标签：#X

Tu repo ya es el contexto de tus agentes, lo hayas diseñado a propósito o no Esa frase me costó entenderla. En este post te ahorro el camino. Era Octubre del 2025, trabajando en el monorepo de Skyward con agentes de IA todos los días. Y todos los días la misma rutina: le decía en el prompt al agente "no uses esto", "no hagas esto así", "reutiliza el componente que ya existe". Lo escribía. Lo repetía. El agente hacía exactamente lo que le dije que no hiciera. No era que no me escuchara. Era que leía el código y ahí veía otra cosa. El agente le cree al código, no a tu prompt Un agente sigue los patrones que ve en el repo, no los que tú le dices en el prompt. Y los subagentes son peores, porque arrancan sin el contexto de la conversación. Toda la pelea que tú diste arriba en el chat, para ellos no existió nunca. Entonces pasaba esto. Creaba un componente nuevo aunque ya había uno que resolvía exactamente el problema. No respetaba las normas de diseño ni usaba los design tokens. Seguía documentación obsoleta porque seguía ahí, viva, sin nada que la marcara como obsoleta. Mi primer instinto fue el de todos, meter más contexto en el prompt. Más reglas, más "por favor no hagas esto", más ejemplos pegados a mano. Funcionaba a medias, y para la siguiente tarea había que volver a agregarlo todo. Hasta que llegaba el siguiente subagente y empezaba de cero. En algún momento, cansado de repetirme, entendí lo obvio. El agente no me estaba desobedeciendo. Estaba leyendo el repo y haciendo caso a lo que el repo decía de sí mismo. Si conviven el componente bueno y tres versiones viejas, no tiene cómo saber cuál es el oficial. Si la doc dice una cosa y el código hace otra, le va a creer al que esté más a mano. Está haciendo justo lo que le pedí. El mismo repo es el contexto que usan los agentes. Si está mal estructurado, las respuestas no van a ser buenas. Punto. No hay prompt que arregle un repo que se contradice a sí mismo. Screaming Architecture me llevó hasta media cancha Lo prim

On June 12, 2026, enterprise developers using the Codex API started seeing an unfamiliar response header: X-Model-Version: kindle-alpha . It appeared on a subset of requests for roughly 18 hours, then vanished. That's the release candidate for GPT-5.6 — OpenAI's next flagship model — leaking through the staging layer. OpenAI's Chief Scientist publicly called the upcoming release "a meaningful leap" the following day. By OpenAI's historically understated communications standards, that's loud. This post covers what the backend traces, developer reports, and Polymarket odds (currently ~80% for a pre-June-30 launch) actually tell you about the model — and what to do before it drops. How the Leak Surfaced Three separate sources converged in the 72 hours after the June 12 header incident. First, developers with ChatGPT Pro OAuth access reported hitting context windows significantly beyond GPT-5.5's supported limit. At least four documented cases logged successful 1.5M-token completions before the backend silently downgraded them to the production model. Second, the Codex enterprise API logs — accessible with full response header exposure enabled — confirmed the kindle-alpha codename across US-east-1 and us-west-2 endpoints. Third, the Polymarket market for "GPT-5.6 public release before July 1, 2026" moved from 61% to 80%+ within 48 hours of the header reports circulating on developer forums. None of this is from OpenAI's press office. No model card, no official benchmark numbers, no pricing. The specifics below are high-confidence inference from multiple corroborating signals — not official spec. Treat it accordingly when making production decisions. The Architecture Shift: Agentic-First, Not Just Smarter GPT-5.5 was trained as a reasoning model with agent capabilities added on top. GPT-5.6 is reportedly designed in the opposite order. The primary optimization target during training was not MMLU or GPQA benchmark scores — it was token efficiency on long-horizon agentic t

