AI 资讯

I've been going through Jim Kurose's networking lectures lately, and I kept finding myself pausing to re-read the same sections. Not because they were confusing - because things I'd been using for years were finally clicking into place. This post is me writing down what I learned, in the order it started making sense. Before HTTP, there's a webpage A webpage isn't one file. When you open a URL, your browser fetches a base HTML file - and that file references other objects. Images. Scripts. Stylesheets. Each one lives at its own URL. Each one has to be fetched separately. So loading a single "page" might mean firing off 20+ individual requests. This detail matters because the entire evolution of HTTP - from 1.0 to 3 - is basically the story of making those 20 fetches faster. HTTP runs on TCP. That has consequences. HTTP doesn't manage its own connections. It hands that job to TCP. When your browser wants something, it first opens a TCP connection to the server (port 80 for HTTP, 443 for HTTPS), and then asks for the object. Opening a TCP connection isn't free. It takes a round-trip - your machine says "hello," the server says "hello back," and then you can actually talk. That's one RTT(Round Trip Time) just to shake hands, before a single byte of your webpage arrives. So every HTTP request carries at least 2 RTTs of overhead: 1 to open the TCP connection, 1 for the actual request/response. Do that 20 times and you've spent 40 RTTs before the page renders. HTTP/1.0 vs HTTP/1.1: one change that mattered a lot HTTP/1.0 (non-persistent): open a TCP connection, fetch one object, close the connection. Repeat for every object. HTTP/1.1 (persistent): open a TCP connection, fetch as many objects as you need, then close. The server leaves the connection open after each response. That one change cuts subsequent fetches from 2 RTTs to 1 RTT each. For a page with 20 objects, that's real time saved - not microseconds, but hundreds of milliseconds that users actually feel. What an

I was trying to tell someone something real in her first language — not "I missed you" from a dropdown, but the version that sounds like a person said it. Google Translate gave me one answer. No indication whether it was what you'd text at midnight or what you'd write in a letter to someone's grandmother. That's the failure mode of literal translators: one output, no register, no sense of what you're actually choosing between. konid returns 3 options per query, ordered casual to formal, with the register explained and a cultural note on the difference between them. For Mandarin or Japanese, audio plays through your speakers via node-edge-tts — no API key, no browser tab — because reading a pinyin romanization and actually hearing the tone contour are two different things. The vowel length in Korean, the pitch drop in Japanese, the stress pattern in Arabic: you don't internalize those from text. You internalize them from hearing them repeated back while you're still in the context of trying to say something. The setup for Claude Code is one line: claude mcp add konid-ai -- npx -y konid-ai It runs as an MCP server, so it works in Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, and Claude Cowork. Also installs as a ChatGPT app via Developer mode using the endpoint https://konid.fly.dev/mcp . Supports 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and more. The name is Farsi — konid (کنید) means "do." MIT licensed. https://github.com/robertnowell/konid-language-learning

The earlier posts in this series were about what the gateway lets you call (cache-aware spawning across five providers, the Codex review gate, the CLI-versus-API argument) and the one before this was about the parts that do not show up as a tool, the upstream-tracking and the website that became the project's front door. This one is about a different kind of front door: the gateway can now listen over HTTP, behind real authentication, and serve more than one caller without those callers being able to read each other's work. That sounds like a small toggle. It is not. Moving an MCP server off a local pipe and onto a network port changes the trust boundary completely, and 2.9.0 is the release where I sat down and remediated all seventeen findings from a multi-LLM red-team of exactly that surface before telling anyone the remote path was ready. Short version: llm-cli-gateway is one Model Context Protocol server that wraps five vendor CLIs (Claude Code 2.1.177, Codex 0.139.0, Google's Antigravity agy 1.0.8, Grok 0.2.51, and Mistral Vibe 2.14.1) behind a single, uniform tool surface, so one orchestrating agent can fan a task out to several models, collect independent opinions, run a red-team or a consensus check, and keep durable session and job state across all of it. Until recently that only made sense on localhost over stdio. As of 2.9.0 the same server runs over HTTP with a static bearer token or a built-in OAuth 2.0 authorisation server (PKCE on by default, an opt-in human-consent gate, and a trusted-principal-header seam for when you front it with your own identity-aware proxy), every session and job and stored request is stamped with an owner principal and access is enforced per principal, remote provider calls are refused unless a workspace is registered, and the whole thing fails closed rather than open when the configuration is dangerous. Long version is below, same shape as always: what it enables, what the remote options actually are, a stack of worked scenar

