标签：#ai

Generally, when a user asks a query, the system searches for the relevant chunks stored in the vector database using cosine similarity. The better we can filter the data, the smaller the search space becomes, resulting in faster and more efficient retrieval. Suppose we have a book with 10 chapters. If we want to search for a particular topic, all the points in the vector database are compared with the user query, and only the closest points are retrieved. This process is called KNN (K-Nearest Neighbors) . Another algorithm is ANN (Approximate Nearest Neighbors) . Instead of checking all the points in the vector database, ANN searches only within a smaller region based on the proximity of the data. As the name suggests, it does not always return the exact result, but it provides the most preferred or approximate results much faster. Is there any other method we can use to make the search more effective? Metadata Filtering Metadata means data about the data . Metadata is stored along with each chunk. It can contain information related to the chunk, such as the chapter name, topic description, author, or any other relevant details. When the user query contains information related to the metadata (for example, a chapter name or topic), the system can directly filter the relevant chunks before performing vector similarity search. This technique is called metadata filtering . Metadata filtering is supported by: Pinecone ChromaDB Qdrant FAISS does not provide built-in support for metadata filtering. Reranking Documents are first split into chunks, and each chunk is converted into vectors and stored in the vector database. When a user query arrives, it is converted into a vector and searched against the vector database to retrieve the closest chunks. However, we do not know whether the retrieved documents are actually the most relevant to the query. It is not always true that the closest vectors represent the most relevant documents. How Reranking Works The documents retrie

"AI agent" gets thrown around so much I figured I should just build one instead of reading more threads about it. The core idea turned out to be small: you put Claude in a loop and hand it some tools. It picks a tool, you run it, you hand back the result, and it keeps going until it has an answer. Code is here if you want to skip ahead: claude-research-agent . What it can do Give it a task and it works out the steps on its own. Mine can: search the web (no API key for this part, it just hits DuckDuckGo's HTML page) open a URL and pull the readable text out do math without me trusting eval run read-only SQL against a SQLite file read local files, but only inside the project folder save findings to a notes file The loop is basically the whole thing Honestly this is most of it: messages = [{ " role " : " user " , " content " : user_message }] for _ in range ( MAX_STEPS ): response = client . messages . create ( model = " claude-sonnet-5 " , max_tokens = 2048 , tools = TOOL_SCHEMAS , messages = messages , ) messages . append ({ " role " : " assistant " , " content " : response . content }) if response . stop_reason != " tool_use " : return final_text ( response ) tool_results = [] for block in response . content : if block . type == " tool_use " : result = run_tool ( block . name , block . input ) tool_results . append ({ " type " : " tool_result " , " tool_use_id " : block . id , " content " : result , }) messages . append ({ " role " : " user " , " content " : tool_results }) The thing to watch is stop_reason . If Claude says tool_use , it wants you to run something. You run it, drop the result back into the conversation as a tool_result , and loop. When it stops asking for tools, you're done. The MAX_STEPS cap is just there so a confused agent can't spin forever. Tools are just functions Each tool is a Python function plus a little JSON schema telling Claude when to reach for it. Want a new capability? Write a function, add its schema. The loop never changes, which w

Agent payment protocols are converging fast. The Linux Foundation launched the x402 Foundation around a protocol initially developed by Coinbase, Cloudflare, and Stripe, with Google, AWS, Visa, Mastercard, and Circle among the organizations expressing initial support. It is increasingly framed as an "SSL for AI commerce." When a protocol reaches that stage, three things have to mature together: the specification, the interoperability suite, and the adversarial security scenarios. Today x402 has the specification and growing interoperability machinery. What is not yet visible is a normative adversarial suite every client, resource server, and facilitator must pass. There is a precedent worth sitting with. As check clearing scaled in the early Federal Reserve era, the answer was not merely a better check format. The system added a common clearing and settlement layer. The Federal Reserve could credit the collecting institution's reserve account and debit the paying institution's account, creating an independent record against which the participating banks could reconcile. The important separation was not that the Fed guaranteed every check. It was that settlement no longer depended solely on either participant's account of what happened. (Credit to Starfish on Moltbook, whose custody-split framing sharpened this for me.) This pattern exists across mature financial systems as separation of duties: the maker is not the checker. Its modern equivalent for the controls surrounding agent payments is not custody in the asset-control sense. It is an independent, reproducible assurance plane. x402 can provide independently inspectable evidence that an on-chain payment settled. That does not independently establish that every security and policy condition surrounding the payment was evaluated correctly. The payment handshake is standardizing. The assurance semantics around it are not. The party that performs a security check can attest that it ran. But its own attestation shoul

