Two weeks later the bill arrived: $87.43 for what I thought would be maybe $15. I wasn't being careless. I had read the docs. I knew about tokens. But the silent costs—the ones nobody puts in a neat table—had quietly multiplied my burn rate by six. That experience taught me that AI API pricing is a lot like buying a printer. The upfront cost is seductive; the real expense hides in the ink cartridges, the proprietary drivers, and the forced upgrades you never planned for. Let's talk about those hidden costs, because I'll bet you've either already hit them or you're about to. The Token Trap That Isn't What You Think Everyone knows tokens are the unit of billing, but the gap between "understanding tokens" and "feeling tokens" is enormous. First, there's the input/output asymmetry . GPT-4 charges double for output tokens. That's fine for short answers, but what about chain-of-thought? If you ask the model to reason step-by-step, those intermediate steps count as output tokens—and they add up fast. I had a single query balloon from 500 output tokens to 2,400 because the model decided to work through a logic puzzle aloud. My cost quadrupled without me changing a thing in my prompt. Then there's the system prompt tax . Many developers stuff context into system messages: instructions, examples, formatting rules. Those are input tokens paid every single time, even when the user's query is tiny. If your system prompt is 1,500 tokens and you handle 10,000 requests, that's 15 million input tokens you're paying for—whether the model uses them or not. And don't get me started on retry costs . You hit a rate limit or your request times out? The token count for that failed request? S

Part of my AWS learning journey — transitioning from Systems Engineer to Cloud/DevOps. This session completes the database picture — Aurora's read/write architecture, ElastiCache caching strategies, and DynamoDB from table creation to production-ready query patterns. 📋 Topics Covered # Topic Type 1 Aurora Endpoints — Writer vs Reader Concept + Interview 2 What Happens When the Aurora Writer Fails Concept + Cert 3 ElastiCache Caching Patterns — Lazy Loading, Write Through, Session Store Concept + Interview 4 Cache Invalidation Concept + Interview 5 DynamoDB — What It Is and When to Use It Concept + Interview 6 DynamoDB Table Creation — Keys and Settings Concept + Lab 7 Table Classes — Standard vs Standard-IA Concept + Cert 8 Capacity Modes — On-Demand vs Provisioned Concept + Cert 9 Warm Throughput Concept + Cert 10 DynamoDB Items & Attributes — CRUD Operations Concept + Lab 11 Query vs Scan — The Critical Difference Concept + Interview 12 Local Secondary Index (LSI) vs Global Secondary Index (GSI) Concept + Cert 13 Bonus Concepts — Streams, DAX, Consistency, Transactions Concept + Interview 14 Interview Questions Interview 15 Practice Tasks Practice Aurora Endpoints — Writer vs Reader Aurora doesn't give you just one database endpoint — it gives you two, each serving a different purpose and routing to different parts of the cluster. Writer Endpoint (Primary Endpoint): Always points to the current primary/writer instance. All write operations (INSERT, UPDATE, DELETE) go here. If a failover happens and a replica is promoted, Aurora automatically redirects this endpoint to the new writer — your application's configuration never needs to change. Reader Endpoint: A load-balanced endpoint that distributes read-only queries (SELECT) across all available Aurora Replicas. You don't manage which replica serves each query — Aurora handles the routing, spreading read traffic evenly across however many replicas exist. Why this architecture matters: In a typical application, read

My Terraform drift pipeline could detect a manual EC2 tag change, classify it as LOW, and run Terraform to remove it. Then the pipeline moved on. The evidence existed, but it was spread across CodeBuild output, Lambda logs, and an SNS message. If I wanted to know what changed, how it was classified, and whether remediation started, I had to reconstruct the event from multiple AWS services. The pipeline could act on drift. It could not remember drift. Phase 4 added that memory: a durable DynamoDB record, a read only API, and a small dashboard that turns the event history into something I can inspect without opening three AWS consoles. The Stack Terraform drift event ↓ SNS ↓ Severity Lambda ├── classifies HIGH / MEDIUM / LOW ├── starts remediation for eligible LOW drift └── writes the audit event to DynamoDB ↓ API Gateway HTTP API ↓ Read only Lambda ↓ DynamoDB Query ↓ CloudFront → static dashboard ↑ private S3 bucket The browser receives static HTML, CSS, and JavaScript from CloudFront. JavaScript calls API Gateway, the API Lambda queries DynamoDB, and the returned JSON becomes the live dashboard. There is no EC2 web server and no application process running continuously. Step 1: Store Every Classified Event I created a DynamoDB table with a composite key: resource "aws_dynamodb_table" "drift_events" { name = "terraform-drift-events" billing_mode = "PAY_PER_REQUEST" hash_key = "project" range_key = "timestamp" attribute { name = "project" type = "S" } attribute { name = "timestamp" type = "S" } } project groups the history for one Terraform project. The ISO 8601 timestamp orders its events. DynamoDB only requires attribute definitions for keys and indexes. Fields such as high_count , changes , and status still belong in each item, but they do not belong in the table schema block. I passed the table name into the existing severity Lambda instead of putting it directly in the code: environment { variables = { DRIFT_EVENTS_TABLE = aws_dynamodb_table . drift_events . name

