标签：#ai

Originally published on angeo.dev . Full tables, p-values and the sealed plan are there. Most claims about AI visibility are untestable by design: publish the signals, wait, attribute anything good that happens to the signals. I wanted a version I could not fudge, so I wrote the analysis plan first, hashed it, and sent the hash to the other party before I had any data. The question Do businesses AI assistants name repeatedly differ, on observable technical signals, from businesses the same assistants name once ? Every business in the corpus was named at least once, so this says nothing about how to enter an answer. It compares repeat against one-off mentions inside a named-business corpus. Four signals, all externally observable: Signal Check Crawler access Does robots.txt block any of 8 AI crawlers Content map Does the site serve /llms.txt Structured data Does a product page emit JSON-LD Product Buyability Does that node carry offers.availability Study setup The answers came from a partner (connexion.me), who ran 44 product-level home-decor buying questions across ChatGPT, Gemini and Perplexity, twice, in two arms — 264 answers per arm. Blinding was deliberate. I did not write the questions and did not see their store list until my plan was sealed; they never saw my frame, my scan results or my thresholds. Roster rows 669 no resolvable domain -186 resolved to a different company -3 marketplaces and listing surfaces -12 duplicate rows collapsed -10 Unique domains analysed 458 scanned successfully 455 Cases: 3+ mentions across both runs and present in both. Controls: exactly one mention across both runs. Head excluded first — anything in 53+ of 264 answers (Amazon, Etsy, Wayfair, Target, Home Depot). The pre-registration Sealed 10 August, SHA-256 9b4ccf12629e… : Under 15% of named businesses would be Magento No signal would separate the groups by more than 15 points Refutation condition: any signal differing by 20+ points with the named group higher Result — generic

You probably don't need one expensive AI subscription. You need the right AI stack. AI subscriptions have quietly become another monthly expense. One tool for writing. Another for research. Another for coding. Another for image generation. Another for PDFs. Before you realize it, you're paying for several AI services every month — even though you use only a fraction of their capabilities. ChatGPT Plus alone is currently $20/month. That's $240 a year before adding anything else. But here's the interesting part: Do you actually need to pay for all of it? I decided to approach the problem differently. Instead of looking for one “best” free AI tool, I built a free AI stack where different tools handle different jobs. The goal isn't to prove that free AI is better than paid AI. The goal is much more practical: How much of a paid AI workflow can you realistically replace with free tools? The Biggest Mistake: Looking for One AI to Do Everything This is how most people use AI: Open ChatGPT → ask everything → hit usage limits → consider upgrading. But AI tools are increasingly specialized. A research engine doesn't need to be your coding assistant. A coding model doesn't need to be your web-search engine. A writing assistant doesn't need to be your data-analysis environment. Instead of asking: “Which free AI is the best?” Ask: “Which free AI is best for this particular task?” That simple change makes the free ecosystem much more powerful. My Free AI Stack Here's the architecture I would use for a zero-subscription workflow. Task Free Option Why General AI assistant ChatGPT Free Everyday questions and writing Web research Perplexity Free Search + citations Coding Gemini / free coding tools Code generation and debugging Research & experimentation Google AI Studio Model experimentation Microsoft workflow Copilot Free Web-based assistance Private/offline AI Ollama / LM Studio Local inference The important point is that these tools don't have identical capabilities or limits. For

Vibe coding has gotten a bad reputation as lazy, AI-driven coding, but that's not where it came from.

