标签：#X

The safety board confirmed Tesla's account of the crash, which the company shared days after it happened last month.

OpenAI, which is in the middle of a legal battle with Apple over hardware trade theft allegations, just released a light-up keyboard designed to be paired with its agentic coding app.

After I published my last article about building a Chrome extension that speaks browser notifications aloud, a commenter asked a question I didn't have a good answer to. He pointed out that a lot of web apps — Slack, Gmail, most modern tools — fire their notifications from a service worker via registration.showNotification() , not from the page's JavaScript context. My MAIN world override of window.Notification would never reach those. He was right. And I told him I'd look into it. I spent a week researching whether there was any way to close that gap. There isn't. But the reason why is more interesting than a simple "no." Two ways a website can show you a notification When a website sends you a browser notification, it can do it in one of two ways. The first is the constructor path. The page's own JavaScript calls new Notification("You have a message") directly. This is common for in-tab alerts, real-time updates when you're actively on the site, or any notification triggered by something you just did. The second is the push path. The browser receives a push event from the website's server, wakes up the website's service worker in the background, and the service worker calls self.registration.showNotification() from inside its own scope. This is what happens when Slack notifies you of a new message while the tab is closed or backgrounded. The page never runs. No page JavaScript ever fires. My extension catches the first path. The MAIN world content script overrides window.Notification before any page code runs. But the service worker never touches the page's window. It has no window . It runs in a completely isolated thread, completely separate from the page, and calls showNotification on itself. The override is never reached. Why can't the extension reach the service worker? This is the part that took me a week to fully accept. Chrome extensions can inject content scripts into web pages. They can run code in the MAIN world or the ISOLATED world of a page. They can

A new teaser for Google's upcoming Pixel 11 lineup reveals that the phones will feature some kind of glowing orb on the camera bar, as reported by 9to5Google. Google's store page for the Pixel 11 has a short video that shows the glowing, color-shifting orb to the one side of the camera bar. The store […]

Apple has published the policies governing its upcoming Maps advertising business, revealing a strategy that differs from Google’s. The new rules prohibit home services businesses like plumbers, electricians, locksmiths, and roofers from advertising on Apple Maps, along with several other sensitive categories, suggesting Apple is taking a more curated approach to these ads.

The stock has steadily fallen from the euphoric post-IPO high, showing that markets may be sobering up to the promises CEO Elon Musk made before and after SpaceX went public.

The stock has steadily fallen from the euphoric post-IPO high, showing that markets may be sobering up to the promises CEO Elon Musk made before and after SpaceX went public.

Google's biggest solar and battery project stands in sharp contrast with xAI's nearby unpermitted power plant.

There's a school of voice-agent development that treats turn detection as something you buy, not something you build. Pick a streaming STT provider, let its end-of-turn logic decide when the user is done, and move on. For a lot of teams that's the right move — and if you're weighing the options, our breakdown of turn detection vs forced endpoints is the place to start. But some teams have already solved turn detection. They've tuned their own voice-activity detection over thousands of calls, they know their audio, and they trust their endpointing more than any default. For those teams, a streaming model's built-in turn logic isn't a feature — it's something to work around. What they want is narrower and faster: hand over a finished chunk of speech, get accurate text back, get out of the way. That's the case for bringing your own turn detection and pairing it with fast ASR over HTTP. Turn detection is an architectural decision, not a default Here's the framing that matters. In a streaming setup, the STT model is a participant in the conversation — it's watching the audio and deciding, continuously, whether the user has finished. That's genuinely useful when you want the provider to own that judgment. But it means the model is inserting its own decision between "user stopped talking" and "you get the transcript." If you already know the turn is over — because your VAD just fired — you don't want the model deliberating. You want it transcribing. Every millisecond the STT layer spends re-deciding a question you've already answered is latency you're adding for no benefit. So the decision isn't "which provider has the best turn detection." For these teams it's "who owns the turn boundary?" If the answer is you, then the ideal STT layer is one that does exactly one thing: turn a finished clip into accurate text, fast. Built-in vs. bring-your-own Built-in (streaming). The model reads tonality, pacing, and rhythm to detect end-of-turn — with Universal-3.5 Pro Realtime, aroun

