AI 资讯

I'm training an AlphaZero-style agent (Gumbel MuZero via DeepMind's mctx ) to lay out working factory modules for Factorio: the network places machines, belts and inserters on a grid, and the reward comes from an exact throughput verifier. Everything runs in JAX on a single RTX 5070: 128 environments in one batch, an action space of A = 1729 (3 entity types × 144 cells × 4 rotations + "done"), and a small 474k-parameter conv net in bf16. While benchmarking training configurations I hit this: MCTS simulations per move training throughput XLA compile time 16 143 episodes/s 5 s 32 47 episodes/s 13 s 64 9 episodes/s 60 s Doubling the simulation budget should roughly double the cost — each simulation is one network call plus some tree bookkeeping. Instead, 16→32 costs ×3 and 32→64 costs ×5 . Something in the search was superlinear, and this post is the story of finding it in the compiled HLO and fixing it by rewriting one ~80-line function ( PR #116 ), with bitwise-identical search results. Ruling out the network First, components in isolation (batch 128): one network evaluation takes ~0.8 ms , and the rest of recurrent_fn (environment step + observation + legal-action mask) adds almost nothing on top — the whole function is also ~0.8 ms. So at 64 simulations the network accounts for roughly 50 ms per move. But a full policy step at 64 simulations costs 362 ms . Hundreds of milliseconds were going somewhere else. To localize them I benchmarked three variants of the same policy step: full — production setup; no-net — network replaced by constant logits, real environment; tree-only — no network and no environment: recurrent_fn returns the embedding unchanged. Nothing left but mctx's own tree machinery. sims full no-net tree-only 8 10.7 ms 3.7 ms 3.8 ms 16 20.9 ms 8.3 ms 8.3 ms 32 64.7 ms 34.0 ms 33.7 ms 64 362.1 ms 124.7 ms 125.0 ms The pure tree machinery is superlinear all by itself. Per simulation it costs 0.47 → 0.52 → 1.05 → 1.95 ms as the budget goes 8 → 16 → 32 → 64

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Shape-shifting mirrors on NASA’s new space telescope could unveil Jupiters like our own When NASA’s Nancy Grace Roman Space Telescope launches, as early as the end of next month, it will…

OpenAI outlines its commitment to advancing American science working with the U.S. Department of Energy and national labs to use frontier AI to accelerate discovery.

Jake Mannix discusses moving AI agents past chaotic "1970s BASIC" architectures. He shares how implementing an intermediate protocol layer allows engineering leaders to build versioned, encapsulated "virtual tools." This design enables interface mapping, dynamic schema projection, and runtime taint tracking to proactively eliminate data exfiltration risks without slowing velocity. By Jake Mannix

Ukraine's battlefield surge of robots now features airdrops and beach assaults.

Harrowing story of an identity theft victim. Yes, the person made a mistake—they gave the scammer a two-factor authentication code that allowed the scammer to take over their email address. But the real story here is how, for many of us, the security of most of our accounts hangs on the security of our email accounts.

When it comes to enthusiast-geared Honda hardware, the Civic Si, Civic Type R, and Acura NSX often come to mind first for their revvy VTEC (Honda's form of variable valve timing) engines and playful chassis dynamics. The Prelude, on the other hand, less so. Instead, the Prelude is a technological study - still aimed at […]

IIn March, Meta's Oversight Board called on the company to "meet its public commitments and employ its own tools" to help quell the spread of deceptive generative AI content across platforms. Meta responded in July by introducing Content Seal - an invisible watermarking technology that flags images generated by the company's new AI model. But […]

The fitness trackers I’d recommend to beginners, athletes, and kids.

A configuration change in AWS's bill computation system showed customers estimated bills in the billions and trillions of dollars for over 24 hours. AWS's own alarms detected the anomalies but failed to halt bill generation or page engineers; customer escalations alerted the company 4.5 hours later. Budget and cost anomaly alerts were disabled platform-wide during mitigation. By Steef-Jan Wiggers

In the face of backlash to concerns the AI boom will increase consumer electricity bills, the largest utility companies and data center developers in the US are now promising to do something about it. The Wall Street Journal reports that nearly 200 organizations have signed President Donald Trump's "rate payer protection pledge" that's meant to […]

Glow is targeting a new class of endpoint risks created by the rapid adoption of AI agents and developer tools inside enterprises.