Rate limiting is one of those things that doesn't feel urgent—until someone hammers your login endpoint at 3am and you wake up to a flooded database and a locked-out user base. I added this to my Nuxt base layer after realising I'd shipped several projects with zero protection on auth routes. Not great. This post walks through the exact setup I now use: Redis-backed, an in-memory fallback when Redis is down, named presets for different sensitivity levels, and a 429 page that shows a live countdown instead of just dying on the user. The structure Three pieces, each with one job: createRateLimiter() — a factory that builds the limiter, using Redis with an in-memory fallback applyRateLimit() — what you call inside handlers to enforce a limit server/middleware/rateLimiter.ts — global middleware so every route gets a baseline for free 1. Install npm install rate-limiter-flexible ioredis rate-limiter-flexible does the heavy lifting: sliding windows, Redis integration, and the insurance fallback pattern we'll use. 2. The factory Create server/utils/rateLimiter.ts : import { RateLimiterRedis , RateLimiterMemory , type RateLimiterAbstract , } from ' rate-limiter-flexible ' import { getRedisClient } from ' ./redis ' export interface RateLimiterConfig { keyPrefix : string // Must be unique per limiter, e.g. 'rl:auth' limit : number // Maximum requests within the window windowSeconds : number } export interface RateLimitResult { allowed : boolean limit : number remaining : number resetAt : number // Unix timestamp in seconds when the window resets retryAfter : number // Seconds until retry; 0 if allowed } function buildLimiter ( config : RateLimiterConfig , ): RateLimiterAbstract { const insurance = new RateLimiterMemory ({ keyPrefix : config . keyPrefix , points : config . limit , duration : config . windowSeconds , }) const redis = getRedisClient () if ( ! redis ) { return insurance } return new RateLimiterRedis ({ storeClient : redis , keyPrefix : config . keyPrefix , points

I benchmarked a vision-language model and scored it at 0.31. The real number was 0.70. Same model, same weights, same hardware, same 100 questions. The only thing that changed was how I read its output. I had already written up the 0.31 as a capability finding and concluded the model was unsuitable. That conclusion was wrong, and the failure was entirely in my harness. Here is the mistake, because I doubt I am the only one making it. The setup I was evaluating a batch of open-weight and frontier models on a multiple-choice benchmark: multi-view driving scenes, four options per question, one correct answer. Standard stuff. The prompt asked for reasoning followed by a final line, Answer: X . My scoring code did the obvious thing: m = re . search ( r " Answer:\s*([A-D]) " , output ) pred = m . group ( 1 ) if m else None # None scores as wrong That last comment is the bug. What actually happened The model I was testing is a "thinking" model. It emits a long internal reasoning trace before it commits to an answer. I had a generation budget of 1024 tokens. On easy questions it reasoned briefly, emitted Answer: B , and scored fine. On hard questions it reasoned at length, hit the token cap mid-thought, and never emitted the answer line at all. So the harness scored every one of those as wrong. 64 of 100 questions returned no parseable answer. Zero of those were image-loading errors or crashes. They were all truncation. And the truncation was not random: Uncertainty 0/8 answered Counterfactual 0/3 answered Safety-critical Planning 1/11 answered Safety-critical Prediction 3/12 answered Look at that distribution. The questions the model failed to answer were precisely the questions that required the most reasoning. My harness was systematically discarding the model's performance on exactly the hard subset I was trying to measure, and reporting the result as a capability ceiling. Of the 36 it did answer, it got 86% right. The model was fine. My measurement was garbage. The fix

When I first started testing APIs, my workflow looked a lot like everyone else's. I'd open a GUI...

