标签：#AI

Qwen's new image paper is easy to read as another benchmark bump. Qwen-Image-2.0-RL takes the existing Qwen-Image-2.0 model, runs a reinforcement-learning pass on top, and reports better scores: 57.84 on Qwen-Image-Bench, up 2.61 points from the base model. Its text-to-image arena Elo moves from 1115 to 1193. Its image-editing arena Elo moves from 1256 to 1349. Those are the headline numbers. They are not the useful part. The useful part is the training story underneath them. The paper is a good reminder that "just optimize the reward" is a dangerously incomplete sentence, especially when the model is not an LLM and the output space is a whole image. The model got better, but not by one simple trick Qwen-Image-2.0-RL is a post-training pipeline for a diffusion image model. In plain English: the base model already knows how to generate and edit images. The RL stage tries to steer it toward outputs humans prefer, including better prompt following, better aesthetics, better portrait fidelity, and more reliable editing. The team builds task-specific reward models. For text-to-image, those rewards cover alignment, aesthetics, and portrait quality. For editing, they cover instruction following and face identity preservation. Then they train with a GRPO-style setup adapted for flow-matching diffusion models. If you only squint at that, it sounds like the same broad recipe people use for language models: generate candidates, score them, push the model toward the better ones. The paper is more interesting because it shows how fragile that story becomes once you touch the actual training loop. The CFG detail is the first real lesson Classifier-free guidance, usually shortened to CFG, is one of those diffusion-model knobs that users mostly experience as "make the image follow the prompt harder." Under the hood, it changes how the model samples. The Qwen team tested three ways to use it during RL. Using CFG during both rollout and training made the images collapse into incohere

Want to go back to your ex... AI assistant? We don't blame you. Take these steps to get Google Assistant back after your fling with Gemini.

The startup, Proception, is taking a unique approach to collecting training data to tackle one of the hardest problems in robotics: hands.

Burn-in is often overblown, but understanding the phenomenon is still important.

Pocket sells a $129 credit card-shaped puck, which sticks to the back of your phone, and promises unlimited recordings, transcriptions, and to-do items.

Quando comecei a trabalhar com aplicações descentralizadas há mais de uma década, lembro bem da frustração de pagar US$ 50 em taxas de transação para mover alguns tokens na rede Ethereum durante um pico de congestionamento. Era um problema técnico que ameaçava inviabilizar todo o ecossistema. Hoje, observo com entusiasmo profissional como as soluções de Layer 2 transformaram radicalmente esse cenário, abrindo portas para casos de uso que antes eram economicamente impraticáveis — especialmente aqui no Brasil, onde a tokenização de ativos e os pagamentos em stablecoins crescem em ritmo acelerado. O problema fundamental: o trilema da escalabilidade Para entender por que as soluções de segunda camada são tão importantes, precisamos compreender o trilema da blockchain proposto por Vitalik Buterin. Uma rede precisa equilibrar três pilares: descentralização, segurança e escalabilidade. O Ethereum, em sua arquitetura original, priorizou os dois primeiros, processando apenas cerca de 15 a 30 transações por segundo (TPS) na camada base. Para se ter dimensão, redes de pagamento tradicionais como a Visa processam milhares de transações por segundo. Quando o DeFi explodiu em 2020 e 2021, e novamente com o boom dos NFTs, a rede simplesmente não dava conta da demanda. As taxas de gas dispararam, e usuários comuns foram literalmente expulsos pelo custo. Em meus projetos de consultoria, atendi empresas brasileiras que desistiram de iniciativas Web3 justamente porque os custos operacionais inviabilizavam o modelo de negócio. A pergunta que sempre me faziam era: "Como cobrar R$ 5 de um cliente se a taxa da transação custa R$ 30?". A resposta estava — e está — nas camadas de segunda geração. Como funcionam as soluções de Layer 2 O conceito central das soluções de Layer 2 é elegante: em vez de processar todas as transações diretamente na blockchain principal (Layer 1), executamos a maior parte do processamento "fora da cadeia" e depois enviamos apenas uma prova compacta de volta para o