Most AI coding tools assume you're sitting in front of a repo. There's a working directory, some source files, tests, maybe a CI pipeline. The AI reads your code, suggests changes, ru****ns tests. Great model — if you're a developer writing code on a Tuesday afternoon. Now picture the other scenario. It's 2am. PagerDuty fires. You SSH into a box you haven't touched in three months. Something is broken, you're not sure what, and the runbook was last updated by someone who left the company in 2022. You're not thinking about repos. You're thinking: what OS is this thing running? What just failed? Is it safe to restart that service or will I make it worse? These are two fundamentally different workflows. But almost every AI terminal tool I've seen is built for the first one. The 30 minutes nobody builds for There's a ton of tooling for the world before you log into the box: Prometheus, Grafana, PagerDuty, incident.io, runbooks, dashboards. All useful. No complaints. But there's this practical 30-minute window after you SSH in where you're basically doing archaeology with journalctl and grep. You check systemd. You look at disk. You read logs that were clearly written by someone who hated future-you. You copy-paste terminal output into Slack so your teammate can squint at it from a different timezone. This is where I think AI could actually help. Not by replacing Grafana. Not by building another dashboard. Just by being present in the shell while you're debugging. Please, for the love of uptime, don't replace my shell I've seen a few AI terminal projects that basically build an entire new terminal experience from scratch. New keybindings, new UI, new everything. Here's the thing: ops people have muscle memory. We have aliases we wrote in 2019 and forgot about. We have tmux configs we'd defend with our lives. We SSH through jump hosts with key forwarding chains that barely work but have worked for years so nobody touches them. If your AI tool requires me to abandon all of

Plaintiffs in the class action complaint allege Rivian falsely promised for years it would bring hands-free driving to its first-generation R1 vehicles.

Expo SDK 56 ships with a custom Kotlin compiler plugin that eliminates reflection from Expo Modules on Android. The result: 70% faster module initialization and a 30% reduction in time to first render. The plugin runs during compilation, so app developers get these performance gains automatically without changing any code. Module authors can unlock even bigger wins with a single annotation. This post walks through how we built it and why this approach succeeded where previous attempts failed. For the Swift side where we now talk to JSI directly, check out our companion post Talking to JSI in Swift . The reflection problem we inherited Before Expo Modules, we had Unimodules. They worked like old React Native bridge modules: you'd sprinkle annotations across methods you wanted to expose, and the runtime would discover everything through reflection. class ClipboardModule ( context : Context ) : ExportedModule ( context ) { override fun getName () = "ExpoClipboard" @ExpoMethod fun getStringAsync ( promise : Promise ) { val clip = clipboardManager . primaryClip ?. getItemAt ( 0 ) promise . resolve ( clip ?. text ?. toString () ?: "" ) } @ExpoMethod fun setStringAsync ( content : String , promise : Promise ) { clipboardManager . setPrimaryClip ( ClipData . newPlainText ( null , content )) promise . resolve ( true ) } } Reflection made sense when we needed metadata about our own code. What methods does this module export? What arguments do they accept? The JVM could answer those questions. But reflection costs time, and on Android that time comes straight out of your startup budget. Every module the runtime introspects adds milliseconds before users see your app. Building the Expo Modules API gave us a chance to fix this. We wanted better ergonomics and less reflection. The Kotlin DSL delivered both in one move, removing most reflection while making modules easier to write. But we couldn't eliminate all of it. Type information for function arguments and Record properties s

This week's tooling landscape is quieter on the AI-native side but dense with infrastructure moves that affect how AI-driven workloads actually run in production. Cloudflare's Workflows scaling overhaul is the clearest signal: agent-triggered execution is now an assumed pattern, not a novelty, and platforms are rearchitecting accordingly. The rest of the week rounds out with a kernel maintenance drop, a meaningful abstraction removal in tRPC, and a Biome beta that's finally making ESLint replacement feel plausible. Linux 7.1 Released with Driver and Networking Fixes 7.1 is a maintenance release. No architectural changes, no new subsystems—just patches you should care about if you're running affected hardware or kernel-adjacent tooling. The two fixes worth flagging are heap overflows in the USB serial io_ti driver ( get_manuf_info() and build_i2c_fw_hdr() ), plus memory leak corrections scattered across drivers and networking subsystems. Trace tooling also gets updates, which matters if you're doing kernel-level performance analysis on production systems. One operational note: Torvalds is traveling, so merge window latency may be irregular. If you're tracking pull request timelines for custom kernel builds, plan for slippage. Verdict: Ship — if you're on 7.0 and running USB serial hardware or affected networking paths, upgrade on your normal kernel cycle. No breaking changes, no new dependencies, nothing to validate beyond your existing regression suite. tRPC Drops Abstraction Layer for React Query This is the kind of change that looks small in a changelog and feels large in daily development. The new tRPC client exposes native TanStack Query interfaces— QueryOptions and MutationOptions —directly, rather than wrapping them in tRPC-specific hooks. The practical effect: if you're already using TanStack Query elsewhere in your app, you stop context-switching between two similar-but-different mental models. You call .queryOptions() and .mutationOptions() factories and pa

One previously unreported SpaceX investor has ties to Chinese military contractors.