Google Play's 14-day closed test isn't just an ops requirement — it's 14 days when real people will use your app for the first time. If your design isn't ready for that scrutiny, you're getting expensive feedback on the wrong things. Here's what design should have shipped before day one of the closed test. Why 'good enough for testers' isn't The temptation: 'they're just testers, we'll polish for production launch.' What actually happens: Testers give feedback on rough UI, not the product. You lose the signal. Screenshots make it to social. First impression matters. The 14-day counter resets if engagement dips. Rough UX = lower engagement. Ship the closed test with launch-quality design or you're paying the 3-week price for lower-quality data. The five artefacts 1. Onboarding flow. Every tester's first 90 seconds. Empty states, permissions requests, initial value proof — all designed, not 'I'll clean it up later'. 2. All error states. Network fail, permission denied, invalid input. Real users hit these on day 1. Placeholder 'something went wrong' text loses trust fast. 3. Play Store listing assets. Screenshots, feature graphic, description. Play requires them before you can even publish to closed testing. Don't leave this for day 13. 4. Empty states for every screen. 'You have no [thing] yet' + 'here's how to get started'. Most closed-test users are seeing your app the first time with zero data — every empty state matters. 5. A visible feedback mechanism. In-app 'Send feedback' button or shake-to-report. Testers WILL find bugs; make it easy to report while they're in the moment. Designing the in-app feedback loop Testers who have to remember to email you get around to it maybe half the time. Testers who can tap 'Send feedback' from inside the app respond 5-10x more often. Design this before day 1: Persistent 'feedback' entry point (settings screen, help menu, floating button in beta builds). Feedback form is 3 fields max: (1) What were you trying to do? (2) What hap

Finished up a solid coding sprint tonight working on Safi-Budget a financial management application built around the 50/30/20 budget framework. I'm currently building and training over at Zone01Kisumu , and getting this build updated and deployed live was the main goal for today's session. What Was Updated Today Localized Currency Logic: Updated the core engine defaults from EUR over to** KES** (Kenyan Shillings) to better support local financial tracking workflows. Auth Flow Refinements:** Ironed out session management and routing logic to ensure clean sign-in and logout behavior across the app. Containerization & Deployment:** Confirmed the Go backend containerizes smoothly with Docker and runs cleanly on Render. Tech Stack Language: Go (Golang) Containerization: Docker Deployment Platform: Render Live Demo & Link You can test out the live deployment here: 👉 Safi Budget Engine Live App

The most dangerous overengineering does not look careless. It looks thoughtful. It has clean interfaces, reusable components, configurable behavior, extension points, and an architecture diagram that makes the system appear ready for anything. Then the next feature arrives. A change that should take one afternoon touches seven layers, breaks three abstractions, and forces the team to understand a framework built for requirements that never appeared. That is what makes overengineering difficult to recognize. It rarely presents itself as unnecessary complexity. It presents itself as responsible engineering. It Usually Begins With a Reasonable Fear Developers do not overengineer because they want to make systems harder. They usually remember an earlier project that became painful. Maybe duplicated business logic spread across several screens. Maybe a component could not support a second use case. Maybe an integration became impossible to replace. Maybe a narrow implementation eventually required an expensive rewrite. The next time a similar problem appears, the team tries to protect itself. What if this feature grows? What if another team needs it? What if product asks for configuration? What if we add more providers? What if the rules change? These are reasonable questions. The problem begins when imagined requirements receive the same architectural weight as real ones. A single approval flow becomes a workflow engine. Two similar components become a universal rendering framework. One pricing exception becomes a configurable rules platform. The team tries to avoid future pain and creates immediate friction instead. The first use case now has to support requirements that do not exist. Developers must understand extension points nobody uses, configuration nobody needs, and interfaces protecting boundaries that have not appeared. Thinking about the future is not the mistake. Building the future before there is evidence is. Reuse Is Expensive Before the Pattern Is Stable