A CFO will not act on a number an LLM eyeballed. They will not act on a number the model "estimated" by reasoning over a usage dump. And they should not — because the moment a language model emits a dollar figure it computed itself, that figure is a guess wearing the costume of a fact. This is the design constraint behind databricks-cost-leak-hunter , the pilot skill of the databricks-pack v2 rebuild shipped in the claude-code-plugins marketplace ( PR #906 ). Given a live, authenticated Databricks workspace, it surfaces real cost leaks across four named categories, ranks them by monthly dollar impact, and emits a report a finance reader can act on. The marketplace validator graded it B (88/100, zero errors). The SKILL.md is 329 lines. The single most important thing in it is a rule the model is structurally prevented from breaking: the LLM never does the dollar arithmetic. Why not just let the agent read the bill and summarize it? Because that is exactly how you ship a confidently wrong cost report. Hand a model a few thousand rows of system.billing.usage and ask it for the top cost leaks, and it will give you a fluent answer. It will add DBUs. It will multiply by a price it half-remembers. It will round. Every one of those steps is a place the model can be plausibly, invisibly wrong — and the output reads identically whether the math is right or hallucinated. The failure mode of an LLM doing FinOps is not a crash. It is a clean, well-formatted, wrong number. The fix is architectural, not prompt-engineering. The model is allowed to decide what to look for and how to explain it . It is never allowed to be the calculator. The dollar primitive: confirmed, never estimated Every confirmed figure comes from the customer's own billing tables — system.billing.usage joined to system.billing.list_prices . Not a model estimate. Not a public price list. The number Databricks actually billed. That join is defined once, as a priced CTE, and reused by every category query. Usage i

Most of the time, when I need to inspect a complex JSON payload, I copy the raw string from my terminal or network tab, open a browser tab, and paste it into one of the many "JSON Formatter" sites that clutter the first page of Google. It’s a ritual we all do. We paste, we click "Format," and we wait. For small payloads, this is fine. But when you are debugging a massive API response, a deeply nested configuration file, or a large dataset, that ritual breaks down. The browser freezes. The site asks you to upload a file. Worse, many of these tools send your data to a server for processing. If that JSON contains API keys, user PII, or internal schema definitions, you are essentially trusting a third-party service with your proprietary data every time you hit "pretty print." I got tired of the latency and the privacy overhead. So I built JSONForge . The core premise is simple: do everything locally. No server-side processing. No file uploads. No network requests for the core logic. Everything happens in your browser, powered by WebGPU for heavy lifting and a small model that runs in your browser for schema inference. The WebGPU Advantage JSON parsing is computationally cheap for a modern CPU, but rendering and diffing large structures is not. When you have a 5MB JSON file, the DOM manipulation required to display it as a tree view can cause significant jank. By offloading the parsing and formatting logic to the GPU via WebGPU, JSONForge handles massive payloads without blocking the main thread. You can open a file, click "Pretty Print," and see the result instantly, even if the file is hundreds of kilobytes or larger. The UI remains responsive because the heavy computation is parallelized on the graphics card. This also means the tool works offline. If you are on a plane, or your internet drops in the middle of a debugging session, your toolkit doesn’t vanish. You can continue to diff, validate, and format without interruption. Schema Generation Without the Server Roun

Omen AI raised a $31 million Series A to monitor chip coolant and stop bacterial outbreaks in data centers.