For the past four years, a sprawling Android-based botnet called Popa has forced millions of consumer TV boxes to relay Internet traffic linked to advertising fraud, account takeovers, and mass data-scraping efforts. This week, researchers from multiple security firms concluded that the Popa botnet is linked to NetNut, a "residential proxy" provider operated by the publicly-traded Israeli firm Alarum Technologies Ltd [NASDAQ: ALAR].

A data breach involving government-issued ID documents affects over three million people in Texas.

General Intuition is in talks to raise around $300 million at a roughly $2 billion valuation from backers including Jeff Bezos. The startup trains AI agents on spatial-temporal reasoning.

Over the last several months, XRP Ledger (XRPL) has fundamentally shifted in how amendments move from concept to mainnet. Historically, amendment development was largely focused on functional correctness, performance testing, traditional security audits, bug bounties and independent validator testing as the last line of defense to catch security vulnerabilities. As XRPL continues to grow in complexity and the value secured by the network increases, we recognized that the previous model was no longer sufficient. Advances in AI are also rapidly reducing the cost of vulnerability discovery, making it increasingly important to identify issues as early as possible in the development lifecycle. With that in mind, we set out to establish a stronger, repeatable, defense-in-depth model that makes it increasingly difficult for critical vulnerabilities, consensus risks, and feature interaction bugs to reach mainnet. The result is a significantly higher bar for amendment activation that combines specification rigor, adversarial testing, multiple independent audits, attackathons with expert security researchers, AI-assisted security reviews and phased deployments. The Lending Protocol ( XLS-66 ) and Single Asset Vault (SAV) - XLS-65 are among the first major amendments to undergo this full review process, making them some of the most rigorously tested amendments in XRPL's history. They also represent some of the most significant new financial capabilities added to the XRP Ledger since 2012, introducing native primitives for lending and borrowing built around Single Asset Vaults. Together, the Lending Protocol and Single Asset Vault bring lending and borrowing capabilities directly into the core XRPL protocol, advancing XRPL's capabilities for Institutional DeFi. Lending Protocol Security and Quality Gates This report provides transparency into the development and security process behind one of the most financially complex features XRPL has ever shipped. As context, the Lending P

Netflix has detailed a cloud-based system for scaling camera file processing across global film and TV workflows. The pipeline handles ingest, validation, metadata extraction, and media transformation at scale using FilmLight API and distributed compute. It standardizes workflows across editorial, VFX, and color pipelines, improving consistency and reducing manual handling across productions. By Leela Kumili

📝 Originally published in Japanese on Zenn. This is the English version. Canonical: https://zenn.dev/uya0526_design/articles/satellite2_haiku-coaching 📚 This is satellite article #2 in my "Read-Aloud Speed Meter dev log" series. For the whole picture, see the main article ; for the AmiVoice integration, see satellite #1 . Where This Sits This article covers the part of the read-aloud speed meter that hands the measured numbers to Claude Haiku to generate coaching as "one compliment + one improvement." Many articles that use generative AI just dump it on the model: "here's the recognized text, evaluate it nicely." This article is the opposite. Don't let the LLM do the math. All computation and rule-based decisions are settled in code, and Haiku only does the wording. I'll go through how I designed this "two-stage" split, and which swamps I sank into during implementation. Four things: Why "code does the math, Haiku only does the wording" (the role-split design) The finalized prompt, and the messages / system / cache_control implementation First-hand findings: the prompt cache that didn't work The moment real data slapped me with "written in the prompt ≠ obeyed" 💡 I'm an ex-Java engineer learning TypeScript in public, so I drop in Java comparisons. Why Not Let Haiku Do the "Math"? Up front: Haiku is more than enough for feedback generation — in fact, you can shape the task into something Haiku is good at. The key is the role split. I settle all the metric computation (speaking speed, stagnation rate, threshold decisions) entirely in code. So by the time it reaches Haiku, it's already settled facts like this (below is from running labelMetrics on a real measurement — reading the Heike sample during development; stagnationRate is a number for percentage display): { "pureSpeakingSpeed" : 322 , "pureSpeakingSpeedEvaluation" : "slightly fast" , "stagnationRate" : 0 , "stagnationRateEvaluation" : "few" } Haiku's only job is to translate these numbers and labels into warm, c