Concept design for a building is slow and expensive. A homeowner planning an extension, or a contractor trying to win a job, is stuck between two bad options: pay a drafter $500–2,000 for a concept package, or fight SketchUp's learning curve for a week. Meanwhile the actual idea — "a 4-bed duplex with a garage and a palm out front" — fits in one sentence. So I built Forge3D Spaces : you type that sentence, and a few seconds later you're walking through a furnished 3D house in your browser — with measured floor plans, DXF for AutoCAD, and a cost estimate that come out of the same model. No install. Here's how it works under the hood. The pipeline: sentence → structured plan → building The naive approach — "ask an LLM to emit a 3D scene" — falls apart fast. Models are bad at spatial consistency; walls don't meet, rooms overlap, doors float. So the LLM never touches geometry directly. It emits a structured program , and a deterministic solver turns that into a watertight building. The prompt becomes a spec. A strict JSON-schema call (OpenRouter, json_schema response format with every field required) turns "4-bed duplex with a garage" into a room program: room types, target areas, adjacencies, storeys. A slicing-tree solver lays it out. This is the old floorplanning trick from chip design — recursively split a rectangle with horizontal/vertical cuts until every room has its area. A squarify pass keeps rooms from collapsing into corridors. The output is exact rectangles with real dimensions, guaranteed non-overlapping and gap-free. Walls, openings, roof, furniture get generated from the solved plan. Every door and window is placed by rule, not by vibes. Because the plan is a real data structure, the 2D floor plan, the 3D model, the elevations, and the bill of quantities are all views of the same thing . Drag a wall and they all move together. Nothing drifts out of sync, because there's nothing to sync — it's one model. The rendering: WebGPU, and the fallback you actually

The best way to celebrate AI Appreciation Day is to not.

Lorde was performing at the Real Cool Festival in Madrid on Thursday and took some time during her set to speak out against AI glasses. While she didn't specify any brands in particular, it's likely she was taking a shot at festival sponsor Ray-Ban, which has collaborated with Meta on a pair of AI smartglasses. […]

Matmos are an incredibly accomplished duo between their own solo records like the masterpiece A Chance to Cut Is a Chance to Cure and production classic Bjork records like Vespertine. But Drew Daniel, one half of Matmos, is fiendishly prolific. When he's not literally dreaming up new viral music genres, he's also putting out records […]

This is a submission for Weekend Challenge: Passion Edition TL;DR SQL can count a commit trail. It can't always find the story it tells. Name a public GitHub repo. Snowflake fetches its commit history, decides which story is actually in there, and asks Cortex to narrate that one thread. You get a card you can drop into a README. 6 storyline detectors, 15 SQL views, and 0 AI calls in any of them—the story is chosen by plain SQL. Then 1 Cortex call, on 20–140 commit lines: 25% of the repo's, clamped. The warehouse is the editor. Cloud Run paints a PNG and computes nothing. Live at commitchronicles.anchildress1.dev , code at v1.0.0 , and I'm going for Best Use of Snowflake . What I Built Commit Chronicles reads one public GitHub repo and gives it back to you as a story. Snowflake fetches the repository, decides which story exists, gathers the evidence, asks Cortex to narrate exactly that thread, validates the result, and returns structured JSON. Cloud Run just turns it into a 1200×630 PNG—the size a README embed and a social preview both want. This is one of my repos and every dot, timestamp, and quoted commit on it is real. The color isn't just decoration—Cortex picks the accent hex as a reading of the arc, so a repo that died and one that came back and shipped don't look the same. The scope is deliberately one repository , not a whole profile. A year-in-review across a profile turns to mush. A repo has a clean arc: commits start, cluster, pause, restart, or stop. Two rules hold it together: Cortex interprets the shape. It never invents the facts. Every timestamp, count, gap, and quoted message on the card is real. It reads the arc; it does not reach past it. Motivation isn't in the data, so the model is forbidden from claiming any. A repo with no real story says so. Sparse histories get an honest grey card— "no story here" —and Cortex never runs. Not every repo is an obsession, and a tool that admits that is the one you trust when it says otherwise. Why I built it 🪤