From knowing to doing — Phase 4 of the Jarvis AI Platform The Problem with Knowledge-Only AI After Phase 3, Jarvis could remember you across sessions and search your documents. But it still had a fundamental limitation. You: "What is the weather in Kathmandu right now?" Jarvis: "I don't have access to real-time weather data." You: "What is 2847 × 391?" Jarvis: "The answer is approximately 1.1 million." ← WRONG An AI that only knows things from training data is useful. An AI that can do things is transformative. That is what Phase 4 built. What Is a Tool Engine? A tool engine gives the AI model the ability to call real functions during a conversation. The flow looks like this: User: "What is the weather in Kathmandu?" ↓ AI Model ↓ "I should call WeatherTool" ↓ WeatherTool.getWeather("Kathmandu") ↓ "22°C, Clear sky, Humidity: 45%" ↓ AI Model ↓ "The weather in Kathmandu is 22°C and clear." The key insight: the AI decides when to call a tool and with what input . We don't hardcode "if user asks about weather, call WeatherTool." The model figures that out from the tool descriptions we provide. The Architecture Decision The most important architectural decision in Phase 4 was the package structure. ai . jarvis . tools / ├── JarvisTool . java ← marker interface ( root ) ├── ToolRegistry . java ← manages all tools ( root ) ├── builtin / ← built - in tools │ ├── DateTimeTool . java │ ├── CalculatorTool . java │ ├── WeatherTool . java │ └── WebSearchTool . java └── mcp / ← MCP protocol └── McpServerConfig . java Why not put tools inside ai/ ? The ai/ package handles HOW Jarvis talks to AI models. Tools define WHAT Jarvis can do. These are fundamentally different responsibilities. Mixing them would mean every new tool requires changes to AI infrastructure code. Keeping them, separate means adding a new tool requires exactly one file. The JarvisTool Pattern Every tool in Jarvis implements one interface. /** * Marker interface for all Jarvis tools. * Spring auto-discovers all @C

We (at DEV and MLH) are covering AI Engineer's World Fair by printing a physical newspaper called...

Ending any story is hard, but that's especially true of mystery-packed TV shows. Series like Lost initially hook viewers with constantly building secrets and questions, to the point that they can often seem incomprehensible. But the promise is that it will all pay off in the end - a feat that few shows ultimately manage. […]

Google was forced to cap Meta's use of Gemini AI due to a lack of capacity.

This virtual panel brings together AI security experts to examine the evolution of AI-driven threats, from prompt injection and data poisoning to agent abuse and AI-powered social engineering. The discussion explores emerging attack patterns, incident response challenges, and the changes security teams must make as AI systems become more autonomous and integrated into critical workflows. By Claudio Masolo, Elham Arshad, Sabri Allani, Vijay Dilwale, Igor Maljkovic

FIFA says hydration breaks protect players from heat. They also create new annoying commercial breaks—and fans are calling foul.

You know what you want from a JSON document. You just don't want to memorize whether it's $[?(@.age > 18)] or $..users[?(@.active)] . Plain English in. JSONPath out. JSONPath Assistant on FormatList lets you paste JSON, describe what you need in natural language, and get a validated JSONPath expression plus the actual results — all in your browser. No account, no API key, no data sent to a server. How it works Paste your JSON — an API response, config file, or test fixture. Describe what you need — e.g. "get all user names" or "find products with tag tech". Generate — the assistant reads your JSON structure and maps your query to JSONPath. Validate & copy — the expression runs against your data immediately. Copy the path or the matched values in one click. If the first attempt returns no matches, the tool retries with a simpler variation automatically. Example queries You type Generated JSONPath Get all user names $.users[*].name Find users older than 18 $.users[?(@.age > 18)] Get names of active users $.users[?(@.active == true)].name Find products with tag tech $.products[?(@.tags.indexOf('tech') >= 0)] Get all order prices $.orders[*].price Get the first user's email $.users[0].email Get all pod names {.items[*].metadata.name} Get all pod IP addresses {.items[*].status.podIP} The tool ships with one-click examples for each of these — load one, hit Generate, and see how it works before trying your own JSON. What kinds of queries it understands Property access — "get all names", "list order prices" Numeric filters — "older than 18", "price under $10", "greater than 100" Boolean filters — "active users", "enabled devices" Tag / category searches — "with tag tech", "beauty category" Multi-condition — "both tech and mobile tags" Quantifiers — "the first user", "the last item" Kubernetes — "get all pod names", "pod IP addresses" It analyzes field names in your JSON — so users , products , orders , or whatever keys you actually have — and builds paths that match your sc