📝 Originally published in Japanese on Zenn. This is the English version. Canonical: https://zenn.dev/uya0526_design/articles/satellite1_amivoice-bff 📚 This is satellite article #1 in my "Read-Aloud Speed Meter dev log" series. For the whole picture, see the main article . Where This Sits In the read-aloud speed meter app, this article covers the part that sends browser-recorded audio to the AmiVoice API to get back recognized text plus timestamps. The theme is calling an external API without exposing your API key to the browser — in other words, implementing a BFF (Backend for Frontend). The main article only touched the highlights, so here I go down to a level you can reproduce yourself. Specifically, four things: Why you must not call AmiVoice directly from the browser (why a BFF is needed) AmiVoice's synchronous HTTP auth, parameters, and multipart order Why the browser's MediaRecorder output (WebM/Opus) passes through as-is Reshaping the raw JSON with a pure-function mapper and testing it with fixtures 💡 I'm an ex-Java engineer learning TypeScript in public, so I drop in comparisons to Java here and there. Why a BFF Is Needed Using the AmiVoice API requires an API key. And that key must never appear in browser-side code. Frontend JavaScript is fully inspectable by the user, so writing the key there leaks it instantly. So I insert a relay that holds the key. [Browser] ──audio Blob──▶ [Next.js API Route (BFF / holds key)] ──▶ [AmiVoice API] record / display audio field reshape into u / d / a speech → text The browser only calls my own API Route ( /api/recognize ), and that Route attaches the key server-side and forwards to AmiVoice. The key is just read from process.env and never ends up in the bundle shipped to the browser. ☕ Java comparison: This is the same as a Spring @RestController reading an external API key from application.yml (env vars) and relaying without showing it to the client. Think "a thin Servlet that hides the secret and relays an external API."

📝 Originally published in Japanese on Zenn. This is the English version. Canonical: https://zenn.dev/uya0526_design/articles/main_article_reading-speed-meter Introduction — What I Built I built a web app that lets you read a Japanese passage aloud, measures your speed and fluency, and has an AI coach return a one-line piece of feedback. 🌐 Demo: https://reading-speed-meter.vercel.app/ 📦 Repository: https://github.com/uya0526-design/reading-speed-meter The flow is simple. You read a passage aloud into the mic (up to 10 seconds) while looking at the script — I prepared the opening lines of two Japanese classics, The Tale of the Heike and Hōjōki — and when you press "Measure," : AmiVoice API recognizes the audio, the code computes your pure speaking speed (characters/min) and stagnation rate from that result, and Claude Haiku returns coaching as "one compliment + one improvement." (Measurement starts on a button press after recording — it never runs automatically.) This article aims to be a single, self-contained piece covering the whole picture, the design decisions, and the reproduction steps . 💡 Where this sits in my journey I'm an ex-Java engineer learning TypeScript and Python in public. This was my first project where I deliberately adopted "AI-collaborative development" as a clear mode. Throughout, I'll drop in comparisons to Java — hopefully useful for anyone coming from a similar background. What You'll Get From This Article How to design evaluation logic that fully exploits the per-word timestamps AmiVoice returns How to build a BFF (Backend for Frontend) so API keys never reach the browser A "two-stage" design where code does the math and Claude Haiku only does the wording First-hand findings you only learn by verifying — e.g., "I thought I'd optimized it, but it wasn't actually working" I include concrete endpoints, parameters, and environment variables so you can reproduce it yourself. My Learning Style (AI Transparency) 💡 Learning companions & how this art

The company has identified at least 13 instances where its robotaxis drove into highway sections closed for construction.