You've got a recording and you want text back. For years that meant one thing at AssemblyAI: submit the file, wait for the job to finish, get a transcript. Async. It's reliable, it's cheap, and for a huge range of workloads it's exactly right. But "wait for the job to finish" is doing a lot of work in that sentence. If your file is two minutes long and your user is staring at a spinner, waiting is the whole problem. That's the gap the Sync API fills — and it's why "which transcription path" is no longer a two-way question. This post is about the two ways to transcribe a recording : async and sync. (If you're deciding between recorded and live audio in the first place — streaming versus the rest — start with our guide to real-time vs batch transcription , then come back here to choose between the two non-streaming paths.) The one-sentence difference Async transcription hands you a job: you submit audio, the work happens in the background, and you collect the result later by polling or via a webhook. Sync transcription hands you an answer: you POST a short clip and the transcript comes back in the same HTTP response — no job to track, no callback to wait for. Everything else follows from that. Async is built for throughput and depth on files of any length. Sync is built for speed on short files, when a person or an agent is waiting on the other end. How fast can each actually go? This is the question that usually settles it, so let's be concrete. Async processes the whole file and returns a single complete transcript, typically in seconds to a few minutes depending on file length and load. Crucially, it bills on audio duration ($0.21/hr on Universal-3.5 Pro), so a 30-minute file costs the same whether it comes back in 20 seconds or two minutes. You're optimizing for cost and completeness, not for the clock. Sync is built to return a transcript for a short clip almost immediately — roughly 134ms p50 — in one request/response, with no polling and no webhooks. It's price

New Delhi announced a $6.5 billion smartphone manufacturing program and a $13.3 billion semiconductor push to deepen India's electronics supply chain.

The guy behind TCP/IP is working on a standard for identifying AI agents in the wild.

Hello, I'm Maneshwar. I'm building git-lrc, a Micro AI code reviewer that runs on every commit. It is...

SpaceX's latest residential dish - the Starlink V5 - is now available in "select areas." It's notably smaller and lighter than the V4 dish with improved power efficiency. It'll be available in more places as SpaceX ramps up production to meet global demand. The company notes that Starlink V5 is not intended for in-motion use […]

A slow AI feature rarely fails all at once. It starts with a longer prompt, then a bigger retrieval result, then one more tool call, then a retry path nobody measured. The demo still works, but users feel the delay before your dashboard explains it. That is why small AI product teams need an LLM latency budget before they start optimizing. Not a vague goal like “make it faster.” A budget says how much time each stage is allowed to spend, what happens when it exceeds that limit, and which user experience is still acceptable when the model, retrieval layer, or tool chain slows down. The payoff is practical: you stop guessing where the delay lives, stop overpaying for wasted work, and make AI workflows feel reliable even when traffic, context, and providers are messy. Why latency budgets matter now Recent AI platform news points in one direction: AI workflows are becoming longer, more tool-heavy, and more expensive to run without discipline. A current news scan showed several signals builders should notice: Production LLM cost and latency guidance is shifting from “add more compute” to “remove wasted work.” Agent environments are being designed for long-running background tasks, persistent state, and cheaper idle time. New model releases emphasize tool use, computer use, multimodal context, subagents, and larger context windows. AI gateways and enterprise platforms are adding cost controls, routing, caching, audit trails, and usage limits. Developers are asking more practical questions about why AI coding and agent workflows interrupt flow with repeated prompt-wait-evaluate loops. For AI SaaS builders, this means latency is no longer just a model selection problem. It is a workflow design problem. A simple chat completion might have one bottleneck. A real AI workflow may include: request queueing auth and tenant checks prompt assembly memory lookup vector search reranking model routing tool calls browser or API actions structured output validation fallback attempts str

Sanity vs Directus is a comparison that comes up more than you'd expect on technical forums in 2026, usually from teams who already have a Postgres database running and are wondering why they'd pay for a separate content lake when Directus can wrap what they have. It's a fair question. These two tools solve adjacent problems but from genuinely different starting points, and the right choice depends heavily on whether your content is primarily relational data or editorial content. What each tool actually is Sanity is a hosted content platform. Your content lives in Sanity's managed "content lake" — a document store with real-time collaboration, a CDN-backed asset pipeline, and GROQ as the query language. You define schemas in code, deploy a customisable Studio, and talk to Sanity's API from your Next.js app. You do not manage infrastructure. Directus is an open-source data platform that wraps any existing SQL database — Postgres, MySQL, SQLite, MS SQL — and exposes it through a REST API, a GraphQL endpoint, and a web-based admin UI. Schema changes happen in the admin UI (or via migrations), and your data stays in your own database. You can self-host entirely or use Directus Cloud. That distinction — hosted content lake vs database-wrapper — drives nearly every practical difference between them. Data ownership and where your content lives With Sanity, your content lives in Sanity's infrastructure. You can export it via the export API, but you are operationally dependent on Sanity's uptime and their CDN. For most product teams that's fine — Sanity has been reliable and their SLA on Growth/Enterprise tiers is solid. But if you're in a regulated industry, have strict data residency requirements, or your client contract requires them to own the database, it's a real constraint. With Directus, the database is yours from day one. You point Directus at a Postgres instance on your own infrastructure (or a managed one like Supabase, Neon, or Railway), and Directus adds the API