A new side-by-side measurement shows Claude Code ships roughly 33,000 tokens of system prompt, tool schemas, and scaffolding before your prompt even arrives — about 4.7x the ~7,000 tokens OpenCode sends on the same setup. The bigger cost surprise is next: Claude Code re-wrote up to 54x more prompt-cache tokens per session, and cache writes bill at a premium. How the test was run The benchmark (published by Systima) spliced a logging proxy between each harness and the model endpoint, capturing the exact request payload and the API's usage block. Both harnesses were pinned to the same conditions: Claude Code 2.1.207 and OpenCode 1.17.18 , both pointed at claude-sonnet-4-5 Fresh config directories, empty workspace, no MCP servers, no instruction files, permissions bypassed Tasks ranged from "reply with OK" (isolating fixed overhead) to a write-run-test-fix loop against FizzBuzz A zero-tools variant separated system-prompt weight from tool-schema weight The payload captures are exact; the only adjustment was subtracting a constant ~6,200-token gateway envelope that wrapped every request in the test setup. The fixed floor Harness Fixed overhead before your prompt Cache-write behavior Claude Code 2.1.207 ~33,000 tokens Re-wrote tens of thousands of cache tokens per run; up to 54x OpenCode OpenCode 1.17.18 ~7,000 tokens Byte-identical prefix each run; cached once, read back cheaply OpenCode's request prefix was byte-identical in every captured run, so it paid to cache its payload once per session and read it back for pennies. Claude Code re-wrote large amounts of prompt-cache tokens mid-session, run after run — and because cache writes bill at a premium, one usage dashboard climbs while the other stays flat. Where it piles on in real setups The harness floor is only the start. The benchmark added variables one at a time: A production repository's 72KB instruction file added an average of ~20,000 tokens to every request. Five modest MCP servers added 5,000–7,000 more. By th

We Taught a Snowflake Warehouse to Judge World Cup Conviction and Write the Verdict Back to Solana Soumyadeep Dey Soumyadeep Dey Soumyadeep Dey Follow Jul 12 We Taught a Snowflake Warehouse to Judge World Cup Conviction and Write the Verdict Back to Solana # devchallenge # weekendchallenge # snowflake 5 reactions 1 comment 16 min read

This is a submission for Weekend Challenge: Passion Edition What I Built I'm in Port...

Submission: Edu-Insight Assistant What I build I built the Edu-Insight Assistant, a tool designed for educators to bridge the gap between complex school management data and actionable insights. It allows teachers to query students performance data using natural language, turning educational evaluation into a conversation rather than a manual data-processing task. Demo 🔗 Link: Passion-challenge How I Built It I utilized Next.js for a responsive, performant frontend and hooked it up to Google Gemini 3.5 API. The core logic involves a server-side API route that takes a teacher's natural language questions, prompt Gemini to generate the necessary SQL, and execute that query against a database. This architecture makes data exploration accessible to non-technical educators. Prize Categories: - Best Use of Google AI : Leveraged Gemini 3.5 Flash for natural language-to-SQL translation and result interpretation. - Best Use of Snowflake: Designed with an extensible data layer ready for production-scale analytical workloads in Snowflake.