You've done it. You're finally done building the website or application you've been working on for quite a while. Proud and elated, you go to share this on your socials or to your buddies — uh oh, what's this now? The preview in WhatsApp shows no headline, your avatar is cropped and dimensions seem wrong. I've been there too. The site looked fine in the browser. The problem was everything outside the browser: link previews, search snippets, and tab icons all use a separate metadata layer most of us skip until something breaks. So, how do you fix this? Use this as a pre-launch checklist — or run it on a site that's already live but sharing badly. What I ran into on my own portfolio When I ran this audit on shwethaadiraj.com , the site rendered fine — but sharing it told a different story. I had pointed both the favicon and Open Graph image at my profile avatar. At tab size the illustration was unreadable; in link previews it got cropped awkwardly. An OG validator then flagged two things I hadn't considered: the image was 512×512 (most platforms expect 1200×630 ), and there was no headline or CTA on the image itself — so Slack and LinkedIn showed a plain square with none of the context from my meta title. I replaced the favicon with a simplified monogram, regenerated the OG image at the correct aspect ratio with my name, tagline, and site URL on it, and re-ran the debuggers. Even then, previews didn't update until I hit Scrape Again — platforms cache OG data aggressively, so fixes on your end won't show up until you bust that cache. None of this required rethinking the app. It was a metadata pass — the kind of work that's easy to defer and annoying to discover at the share button. Before we get into the specifics, here's a primer on what metadata can actually impact: What is metadata for? Metadata, simply put, is data about data. Search engines, crawlers and social sites all parse different metadata from your app. Search & Discovery: The title and description in your

Everyone's talking about agentic AI shipping production code. Nobody's talking about what happens when you actually sit down and grade thousands of lines of it against a rubric, line by line, for months. I have. And the failure pattern that shows up over and over isn't the one Twitter/X is arguing about. The job title that didn't exist two years ago "AI evaluator." "AI trainer." "Expert contributor to frontier model training data." None of these existed as job titles when I started my career. Now they're where a chunk of the most interesting engineering signal in the industry is actually happening — quietly, behind NDAs, far from the demo videos. Here's what the job actually is: agentic coding outputs land on your desk, and you grade them against a structured rubric — correctness, instruction adherence, quality, edge-case handling. You design adversarial prompts to find where the model's reasoning breaks. You decide which checks can be programmatic and deterministic, and which genuinely need a human who's shipped production systems to make the call. This is RL environment design and LLMOps in its rawest form, and it's a completely different skill from "prompt engineer" or "ML researcher." It's closer to being a QA lead for a junior engineer who never sleeps, never gets embarrassed, and will confidently ship the wrong answer with perfect syntax. The pattern: agents are great at code, bad at consequences Here's the uncomfortable part. The failure mode people are loudest about — hallucinated APIs, made-up library functions — is the easy failure mode. It's loud, it's obvious, and any decent test suite catches it in seconds. The failure mode that actually matters, the one that slips past a surface read and even past a naive test suite, looks like this: The code is syntactically perfect and semantically wrong about failure. It handles the happy path beautifully and quietly assumes the retry, the timeout, the partial write, the duplicate message never happens. It optimises

Strict prop types and typed emits get most of the attention in Vue 3 + TypeScript setups, but provide / inject is where type safety quietly falls apart if you use the API the way the docs show it by default. inject() without a type hint returns unknown , which means every consumer of an injected value either casts it blindly or loses autocomplete entirely — and a typo in the injection key becomes a runtime undefined instead of a compile-time error. The Default Setup Is Untyped by Construction The naive version compiles, but gives you nothing: // Provider provide ( ' theme ' , currentTheme ); // Consumer const theme = inject ( ' theme ' ); // type: unknown Nothing here catches a typo in the key string, and nothing tells the consumer what shape theme actually has. Both problems come from using a plain string as the injection key. InjectionKey Fixes Both Problems at Once Vue exports an InjectionKey<T> type specifically for this. Define it once, typed, and both provide and inject become fully type-checked against the same symbol: // keys.ts import type { InjectionKey } from ' vue ' ; export interface Theme { mode : ' light ' | ' dark ' ; accentColor : string ; } export const ThemeKey : InjectionKey < Theme > = Symbol ( ' theme ' ); // Provider import { ThemeKey } from ' ./keys ' ; provide ( ThemeKey , { mode : ' dark ' , accentColor : ' #4f46e5 ' }); // Consumer import { ThemeKey } from ' ./keys ' ; const theme = inject ( ThemeKey ); // type: Theme | undefined The | undefined in that last type isn't a quirk — it's inject being honest that a consumer might render without a matching provider above it in the tree, which is a real runtime possibility TypeScript is right to force you to handle. Handling the undefined Case Without Littering ?. Everywhere The common mistake is providing a default value to silence the undefined type instead of actually checking for it: const theme = inject ( ThemeKey , { mode : ' light ' , accentColor : ' #000 ' }); // default masks missing pro