The famous American philosopher Smashmouth once said that the years start coming and they do not stop coming. If you follow the artificial intelligence industry, you have never felt the weight of those words more deeply than right now. For years, science fiction fans have watched Tony Stark talk to Jarvis, his brilliant, autonomous, and conversational AI assistant. We watched Jarvis build 3D models, hack into secure mainframes, and seamlessly manage Stark's entire life. We all thought that level of technology was decades away. But over a span of just three days in September 2026, the entire landscape of technology completely shifted. Anthropic launched two massive models. Meta dropped a frontier model so cheap it is practically free. And OpenAI released GPT 6 Astra in an event so chaotic it literally took down the internet. Grab a coffee and buckle up. We are going to break down this insane week of AI news, dive into the real engineering breakthroughs, and figure out if we just witnessed the birth of the real Jarvis. The Day The Internet Went Dark Usually, the first week of September is quiet in the tech world. But this year, AGI apparently waits for no one. Right before OpenAI was scheduled to announce GPT 6 Astra, something bizarre happened. ChatGPT, Claude, Grok, and Cursor all went completely dark at the exact same time. The most logical explanation is a massive Azure cloud infrastructure outage. But the timing was so perfectly cinematic that people immediately started joking that Astra's first act as a public model was to assassinate its competitors. When the lights finally came back on, the OpenAI rollout was spectacularly messy. They published their launch page, major news outlets released their carefully written embargoed stories, and then, for reasons nobody fully understands, OpenAI ripped the page down for 90 minutes. Tech influencers immediately began playing the ultimate status game, flexing about how long they secretly had early access to Astra. Meanwh

Hello, I'm Shrijith Venkatramana, and I'm building LiveReview — a blast-radius aware AI code review built for your business-critical systems. Star us to help devs discover the project, give it a try, and share your feedback to help improve the product. A language model can write Python, explain quantum mechanics, and imitate Shakespeare. Show it a screenshot of a production dashboard, however, and suddenly the central question becomes: How does a transformer that was trained on text learn what a pixel means? The naïve answer is: “Give the image to the LLM.” That description hides almost all of the interesting engineering. Modern multimodal systems are usually compositions of several models: a vision encoder turns pixels into vectors, a connector translates those vectors into something the language model understands, and the LLM then reasons over the resulting representation alongside ordinary text tokens. That architectural trick has turned the transformer from a language architecture into something much closer to a general-purpose interface for heterogeneous data. The evolution is worth understanding because it reveals a useful engineering pattern: you often do not need to retrain a giant model to give it a new sensory modality. You need a good representation and a sufficiently expressive interface between representations. 1. The basic mental model: pixels become tokens Start with an ordinary LLM. Its input looks conceptually like: "The server returned HTTP 500. What should I check?" | v tokenizer | v [t1, t2, t3, ..., tn] | v Transformer | v answer Everything is eventually represented as vectors. Multimodal transformers exploit this fact. An image is first converted into a sequence of vectors: image | v vision encoder | v [v1, v2, v3, ..., vm] | v multimodal connector | v [z1, z2, z3, ..., zk] | +------ text tokens [t1, t2, ...] | v LLM | v answer The important conceptual shift is this: The LLM does not have to understand pixels directly. It only has to understand

Every small business has one person who is the office search engine. Where is the refund form. What goes in the Friday report. Which supplier do we use for rush jobs. The answers are written down somewhere, but asking that person is faster than finding them, so the questions keep coming and that person never gets a full hour of their own work. An SOP chatbot fixes exactly that. Staff type the question, and the bot answers with the steps from the procedures you already wrote, and shows which document it took them from. Nothing else. This article explains how that works using one picture, shows two bots we actually run, and is honest about what the bot will never do. The box Picture the AI as a very capable temp worker who shows up every morning with no memory of your business at all. Not the address, not the prices, not how you handle a late delivery. Smart, fast, and completely blank. Before you ask them anything, you hand them a box. In the box are your documents. The rule, taped to the lid, says: answer only from what is in the box, and if the answer is not in the box, say so. That box is what people in the AI world call the context. Everything the bot knows about you at the moment it answers is what you put in the box for that one question. It does not learn your business over time. It reads the box, answers, and forgets. Next question, new box. Two things follow from this picture, and they explain almost everything about SOP chatbots. The box has a size. Anthropic, the company behind the Claude models, says in its engineering write-up on contextual retrieval that a knowledge base under about 200,000 tokens, roughly 500 pages, can simply be included with every question, with no extra machinery. Most small businesses have far less than 500 pages of procedures. So for most of you, the whole manual fits in the box every time. If the manual is bigger than the box, someone has to pick. Then a librarian step runs first: it reads the question, pulls the few pages that m

The idea of your child using ChatGPT can be daunting, but OpenAI's new teen-focused tools provide granular controls over what they can and can't access.