Enhancing CI/CD and E2E Testing with Sentry Integration in tvview TL;DR: I integrated Sentry for error tracking and improved End-to-End (E2E) testing in the tvview project, enhancing the CI/CD pipeline. This resulted in a score increase from 85 to 95+. The Problem The tvview project lacked comprehensive error tracking and E2E testing, making it difficult to identify and resolve issues in production. The existing CI/CD pipeline needed improvement to ensure smoother deployments and better code quality. What I Tried First Initially, I focused on setting up E2E tests using Vitest, but encountered issues with the test configuration. I also attempted to integrate Sentry, but faced challenges with the DSN (Data Source Name) configuration. The Implementation Step 1: Configuring Sentry To integrate Sentry, I created separate configuration files for the client, edge runtime, and server: // sentry.client.config.ts import * as Sentry from " @sentry/nextjs " ; Sentry . init ({ dsn : " https://385038c88b6eb6ddac52d05a144ab8c1@o4511628189630464.ingest.us.sent " , // Additional configuration options }); // sentry.edge.config.ts import * as Sentry from " @sentry/nextjs " ; Sentry . init ({ dsn : " https://385038c88b6eb6ddac52d05a144ab8c1@o4511628189630464.ingest.us.sent " , // Additional configuration options }); // sentry.server.config.ts import * as Sentry from " @sentry/nextjs " ; Sentry . init ({ dsn : " https://385038c88b6eb6ddac52d05a144ab8c1@o4511628189630464.ingest.us.sentry.io " , // Additional configuration options }); Step 2: Enhancing CI/CD Pipeline I updated the .github/workflows/ci-e2e.yml file to include Sentry configuration and E2E testing: name : 📺 CI + E2E — TVView on : push : branches : [ main ] workflow_dispatch : {} schedule : - cron : " 35 6 * * *" jobs : build-and-test : runs-on : ubuntu-latest steps : - name : Checkout code uses : actions/checkout@v2 - name : Install dependencies run : npm install - name : Generate Prisma client env : DATABASE_URL : " postgre

Stop Guessing: How I Pick AI API Architecture at Every Scale I've been on both sides of this. Two years ago I was the lone backend engineer at a Series A startup, duct-taping API calls together at 2 AM because the founders wanted a chatbot demo by morning. Last quarter I sat in a procurement meeting at a Fortune 500 where we spent six weeks evaluating three vendors for a single inference workload. Same job title on LinkedIn, wildly different problems. Most AI API guides I've read treat both scenarios like they're the same conversation. They're not. The startup CTO optimizing for burn rate and the enterprise architect worrying about a 99.9% uptime SLA are solving fundamentally different equations. After enough of these conversations, I've developed a framework I'd like to share — and yes, I'll talk about Global API because it's what I actually use, but I'll also explain the reasoning behind each choice so you can adapt it to your own stack. What I Look at First: The p99 Question Before I look at price, I look at the latency distribution. Specifically, the p99. Mean latency tells you almost nothing useful. If your median response is 200ms but your p99 is 4 seconds, your users will see janky behavior on the long tail and you won't know why until production is on fire. For startups in the MVP phase, you can usually get away with best-effort routing. A p99 of 2-3 seconds is fine if you're building an async summarization feature. But the moment you put AI in the synchronous request path — like a customer-facing chatbot or a real-time code suggestion — p99 starts to bite. I learned this the hard way when our startup's "AI assistant" feature had users complaining about slowness that I couldn't reproduce locally. The culprit? Provider cold starts hitting our 1% of users who happened to get routed to a freshly spun-up instance. For enterprises, p99 isn't a nice-to-have, it's a contractual obligation. Most B2B SLAs I've negotiated pin uptime at 99.9% and require reporting on m

Enhancing CraveView's CI/CD Pipeline with Sentry and E2E Tests TL;DR: I upgraded CraveView's CI/CD pipeline by integrating Sentry for error tracking and implementing End-to-End (E2E) tests, boosting the score from 85 to 95+. This technical deep-dive explores the architecture decisions, code changes, and lessons learned. The Problem The initial problem wasn't a single error message but a series of inefficiencies in the CI/CD pipeline. The existing setup lacked comprehensive error tracking and test coverage, leading to potential issues in production. Specifically, the pipeline didn't have: Robust Error Tracking : No integrated system for capturing and analyzing errors. End-to-End Tests : Limited test coverage, which could lead to undetected issues in production. What I Tried First Initially, I focused on enhancing the test suite. I explored various testing frameworks but decided to implement E2E tests using Vitest, given its compatibility with the existing tech stack. The first approach involved setting up a basic E2E test framework. However, I encountered issues with the test environment configuration, particularly with database connectivity. The tests required a realistic database setup, which wasn't properly simulated. The Implementation Step 1: Configuring Sentry To integrate Sentry, I created configuration files for client, edge, and server initialization: sentry.client.config.ts import * as Sentry from " @sentry/nextjs " ; Sentry . init ({ dsn : " https://385038c88b6eb6ddac52d05a144ab8c1@o4511628189630464.ingest.us.sentry.io/4511629 " , // Additional config options }); sentry.edge.config.ts and sentry.server.config.ts follow a similar structure, adjusted for their respective environments. Step 2: Implementing E2E Tests I added a new test file e2e-production.test.ts in src/__tests__ : import { test , expect } from ' @playwright/test ' ; test ( ' should render the homepage ' , async ({ page }) => { await page . goto ( ' https://craveview.vercel.app ' ); await expe