Hello, I'm Maneshwar. I'm building git-lrc, a Micro AI code reviewer that runs on every commit. It is free and source-available on Github. Star git-lrc to help devs discover the project. Do give it a try and share your feedback. A few years ago the AI conversation looked like this. "Should I pay $20?" "No, $200." "Actually this new tool is $39/month." My wallet started looking like it had gone through a startup funding winter. Then Google quietly walked into the room and started dropping free AI tools like Oprah handing out cars. "You get an AI IDE!" "You get a workflow builder!" "You get a GitHub coding agent!" ...except nobody really noticed because Google announced them across five different events, Labs pages, GitHub repos, and random blog posts. So I spent some time collecting the ones developers will actually find useful. No "AI that writes your wedding speech." No "AI that guesses your spirit animal." Just tools that can actually help you ship software. Bookmark this one. 1. Pomelli https://labs.google/pomelli If you've ever launched a side project, you already know the painful truth. Building the product is fun. Writing 37 LinkedIn posts explaining the product... not so much. Pomelli takes your website, understands what your product does, builds a brand profile, then generates social posts around it. Think of it as hiring an intern that actually reads your landing page before tweeting. Would I let it post automatically? No. Would I happily let it generate the first draft so I don't stare at a blinking cursor? Absolutely. Perfect for: Indie hackers SaaS founders Open-source maintainers pretending they enjoy marketing 2. Stitch https://stitch.withgoogle.com Remember when designing an app meant opening Figma... ...moving a button 3 pixels... ...asking for feedback... ...moving it back 3 pixels? Stitch skips a surprising amount of that. You describe the interface. Or upload a sketch. Or even paste a wireframe. It generates modern UI designs and can even produce

Ford's CEO has described the new Ford Fathom as a "Model T moment" for the company. But the EV pickup is unlikely to live up to lofty expectations, and maybe it doesn't have to.

The California Teamsters union says that new self-driving truck regulations could eventually cause around 400,000 people to lose their jobs.

There's a workflow that basically didn't exist three years ago and now half the job-seekers I know use it: ask ChatGPT, Claude, Gemini, or any AI to write your resume bullets, get back beautifully structured text… and then spend forty minutes mangling it into Word or a drag-and-drop resume builder, fixing bullet indentation and font sizes by hand. Here's the thing that bugged me: LLMs already speak Markdown. Ask any chatbot for a resume and you get ## Experience , **Senior Engineer** , - Shipped X — clean, structured Markdown. Then every resume tool on earth makes you throw that structure away and re-enter it into form fields. So I built ResumeMD: a split-pane editor where you paste Markdown on the left, see a typeset resume on the right, pick a template, and download a PDF. No signup to start, everything in localStorage by default. This post is about the parts that fought back. Decision 1: Markdown is the source of truth Most resume builders store your resume as a proprietary JSON blob mapped to form fields. I wanted the document itself to be portable text. That means the entire product is "just" a Markdown renderer with opinions: h2 = section headers (Experience, Education) — these get the decorative treatment per template: uppercase, border, background, prefix glyphs. h3 = job titles — plain, bold, primary color. One weird trick I'm genuinely fond of: the sidebar template splits a single Markdown document into main column and sidebar using an HTML comment ( <!-- sidebar --> ) as the split marker. Content above the marker is the main column; below is the sidebar. It keeps the document valid Markdown everywhere else. The preview is react-markdown + remark-gfm with a 300ms debounce, styled by a template system that turned out to need three parallel implementations of every template: CSS classes for the live preview, inline-style functions shared between preview and template cards, and pure-JS styles for the PDF renderer. Thirty-two templates, three layers each. When