The Uber founder has said that Atoms will allow him to complete "unfinished business."

My harness flagged the model for sending the wrong arguments. It compared what the model actually...

To get started with competitive Pokémon battles, all you need is your phone. But to compete at the highest level, like at the Pokémon World Championships, you're going to need a Switch. And there doesn't seem to be a good reason why. Battle simulator Pokémon Champions originally launched on the Switch and Switch 2 in […]

This is a submission for Weekend Challenge: Generosity Edition What I Built A donation ledger for a group too small to buy software. It is a Google Sheet, some Apps Script, and one public page a donor can open. The group I had in mind is the kind that exists on every street: a neighbourhood fund, a school parents' group, a committee that collects for winter coats. Money arrives over WhatsApp and leaves in cash, and somebody keeps it in a notebook. The arithmetic is not the hard part. The hard part arrives three months later when a donor asks where their money went, and answering needs the notebook, the person holding it, and an afternoon. Software for this exists and is priced for organisations with a finance team. So the ledger stays in a spreadsheet a volunteer already knows how to open, and the only thing added is what a spreadsheet cannot do alone: read messy human messages, refuse to trust its own reading , and publish the page that answers the question before it is asked. Demo The public page a donor opens → That page is the deployed page, byte for byte, with one line changed: where the Apps Script version writes <?= data ?> , the demo fetches the same JSON from a file so you can read it without a Google account. The JSON is produced by running the sample month through the same recordEntry() and publicView() the real script uses, so if the ledger rules change, the demo changes with them or the build fails. The sample month deliberately includes the things that go wrong: a receipt two volunteers forwarded, a donation typed with one zero too many and later corrected, and a reading the checks refused to trust. What the model read, and whether it was allowed to count → The ledger page shows the result. This one shows the part worth showing. Pick any of six real donation messages and it highlights the exact characters Gemini says it read the amount from, lists the three checks with their outcomes, and says why the row was posted or held. It is fed by a recorded run

An AI agent that has no job description will invent one. That is the whole reason agents drift, and it is the reason most of the agents I have seen deployed inside Indian businesses are quietly switched off within a few months of going live. Nobody would hire a person, point them at the office, and say "handle things". Yet that is exactly how most owners deploy an agent. They connect it to WhatsApp or email or the accounts folder, give it a paragraph of instructions, and let it run. Then they are surprised when it starts answering questions it was never meant to answer, promising delivery dates it cannot know, or filing something that a human should have looked at first. The fix is not a better model. It is the same discipline you already use for people: defined duties, an escalation path, a probation period with a review date, and one named person who is accountable for it. What drift actually looks like Drift is not a dramatic failure. It is a slow widening of scope that nobody approved. A distributor in the FMCG trade sets up an agent to acknowledge incoming orders on WhatsApp and log them into a sheet. Week one, it does that. Week three, a retailer asks "when will my stock reach?" and the agent, being helpful, answers with a guess. Week five, a retailer asks for a discount, and the agent, having seen discounts mentioned in earlier messages, offers one. None of this was in the brief. All of it followed naturally from "be helpful to customers", which is what the owner wrote because they did not know what else to write. By the time the owner notices, the agent has made commitments in writing to twenty retailers, and the sales team is cleaning up after it. The agent did not malfunction. It did what an unsupervised new employee does: it filled the vacuum with its own judgement. The mistake was upstream, at the moment of deployment. The job description A job description for an agent is not a prompt. It is a one-page document the owner can read and sign off, written in

AI has been known to mess up some pretty basic things, but Claude wants to help you with your emails.