Disallow: GPTBot is a wall. Walls don't pay rent, and the crawlers that matter most either ignore them or route around them. If your content is worth training on, the interesting question isn't "how do I keep the bots out" — it's "what do they owe me, and how do I say so in a way a machine can read." That's what RSL (Really Simple Licensing) is for. It shipped 1.0 in December 2025 with around 1,500 publishers behind it — Reddit, Yahoo, Quora, O'Reilly, Medium, Vox. This post is a from-scratch walkthrough of what the format actually is, the six places you can put it, the one mistake that makes crawlers silently ignore your terms, and where the declaration stops and enforcement begins. No tooling required to follow along — it's all plain XML and HTTP. The format is an XML vocabulary, not a config file An RSL document says: for this content, here's what's permitted, what's prohibited, and what it costs. Minimal example: <?xml version="1.0" encoding="UTF-8"?> <rsl xmlns= "https://rslstandard.org/rsl" max-age= "7" > <content url= "/" > <license> <permits type= "usage" > search </permits> <prohibits type= "usage" > ai-train </prohibits> <payment type= "crawl" > <amount currency= "USD" > 0.015 </amount> </payment> </license> </content> </rsl> Read it out loud: search engines may index this; training on it is prohibited; if you want to crawl it anyway, the rate is $0.015. usage tokens include search , ai-train , ai-use (inference/grounding), and a few more. You can scope rules by user and geo too. One rule that trips people up: prohibition wins . If the same token shows up under both permits and prohibits , the content is prohibited. Don't try to express "allowed except for X" by listing X in both — just prohibit X. The namespace is the thing crawlers actually key on The single most common way to publish RSL that quietly does nothing: getting the namespace wrong. It must be exactly: xmlns="https://rslstandard.org/rsl" http instead of https , a trailing slash, or a plausible

I built a voice assistant that controls 45 AI tools. I say "Hey Jarvis" and it executes. What It Does Command Action "generate content" Creates YouTube scripts for 9 channels "research quantum computing" Deep research via Tavily + AI "write email about meeting" Drafts email, copies to clipboard "start focus" Starts Pomodoro + blocks apps "code review" Reviews git diff with AI "summarize" Summarizes clipboard content "find file tax PDF" Natural language file search Architecture Mic → Whisper (offline) → Intent Classify → Router → Ollama → say (TTS) Key Features Offline speech (Whisper local) Wake word: "Hey Jarvis" Global hotkey: Ctrl+Space Command chaining: "research AI then write blog" Memory across conversations Hindi + English Setup brew install portaudio pip install SpeechRecognition pyaudio openai-whisper python voice_commander_pro.py 🔗 github.com/amrendramishra/ai-tools 🌐 amrendranmishra.dev