Most AI image generators look like a prompt box with a Generate button. That is also how my first version started. But once real users entered the workflow, the difficult problems appeared somewhere else: browser refreshes, external task IDs, reference images, partial failures, credit refunds, private assets, and public artwork moderation. While building Magggic , I learned that an AI image generator is less like a form submission and more like a small distributed job system. This article covers the decisions that made that workflow more reliable. The code samples below are intentionally simplified. The important part is the shape of the workflow, not a specific database or image provider. The prompt box is only the beginning A synchronous prototype is easy to imagine: const images = await provider . generate ( prompt ); return images ; That version works until the request takes a minute, the provider times out, one of four requested images fails, or the user refreshes the page. The production workflow I needed looked more like this: Prompt + references ↓ Create a local queued task ↓ Charge credits with an idempotency key ↓ Submit work to the image provider ↓ Persist every completed output immediately ↓ Finalize the task and refund failed outputs ↓ Keep the result private until the user publishes it The provider request is only one step. The local task is the source of truth for what the user sees. 1. Persist the task before calling the provider The first important decision was to create a generation record before making the external API request. A generation stores the information needed to reconstruct the job: type Generation = { id : string ; userId : string ; idempotencyKey : string ; prompt : string ; referenceImages : string []; model : string ; ratio : string ; resolution : string ; count : number ; cost : number ; status : " queued " | " generating " | " completed " | " failed " ; outputs : string []; providerRequestIds : string []; failureReason : string |

Ford's first midsize electric truck has a name and a price, but we're still waiting for a proper design reveal.

Due in "fall 2027," Ford said Thursday that it won't reveal what Fathom looks like until early next year.

I enjoy using display-mirroring glasses as a way to play Steam Deck games on a big screen, and I'm always on the hunt for the best pair. There's still not one perfect model I can recommend to everyone, but Viture's new $299.99 Pro 2 have sharper-looking screens than models that cost hundreds more. It's mostly […]

If you've ever tried coding on an iPad, Galaxy Tab, or Chromebook, you know the frustration: Standard desktop tutorials assume a Mac or high-spec Linux laptop. Neovim configuration takes 4 hours of plugin debugging. Touch input on mobile terminals sucks without a dedicated extra-keys bar. I built DevDock (dock) to solve this permanently. What is DevDock? DevDock is a turnkey developer environment manager built specifically for mobile devices, Termux, Chromebooks, and low-spec hardware. Instead of fighting configuration files, one command installs a complete, high-performance terminal stack: bash curl -fsSL https://get.devdock.io | bash -s -- --profile=fullstack ⚡ Key Features Sub-5ms Terminal Rendering: Uses Starship prompt + Zsh lazy-loading tuned for ARM chips. Termux Touch Optimization: Automatically injects an ESC/TAB/CTRL touch bar and enables mouse scrolling in Tmux. Low-Memory Neovim: Starts in <50ms and uses under 50MB RAM while providing full Language Server Protocol (LSP) support for TS, Go, Python, and Rust. Curated Profiles: fullstack: Web + API tools frontend: React, TS, Vite & Tailwind preset backend: Go, Rust, Python, Postgres & Redis CLI tools devops: Kubectl, Helm, Terraform, and Cloud CLIs 🛠 Trying It Out bash Check your mobile terminal health: dock doctor View available developer stacks: dock profiles Initialize a frontend stack: dock init frontend 🔗 Open Source & Community DevDock is 100% open source under the MIT License! GitHub Repo: github.com/devdock/devdock Web Showcase: devdock.io Give it a spin on your Android phone, tablet, or cloud shell and let me know what you think in the comments below!

A coding agent with no context usually hesitates, searches, or asks a question. A coding agent with stale context can be much more confident. That is the dangerous case. The file still exists. The instructions look deliberate. The generated JSON is valid. The agent follows it exactly — into a package that stopped owning the feature two weeks ago. Nothing looks broken until the edit is already in the wrong place. I wanted repository context to have an expiration signal that CI could verify, not a date someone had to remember to check. The failure is not missing documentation Imagine a monorepo where packages/auth owns token validation. The repository publishes a machine-readable handoff: { "startHere" : "docs/for-agents/packages/auth.md" , "editRoots" : [ "packages/auth" ], "checks" : [ "pnpm --filter @example/auth test" ] } Later, token validation moves to packages/security . A maintainer updates the source documentation but forgets to regenerate the handoff index. There are now two internally consistent answers in the same repository: the source documentation says packages/security ; the generated agent context still says packages/auth . The old answer is not malformed. That is precisely why it is risky. I reproduced the drift with one edit I tested this against the public fixture in Doc Bridge , using version 1.2.6. The first index and freshness check passed: Index is fresh expected: 359355e5... actual: 359355e5... Then I changed one agent-facing source document: - Package: packages/os-core - Layer: L1 + +Token validation now belongs to packages/security. I did not touch the generated index. The next check returned exit code 1: ak-docs gate run index-freshness Index is stale. Run: ak-docs index expected: b099695d... actual: 359355e5... After I ran ak-docs index , reviewed the generated change, and ran the gate again, both hashes matched and the check passed. The hashes are not trying to prove that the documentation is true. No checksum can do that. They prove a na