TL;DR 📌 🕰️ Then: I built google-chat-electron by hand , over months , reading tutorial after tutorial. ⚡ Now: I rebuilt the whole thing as google-chat-tauri in less than 24 hours — and I did not write the code. Claude did. 🦀 Plot twist: I don't know Rust. Not a little — at all . The AI wrote every line of it. 📦 Result: a ~ 3 MB Linux installer instead of a bundled Chromium. 🧪 Status: pre-release. Fun project. Stable version coming after real-world testing. Let's dive in. 👇 The Electron Era: Months of Honest, Manual Labour 😅 A few years ago I wanted Google Chat in a real window — with a tray icon, an unread badge and native notifications — instead of a browser tab that disappears among thirty other browser tabs. So I built it. In Electron. By hand. And it took months . Not because Electron is bad, but because every single thing was a tutorial: How do I make a tray icon that actually behaves? How do I keep the app alive when the window closes? How do I intercept a link and open it in the real browser? How do I package a .deb ? A .dmg ? An installer for Windows? Why does this work on my machine and nowhere else? 🙃 Every answer was a blog post, a GitHub issue thread, or a Stack Overflow reply from 2017 that almost applied. It shipped, people used it, and I was genuinely proud of it. The Tauri Rewrite: One Evening, One Prompt Loop 🤖 Last week I opened Claude Code and asked it to port the app to Tauri v2 . I did not open the Rust book. I did not read the Tauri docs. I described what the app should do, reviewed what came back, ran it on my actual laptop, and reported what broke. Timeline: Time What happened 15:42 git init 17:01 Full Electron → Tauri v2 port committed 17:25 Desktop notifications working < 24h later v0.0.1 tagged and released 🎉 38 commits. ~3,600 lines of Rust and JavaScript. Zero lines typed by me. This is what people mean by vibe coding — and honestly, it felt less like programming and more like directing . My job became: describe the behaviour, test it on

This is a submission for Weekend Challenge: Generosity Edition What I Built Every...

AI can write code now. That part is no longer surprising. You can describe a feature to Copilot, Claude Code, Cursor, Codex, or another coding agent and get a working implementation in minutes. Sometimes it is genuinely impressive. But there is a bigger question: Can you actually trust the code enough to ship it? According to the Stack Overflow 2025 Developer Survey, 84% of developers use or plan to use AI tools . At the same time, trust in AI-generated output is still limited. One of the biggest frustrations developers report is getting an answer that is almost right, but not quite . Source: https://survey.stackoverflow.co/2025/ai And that “almost right” part is exactly where developers still matter. AI may write more code. But humans still need to decide whether that code is correct, secure, maintainable, and actually worth merging. So here is a simple review workflow I think every developer should practice. 1. Start With the Requirement, Not the Diff Imagine you tell an AI agent: Add password reset support. A few minutes later, it generates the full feature. The code may compile. The UI may work. The tests may even pass. But before reading the implementation, ask: How long should reset tokens remain valid? Can the same token be used twice? What happens if the email does not exist? Should existing sessions be logged out? Are we exposing whether a user account exists? This matters because AI can build the wrong thing very cleanly. So before asking: Does this code work? Ask: Does this solve the correct problem? That one question can save a lot of time. 2. Check the Architecture Before the Syntax AI is usually good at writing a function. It is not always good at understanding where that function belongs inside your system. For example, an agent might create something like: components/ ├── PaymentForm.tsx ├── PaymentAPI.ts ├── StripeService.ts └── Database.ts Everything may technically work. But should database access really live beside your UI components? Probably no

AI agents don’t use web browsers. They don’t click buttons, submit forms, or trigger standard HTTP requests that pass through your middleware stack. They execute logic via API calls, background queues, or CLI commands using tool definitions. When an LLM decides to "fetch the latest invoices," it usually calls a tool function. If that tool function just runs Invoice::all() , your AI agent just became a god-mode data leak. The fundamental problem with integrating LLMs into existing applications is that agents operate in a detached, stateless execution context . They don't have a session cookie. They don't inherently know who invoked them. If you rely on the system prompt to tell the LLM, "Only show John his own data," you are trusting a probabilistic text generator to enforce your security boundary. That is a production incident waiting to happen. To build a secure AI agent in Laravel, you must treat the LLM not as a user, but as a proxy for the user. The agent must inherit the exact Role-Based Access Control (RBAC) and Attribute-Based Access Control (ABAC) constraints of the human sitting behind the keyboard, and it must enforce those constraints at the database query level, not the prompt level. TL;DR AI agents bypass traditional web middleware because they execute logic through background tools and function calling. Never trust the LLM to filter its own results. Force filtering through Eloquent scopes and authorization gates. Pass the acting user's identity explicitly into the agent's execution context using Laravel's auth guards or custom context DTOs. For complex rules (hierarchies, multi-tenancy, ABAC), standard role packages fall short. Tools like hosseinhezami/laravel-permission-manager are required to evaluate deep permission trees inside agent tools. Audit every tool execution with the acting user's ID, not the system service account. 📋 Table of Contents 1. The "God-Mode Tool" Problem 2. Passing Identity Down the Execution Chain 3. Enforcing RBAC Inside LLM