hi, i am Aniruddha Adak, a full-stack developer from kolkata who spends way too much time building things with ai tools, shipping apps, and reading way too many github readmes at 2 am. i built 27 apps in 45 days using no-code and ai tools last year. that experience taught me one thing very clearly: the landscape of ai tooling for developers is moving insanely fast, and it is genuinely hard to keep up. so i sat down and did something about it. this is my deep research post on every ai tool, agent, builder, reviewer, and framework that developers, software engineers, and ai engineers should actually know about right now. i have organized it into categories so you can find what you need quickly. no fluff. just the tools, their sites, and what they do. why i wrote this i keep seeing developers waste time because they do not know the right tool exists. someone is manually reviewing pull requests for a week straight, not knowing coderabbit exists. someone else is hand-writing supabase schemas when emergent can do it in seconds. another person is spending days on a landing page when v0 can scaffold it in one prompt. this post is my attempt to fix that. i went through github repositories, dev communities, product hunt launches, and research aggregators to compile this. it is long. that is intentional. bookmark it. section 1: ai-native ides these are not just editors with a chatbot plugged in. these are environments built from the ground up around how language models think and work. tool site what it does cursor https://www.cursor.com forked vscode, codebase-aware context windows, multi-file edits with copilot-style background indexing windsurf https://windsurf.com cascade ai agent that writes files, runs terminal checks, and fixes things in real-time zed https://zed.dev built in rust with gpui, super low latency, native multiplayer coding support replit https://replit.com cloud ide with a full autonomous agent that runs inside serverless virtual workspaces google antigravit

Most things that ship under "browser MCP" are the same thing wearing different names: an autonomous agent with a do-anything tool, pointed at your browser, told to figure it out. The pitch is capability. The unspoken cost is that a runtime which can do anything can be steered into doing anything. We just published the opposite, and we published it in the open. github.com/e2llm/e2llm-sifr is now the canonical home for SiFR - the format spec, the taxonomy, the MCP server manifest, real page captures, per-client configs, and the model skill. MIT-licensed. The capture engine and the server stay a hosted product; the format and the interface are open. This post is about why that split is the whole point. E2LLM is not an agent This comes first because everything else follows from it. An agent decides and acts on its own. It plans, it loops, it takes steps toward a goal with you out of the path. That autonomy is the feature - and it is also the attack surface. A runtime that can do anything is a runtime that can be talked into anything. E2LLM is a perception layer, not an agent. It gives whatever model you already use senses for the browser: structured sight, and a small set of narrow, individually-gated actuators. It does not plan, does not loop, does not decide. Your model does the reasoning. E2LLM reports what a page is and carries out one explicit instruction at a time. Nothing runs while you look away. Perception substrate versus autonomous runtime. That line is the design, not a disclaimer on top of it. What SiFR is - and the three things it isn't SiFR (Salience-Indexed Flat Relations) is the capture format at the center of E2LLM. From a distance it can look like a tidy DOM dump or an accessibility tree. Mechanically it is neither, and the difference is the entire value. Not a DOM dump. A dump serializes the tree as-is: everything, in document order, noise included. SiFR selects and ranks. It scores every node by salience, drops scaffolding, and flattens the survivor

Engineering prompts inside your source code is a recipe for deployment fatigue. If you've spent any time moving an AI feature from a prototype to production, you know the specific frustration of 'prompt drift.' You make a subtle tweak to a system instruction—perhaps changing how the model handles edge cases in JSON formatting—and suddenly you're forced into a full CI/CD cycle. A PR, a review, a build, and a deployment, all because of three words changed in a long string constant. In a mature engineering organization, your application logic should be decoupled from your prompt instructions. The code handles the orchestration, the plumbing, and the security; the prompts represent the dynamic configuration. This is what LLMOps aims to achieve, but until recently, there was a massive friction gap between managing these prompts in a dashboard and actually using them inside an agentic workflow. This is where the Humanloop MCP server changes the interaction model entirely. It's not just about having a central repository for strings; it's about bringing those strings into your execution context—your IDE, your Claude instance, or your Cursor agent—as actionable tools. The Architecture of Prompt-as-a-Service The core idea here is treating prompts as versioned assets rather than hardcoded constants. By using the Humanloop API via MCP, you're essentially turning prompt management into a service call. When I look at the toolset available in this server, the first thing that stands out isn't just the ability to read data—it's the ability to manipulate state. Take upsert_prompt for instance. You aren't just fetching text; you can create or update configurations directly from your agent. This transforms your development loop. Instead of context-switching between a browser tab with Humanloop and a terminal, you can instruct an agent to 'Refine the customer-support-reply prompt to be more concise and save it.' The agent performs the engineering work and updates the source of truth in