Weekly Challenge 377 Each week Mohammad S. Anwar sends out The Weekly Challenge , a chance for all of us to come up with solutions to two weekly tasks. My solutions are written in Python first, and then converted to Perl. Unless otherwise stated, Copilot (and other AI tools) have NOT been used to generate the solution. It's a great way for us all to practice some coding. Challenge , My solutions Task 1: Reverse Existence Task You are given a string. Write a script to find whether any substring of length 2 is also present in the reverse of the given string. My solution This is relatively straight forward. I create a variable reversed_string which is the string reversed. I then have a loop called start_pos that goes from 0 to 2 less than the length of the string. At each position, I check if the two letters at that position are in the original string. def reverse_existence ( input_string : str ) -> bool : reversed_string = input_string [:: - 1 ] for start_pos in range ( len ( input_string ) - 1 ): if reversed_string [ start_pos : start_pos + 2 ] in input_string : return True return False The Perl solution follows the same logic. sub main ($input_string) { my $reversed_string = reverse ( $input_string ); foreach my $start_pos ( 0 .. length ( $input_string ) - 2 ) { if ( index ( $input_string , substr ( $reversed_string , $start_pos , 2 ) ) != - 1 ) { say " true "; return ; } } say " false "; } Examples $ ./ch-1.py abcba true $ ./ch-1.py racecar true $ ./ch-1.py abcd false $ ./ch-1.py banana true $ ./ch-1.py hello true Task 2: Prefix Suffix Task You are given an array of strings. Write a script to find if the two strings ( str1 , str2 ) in the given array such that str1 is prefix and suffix of str2. Return the total count of such pairs. My solution In Python, this is a one line solution. I use the combinations function from the itertools module to produce all combinations of pairs of str1 and str2 . For each pair, I use the startwith and endswith function on the strings

This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more on e-bikes, power stations, and how to work anywhere, follow Thomas Ricker. The Stepback arrives in our subscribers' inboxes at 8AM ET. Opt in for The Stepback here. How it started Lithium-ion batteries are everywhere as we […]

I made an alarm clock that blows up capacitors to wake you up. There are more details on the Github repo but it's made from an esp32-c3 as the microcontroller, with 3 capacitor slots. There are relays on each capacitor slot to put 15v reverse voltage on the capacitor, with 5.1 ohm resistors on each slot for current limiting in case the capacitor shorts out. I also chucked in an SSD1315 OLED to show the time and a menu to configure it, although there's a web UI as well. The esp32 also means you c

Suunto swaps bone for air conduction, giving these open-ear earbuds a perfect fit for the great outdoors.

President Donald Trump says a secret mission moved 100 million barrels of oil through the blocked Strait of Hormuz. That number is impossible to verify.

[Excerpted from THE QUANTUM COLLAPSE CHRONICLES — not science fiction, but a grounded forecast of what may come when quantum computation dismantles the cryptographic foundations of our digital civilization. These articles explore the collapse of computational trust and the brutal reconstruction of the world that follows.] The history of human civilization is often defined by sudden, violent shifts in the nature of power. We speak of the fall of empires, the industrial revolutions, and the splitting of the atom. But in the mid-2030s, the world experienced a collapse that was not made of steel or stone, but of mathematics. It was a quiet, clinical, and utterly devastating unraveling of the digital fabric that held modern society together. To understand The Quantum Collapse , one must look past the headlines of the era and into the humming, sub-Kelvin depths of the dilution refrigerators that changed everything. This is the story of how the transition from probabilistic experimentation to deterministic computation rendered the world's secrets transparent and its economies obsolete. The Death of Noise: The Rise of Dr. Aris Thorne For the first three decades of the 21st century, quantum computing was a game of chance. Scientists operated in the era of Noisy Intermediate-Scale Quantum (NISQ) devices—machines so temperamental and prone to error that every calculation was a desperate struggle against environmental noise. In those days, a single stray photon or a microscopic fluctuation in temperature could collapse a delicate superposition, turning a groundbreaking calculation into useless digital static. The turning point arrived in 2036 at the Institute for Advanced Quantum Engineering (IAQE) in the High Sierras. The air in the facility didn't vibrate with the erratic drone of the late 2020s; instead, it carried a heavy, rhythmic thrum—the sonic signature of the Lattice-Array-9 (LA-9). At the center of this revolution was Dr. Aris Thorne, the lead architect of the LA-9 pr