The first version of a RAG system always looks convincing. You connect a document loader, a vector database, and a large model, ask a question, and the answer comes back with impressive confidence. Then production happens. A support agent asks about a refund policy that changed last week, and the bot answers with the old policy. A user from the finance team sees chunks they should never see. Gemini starts returning 429 errors during a reindex. A 3,000-document ingestion workflow fails at document 2,412, and you have no idea how to resume safely. That is the gap between a RAG demo and a production RAG pipeline. This walkthrough focuses on building a maintainable retrieval-augmented generation pipeline using n8n for orchestration, Qdrant for vector storage and filtered retrieval, and Gemini for embedding and answer generation. The goal is not just “make it answer.” The goal is to make it operable: idempotent ingestion, access-controlled retrieval, retry-safe automation, grounded answers, and a path for evaluation. TL;DR Treat RAG as two separate pipelines : ingestion and query. Store more than vectors in Qdrant: source_id , acl , version , updated_at , chunk_index , and text. Make ingestion idempotent so reprocessing documents does not create duplicate truth. Use Qdrant filters for permissions, freshness, and document status. Force Gemini to answer only from retrieved evidence and return citations. Add retries, timeouts, dead-letter handling, and evaluation before users do the testing for you. 📋 Table of Contents The Production Problem with Demo RAG 1. Split RAG Into Two Pipelines Before You Automate Anything 2. Design the Qdrant Collection Around Access Control and Freshness 3. Chunk for Retrieval, Not for Reading 4. Make Ingestion Idempotent and Resumable 5. Embed in Controlled Batches Without Dropping Documents 6. Retrieve With Filters, Not Blind Similarity 7. Make Gemini Prove It Used the Evidence 8. Add the Production Guardrails: Retries, Timeouts, and Dead Lette

The first version of most AI features is not a system. It is one giant prompt doing too many jobs. It is supposed to understand the user, check account facts, retrieve policies, write a response, avoid legal risk, match the brand tone, and maybe decide whether to escalate. Then one edge case arrives — a refund request with a partially used subscription — and the prompt starts negotiating with itself. That is usually when teams say, “We need better prompting.” Often, the real problem is architectural. A single prompt becomes a god object. It holds competing responsibilities, hidden assumptions, and constraints that are hard to test. Multi-agent orchestration is not about creating a mystical swarm of autonomous bots. It is about doing the boring, Laravel-style thing: breaking a large problem into bounded services, coordinating them with typed contracts, and using queues, events, validation, logging, and failure policies to keep the system honest. Laravel is a good place to build this because it already gives you the pieces: service container, queues, batches, events, validation, HTTP client, rate limiting, caching, structured logging, and database persistence. The hard part is not calling a model. The hard part is coordinating specialists safely. TL;DR A giant prompt becomes fragile when it tries to be researcher, analyst, writer, reviewer, and policy engine at once. Model agents as bounded specialists with explicit inputs, outputs, tools, and permissions. Use a lightweight router to classify work, not to do the work. Coordinate through typed messages, not loose prompt fragments. Use Laravel’s container, queues, batches, validation, and logging to make orchestration operational. Give specialists permissioned tool adapters instead of implicit knowledge. Add budgets, timeouts, retries, and escalation paths from the beginning. Do not use multi-agent orchestration when one deterministic service or one simple prompt is enough. 📋 Table of Contents The Giant Prompt Is a God

The engineering team at software company Figma recently documented how they built AI agents to help their security team investigate alerts, search past incidents, check company systems, and even prepare code fixes. The agents learn from previous investigations, reducing repetitive work and helping engineers resolve complex alerts about 70% faster. By Renato Losio