Hello Dev Community! 👋 It is officially Day 41 of my continuous streak toward full-stack MERN engineering! Yesterday, I migrated my codebase from native Node boilerplate to Express.js. Today, I dived straight into the absolute core mechanism that makes Express so incredibly powerful in Prashant Sir's (Complete Coding) masterclass : Middlewares . Before today, I thought requests hit an endpoint and immediately returned a response. Today, I learned how to intercept, inspect, and modify that request before it ever reaches the final route handler! 🧠 Key Learnings From Node.js Lecture 9 (Middlewares) A middleware is essentially a function that executes during the Request-Response cycle, having full access to the req , res , and the next middleware function in line. Here is the technical breakdown: 1. The Anatomy of Middleware Unlike a standard route handler that takes (req, res) , a middleware takes a third powerful argument: next . If you don't invoke next() , your request will hang forever and the browser will eventually timeout! 2. Built-in vs. Custom Middleware Custom Middleware: Wrote my own custom functions using app.use((req, res, next) => { ... }) to act as a security guard or global logger. Built-in Middleware: Explored how Express natively handles data types using structures like express.json() and express.urlencoded() , which automatically parse inbound request bodies so we don't have to manually handle streams anymore! javascript const express = require("express"); const app = express(); // Custom Global Logging Middleware app.use((req, res, next) => { console.log(`[${new Date().toISOString()}] ${req.method} request to ${req.url}`); next(); // Pass control to the next handler in line! }); app.get("/dashboard", (req, res) => { res.send("Welcome to the secure dashboard layer!"); }); app.listen(8000);

Peter Parker to Bruce Banner: "I didn't know you could get that big."

The problem I kept running into I'm a chronic tab hoarder. At any given time I've got 40–80 tabs open across two windows. Chrome's built-in Memory Saver is aggressive in the wrong ways — it hibernates tabs I'm actively referencing. And the built-in task manager is a two-step detour that still doesn't tell me which tabs I should actually close. So I built Tab Memory Manager. What it does Per-tab memory estimates — A live MB count next to every open tab. Sorted by memory usage by default. There's a live total on the toolbar icon so you always know what Chrome is consuming right now. Smart suggestions — The extension flags your biggest, stalest tabs: ones that are idle the longest and consuming the most. It never suggests your active tab, pinned tabs, tabs playing audio, or domains you've whitelisted. Hibernate, don't close — This was the core design decision. Hibernating frees the memory but keeps the tab alive in your strip — it reloads when you click it. Much safer than closing, especially mid-research. Bulk cleanup — Select multiple tabs or hit Apply on the suggestions panel. See the total memory you'll reclaim before you commit. Undo list — Closed something by mistake? There's a "Recently cleaned" panel. One click to restore. Tab grouping — Groups all your open tabs by domain into color-coded Chrome tab groups, instantly. The interesting technical bit: memory estimates Chrome's stable extension API doesn't expose exact per-tab memory. The chrome.processes API that does exists only on Dev and Canary builds — not the Chrome that 99% of people use. So Tab Memory Manager uses calibrated estimates based on tab state, domain patterns, and known Chrome process overhead. These are clearly labeled "est." in the UI. If you're on Dev or Canary, you can switch on real per-tab memory in settings. The warning Chrome shows about "processes requires dev channel" is a Chrome-generated note about that optional API — the extension works completely normally without it. It's not a bug

Dialogue management is the process of tracking conversational state and deciding what an agent should say or do next. Classical systems split this into isolated modules: natural language understanding, dialogue state tracking, a policy engine, and response generation. Large language models can collapse these boundaries into a single inference step, but doing so reliably requires careful architecture choices. This article examines practical patterns for using LLMs as dialogue managers, with a focus on structured reasoning, tool use, and cost-efficient inference. What Is LLM Dialogue Management? An LLM-based dialogue manager treats conversation as a partially observable decision process where the model itself reasons over history, user intent, and available actions. Instead of hand-written rules or separate slot taggers, the model receives the full transcript, a system prompt defining the task, and optionally a schema of tools it can invoke. The model then emits either natural language or structured JSON representing the next system action. This approach excels in open-domain or rapidly changing domains where maintaining a rigid ontology is impractical. Architecture Patterns for LLM-Based Dialogue Most production implementations fall into one of four patterns. The right choice depends on how much control you need over state transitions and how willing you are to trade complexity for flexibility. End-to-end generation. The LLM receives the full chat history and outputs the next response. It works well for unstructured chit-chat but can hallucinate state or ignore business rules without additional guardrails. Structured state extraction. The LLM is prompted to output a JSON object representing dialogue state, such as slots, user intent, and confirmed facts. A lightweight policy layer reads this state to decide whether to ask a question, call an API, or close the task. This separates reasoning from control and makes debugging easier. Tool-augmented manager. The LLM uses