I used to think AI search readiness was just SEO with a new name. It's not. The more time I spend on this, the clearer the distinction becomes. The core difference Traditional SEO optimizes for ranking in a list of links. You want to be the #1 blue link on Google for "best project management software." The user clicks through to your page, you get the traffic, you monetize. AI search optimizes for being the source of an answer. When someone asks Perplexity or ChatGPT "what's the best project management software?", the AI reads multiple sources, synthesizes an answer, and cites the ones it used. The user may never click through. The fundamental units are different: SEO operates on pages and rankings AI search operates on facts, claims, and citations You can be #1 on Google for a keyword and never appear in a single AI-generated answer. And you can be cited in AI answers without ranking in the top 10 for anything. What still matters Some things carry over from SEO: Technical quality — Fast pages, HTTPS, crawlable content. AI crawlers care about this just like Googlebot. Clear content structure — Headings, lists, tables. Well-structured content is easier for AI models to parse. Internal linking — AI crawlers follow links like any other crawler. Good information architecture matters. Backlinks from authoritative sources — Being cited by Wikipedia, academic papers, and major publications signals trust to AI models just like it does to search engines. What matters for AI search that barely matters for SEO A few things that are critical for AI search but don't move the needle much for traditional rankings: LLMs.txt / LLMs-full.txt — These files don't affect your Google ranking at all. But they give AI models a clean, structured map of your site. I've seen sites with great LLMs.txt files get cited more consistently than sites with better backlink profiles but no AI-readable summary. Structured data for disambiguation — In SEO, schema markup helps with rich snippets. In AI s

AI crawlers now appear in many server logs alongside traditional search bots. Some are used for search retrieval, some for training, and some for broader web indexing. If you care about AI search visibility, you need to know which ones can access your public pages. The most common accidental blocker is simple: a robots.txt rule or CDN bot setting that prevents AI crawlers from reaching the content you want discovered. The major AI crawler tokens to check Here are crawler tokens you may see in logs or robots.txt rules: Crawler token Company Notes GPTBot OpenAI Documented OpenAI crawler token OAI-SearchBot OpenAI Documented OpenAI search-related crawler token ChatGPT-User OpenAI Documented OpenAI user-triggered agent token ClaudeBot Anthropic Documented Anthropic crawler token Claude-SearchBot Anthropic Documented Anthropic search-related crawler token Google-Extended Google Google control token for Gemini Apps and Vertex AI use CCBot Common Crawl Web corpus crawler used by many downstream systems PerplexityBot Perplexity Commonly referenced Perplexity crawler token Crawler names and purposes change. Always confirm against official platform documentation before making sitewide access decisions. First, check what is actually happening Before you change anything, find out who is already crawling. If you have server logs: grep -E "GPTBot|OAI-SearchBot|ChatGPT-User|ClaudeBot|Claude-SearchBot|Google-Extended|CCBot|PerplexityBot" access.log If you use Cloudflare, check bot and security events and filter by user agent. Three quick diagnostic steps: Open https://yourdomain.com/robots.txt and look for broad Disallow: / rules. Confirm the sitemap is listed in robots.txt or discoverable at /sitemap.xml . Use our AEO Checker to validate robots.txt and flag restrictive AI crawler rules. The most common mistake The blunt rule that makes sites invisible to many crawlers: User-agent: * Disallow: / This blocks every well-behaved crawler that follows the wildcard rule. If you see it on

OpenAI is investigating issues with Codex usage limits Tibo wrote that the Codex team spent Sunday in a war room, digging through logs and looking for anything that could have caused faster usage drain for some users. As the investigation continues, OpenAI has issued a full reset of Codex usage limits for everyone. The funny part: this week at OpenAI is called RESET week. In US corporate culture, that usually means a lighter week to slow down, clear the calendar a bit, and recharge.

Hello Devs 👋 When teams talk about engineering metrics, the conversation usually moves toward speed....