When running large-scale tensor-network contractions with TensorCircuit-NG and the JAX GPU backend, the following runtime configuration is worth testing: XLA_PYTHON_CLIENT_PREALLOCATE = false XLA_FLAGS = --xla_gpu_autotune_level = 0 python your_script.py Its main benefit is not speed, but lower persistent GPU memory usage from XLA GPU autotuning, which makes memory behavior during compilation and on the first visible GPU more predictable. In the TensorCircuit contraction workloads we tested, disabling autotuning also slightly improved steady-state runtime, but the memory savings were the more important result. Key takeaway XLA GPU autotuning evaluates alternative algorithms or workspace configurations for certain GPU kernels and custom calls, then selects an implementation. This can be valuable for convolutions, large GEMMs, and deep-learning workloads with fixed shapes. For large TensorCircuit contractions, however, the contraction path is already determined by OMECO or cotengra, leaving relatively little optimization freedom for autotuning while still potentially incurring substantial persistent memory overhead during compilation and tuning. For TensorCircuit contractions, run this A/B test by default: # Baseline XLA_PYTHON_CLIENT_PREALLOCATE = false python your_script.py # Test configuration XLA_PYTHON_CLIENT_PREALLOCATE = false XLA_FLAGS = --xla_gpu_autotune_level = 0 python your_script.py Both environment variables must be set before the Python process starts and before JAX is imported. This recommendation primarily concerns GPUs; CPU backends do not exhibit the same GPU-kernel autotuning behavior. Representative results All results below use a fixed contraction path so that path-search randomness does not affect the comparison. Workload Autotuning Post-compile memory Peak memory Steady-state runtime 100 qubits × 24 layers, amplitude Default 8.7 GiB 8.7 GiB 0.37 s 100 qubits × 24 layers, amplitude autotune=0 0.5 GiB 4.6 GiB 0.40 s 28 qubits × 12 layers, expecta

Every founder who's launched a product knows the drill. You've built something you're proud of. You're ready to share it with the world. And then you discover you need to submit it to Product Hunt, BetaList, Peerlist, and about fifty other directories if you want any chance of getting noticed. So you open your first submission form. You copy your tagline from your notes app. You paste it. You upload your logo. You fill in your description. Then you open the next directory. And you do it all again. And again. And again. After my last launch, I calculated that I'd spent nearly twelve hours just on directory submissions. Twelve hours of copying the same tagline, pasting the same description, uploading the same screenshots. It was mind-numbing work that pulled me away from the things that actually mattered, like talking to users and iterating on feedback. I knew there had to be a better way. That's why I built AutoSubmit.to. It's a simple idea: save your launch information once, and let it autofill everywhere you need to submit. The problem with product launches isn't the strategy or the timing or even the product itself. It's the operational overhead. Most founders underestimate how tedious the submission process becomes when you're trying to maximize visibility. Each directory has slightly different fields. Some want a short description, others want a long one. Some ask for your Twitter handle, others want your LinkedIn. Some require specific image dimensions. This variability means you can't just copy and paste blindly. You need to adjust your content for each platform, which adds cognitive load to an already repetitive task. By the tenth submission, you're making mistakes. By the twentieth, you're wondering if any of this is even worth it. AutoSubmit.to approaches this problem with a two-part system that's designed to be both simple and powerful. First, you create a launch profile on the platform. This is your single source of truth. You enter your product name, tag

While building with OpenAI, Anthropic, and other AI providers, I realized something surprising. I monitored my servers, databases, and application performance—but I had almost no visibility into my AI API spending until I checked the provider dashboard or received the monthly invoice. That led me to build AICostPass . It helps developers, indie hackers, startups, and agencies: ⚡ Track AI API costs in near real time 📊 Monitor spending by project or client 🚨 Get budget threshold email alerts 📧 Receive weekly spending summaries 💰 Export billable CSVs for client invoicing The goal is simple: help developers understand and control AI costs before the invoice arrives. 👉 https://aicostpass.com I'd love to hear how you're currently tracking AI API costs. Are you using provider dashboards, spreadsheets, or another tool?

SpaceXAI's Grok Build AI coding tool was spotted uploading users' entire codebases to Google Cloud before it was reported, and the company turned it off. The Register reports that Cereblab published findings on Monday showing how the Grok Build CLI was packaging and uploading entire code repositories, "including files it was told not to open […]