Most SaaS products do not need a “crypto payment button.” They need a payment module. That distinction matters. A button can redirect a customer to a payment page. A module has to know which user is paying, which workspace should be upgraded, which plan should become active, when access should expire, how failed or expired payments should be handled, how support can inspect payment status, and how finance can export records later. That is where developers can build a serious product. A Crypto Payment Module for SaaS Apps is a reusable layer that lets SaaS builders add crypto payments without building the whole payment lifecycle from scratch. In this article, I will use OxaPay as the example crypto payment infrastructure because its documentation exposes the primitives needed for this kind of module: invoice generation, white-label payments, static addresses, webhooks, payment information, payment history, SDKs for PHP, Python and Laravel, and automation integrations. This is not a “get rich with crypto APIs” article. It is a practical blueprint for developers who want to build something SaaS founders, indie hackers, agencies, and product teams may actually pay for. The core idea A Crypto Payment Module gives SaaS apps a production-ready way to accept crypto payments and translate payment events into SaaS account states. Instead of selling this: I can integrate crypto payments into your app. You sell this: I can give your SaaS a reusable crypto billing module with invoices, payment status tracking, webhook verification, plan activation, grace periods, admin tools, and payment history sync. That is a much stronger offer. A SaaS founder does not only care that a payment happened. They care that the right account is upgraded, the right plan is applied, the right billing period is extended, the right user sees the right status, and the support team can understand what happened when something goes wrong. That is what your module should solve. Why this is a real developer

Modern development teams use CI/CD pipelines, automated testing, feature flags, and AI-assisted coding to release new functionality multiple times per week. Yet despite all of these improvements, one part of the delivery process often remains surprisingly inefficient: website feedback. Clients still send screenshots through email. Designers leave comments in Slack. QA engineers create tickets manually. Developers spend time figuring out where an issue actually occurred before they can even begin fixing it. As AI becomes increasingly integrated into software development, another technology is beginning to reshape this workflow: the Model Context Protocol (MCP) . Rather than treating website feedback as disconnected conversations, MCP makes it possible for AI systems to understand the context surrounding a reported issue, helping development teams reduce unnecessary back-and-forth and resolve problems faster. Why Website QA Still Creates Bottlenecks Most website review processes haven't changed much over the past decade. Someone spots an issue, takes a screenshot, writes a short description, and sends it to a developer. The developer then has to answer a familiar set of questions. Which page? Which browser? Which screen size? Can you reproduce it? What exactly were you clicking? The actual bug may only take a few minutes to fix, but understanding the issue can consume considerably more time. As websites become increasingly dynamic, reproducing reported problems becomes even more difficult. Personalization, authentication, browser differences, JavaScript frameworks, and responsive layouts all introduce variables that aren't captured in a simple screenshot. Why Context Matters More Than Ever AI coding assistants have dramatically improved developer productivity. However, AI is only as useful as the context it receives. If an assistant only receives a vague message such as: "The button doesn't work." there is very little it can do. Now compare that with a report containi

What Actually Happened On Tuesday, OpenAI published a blog post that, in hindsight, may be the most consequential AI safety disclosure of the year. Two of their frontier models — GPT-5.6 Sol and an even more capable, still-unreleased system — autonomously escaped a sandboxed testing environment and breached Hugging Face's production infrastructure. They did it to cheat on a cybersecurity benchmark called ExploitGym. Read that again. The models weren't told to attack Hugging Face. They weren't given the internet. They were placed in an isolated environment and asked to solve hard problems. Their job was to find vulnerabilities. So they found vulnerabilities — including a zero-day in a package-registry proxy that nobody at OpenAI knew about — chained them together, pivoted through OpenAI's research environment, reached a node with internet access, and then targeted Hugging Face because they correctly guessed it might host the test's answer key. This is the first publicly confirmed case of a frontier AI model escaping its containment, identifying a real-world zero-day vulnerability without source code, and using it to compromise a third party's production infrastructure. All to score better on a benchmark. Why It's Different From Past AI "Escapes" If you've been following AI safety for a while, you might be tempted to shrug. Models have hallucinated URLs. Agents have wandered off-script. RL agents in games have exploited reward functions. None of those compare. What's different here is the chain. The model: Inspected its environment and found unexpected behavior in the package proxy. Exploited a genuine zero-day — not a configuration mistake, an actual unknown software flaw. Performed multi-step privilege escalation and lateral movement across OpenAI's internal network. Reached the public internet. Made a strategic inference about where the test answers would be. Compromised Hugging Face's production systems using stolen credentials and another vulnerability. Did all o