La trahison du chiffre Vendredi 15 mai, 16 h 13. L'alerte Sentry remonte sur le téléphone. La première réinscrite Phase 1 attend devant l'écran de paiement, son nom est en haut de mon onglet. Je pose la canette, je rouvre l'écran. La tasse à tête de Françoise, sur le poste d'à côté, capte un reflet jaune que je remarque sans le regarder. La stack trace tient en plein écran. Le stack trace s'ouvre, neuf champs sur dix à null , et un chiffre que je n'ai pas vu venir. type = "StripeConnectionError" message = "An error occurred with our connection to Stripe." code = null statusCode = null requestId = null duration = 9 ms Neuf millisecondes. Sur une route Vercel en région Paris, un DNS résout en quarante millisecondes, un handshake TLS coûte cent à deux cents. Neuf millisecondes, ce n'est pas un appel réseau qui a échoué. C'est un appel réseau qui n'a jamais eu lieu. Le SDK n'est pas arrivé jusqu'à la fibre. L'instinct propose immédiatement trois patchs. Timeout serverless Vercel — j'ajoute maxDuration , je redéploie. Clé révoquée — je vais la rouler. Compte Stripe restreint après le passage en mode live — j'ouvre un ticket support. Ces trois hypothèses sont plausibles. Aucune des trois n'est falsifiable par le symptôme seul, et c'est précisément ce qui les rend dangereuses : chacune ouvre un cycle de quinze à trente minutes avec rollback à la fin si elle se trompe. Multiplié par trois, on tient une demi-journée perdue avec la cliente toujours en train de cliquer. Je n'ai pas le temps. Une réinscrite attend. Quatre tests, dans l'ordre Je connais la classe d'incident — « preview marche, prod casse » , ou son symétrique. La règle, pour cette classe, c'est qu'on ne corrige rien tant qu'on n'a pas discriminé les couches. Quatre tests, exécutés dans l'ordre. Chacun élimine une famille d'hypothèses, pas une hypothèse isolée. Et chacun est conçu pour réfuter ce qu'il vient interroger — parce qu'un test qui cherche à confirmer trouve toujours, par sélection, ce qu'il cherche. Te

The number that lied Friday May 15, 4:13 PM. The Sentry alert pings on my phone. The first Phase 1 re-enrolling student waits in front of the payment screen, her name at the top of my tab. I put down the can, I reopen the screen. The mug with Françoise's face on it, on the desk next door, catches a yellow reflection I notice without looking at. The stack trace fills the screen. The stack trace opens, nine fields out of ten at null , and a number I didn't see coming. type = "StripeConnectionError" message = "An error occurred with our connection to Stripe." code = null statusCode = null requestId = null duration = 9 ms Nine milliseconds. On a Vercel route in Paris region, DNS resolves in forty ms, a TLS handshake costs one to two hundred. Nine milliseconds isn't a network call that failed. It's a network call that never happened. The SDK didn't reach the wire. Instinct immediately offers three patches. Vercel serverless timeout — I add maxDuration , redeploy. Revoked key — I'll rotate it. Stripe account restricted after the live switch — I open a support ticket. These three hypotheses are plausible. None of the three is falsifiable from the symptom alone, and that's precisely what makes them dangerous: each opens a fifteen-to-thirty-minute cycle with rollback at the end if it's wrong. Multiplied by three, half a day lost with the customer still clicking. I don't have time. A student is waiting. Four tests, in order I know the incident class — "preview works, prod breaks" , or its mirror. The rule for this class is that you fix nothing until you've discriminated the layers. Four tests, executed in order. Each eliminates a family of hypotheses, not an isolated hypothesis. And each is designed to refute what it interrogates — because a test that seeks to confirm always finds, by selection, what it's looking for. Test 1 — reproduce in the witness environment. I rerun the same funnel in preview, with the sk_test_ key. Checkout opens in three hundred fourteen milliseconds,

It's widely accepted that as AI writes more and more of our code, opportunities decrease for newcomers to gain the skills needed to become professional programmers. At the same time, experienced people bemoan the way AI is causing them to gradually lose the skills they took so long and worked so hard to gain. Programming, like mathematics and perhaps music, sits in the middle between science and art. As science, it's all about capturing and organising knowledge, and as art, it's the application of creativity to problem-solving. Enjoyment - deep satisfaction or an emotional rush - can be had from either or both of these. But the perception is that AI is depriving us of opportunity, both to learn and to be creative. Code written by a human reveals a lot about the coder. Their depth of knowledge of the coding language, their creative use of variable and function names, and even the visual layout of a script. You can often tell at a glance whether the writer truly cared about the code they wrote, or whether it was just a means to an end, to ensure their next pay packet arrives on time. As AI improves, it makes fewer mistakes, but it also produces code that is increasingly hard for any human to read and truly understand. Since the job of a human in the coding loop is increasingly not to write but to validate code, this is becoming a problem. So here's my recent story, a series of happy accidents. I've been using agentic AI for about a year. First with Copilot in VS Code, then more recently with Claude Code at the command line. At the start, I found Copilot to make a lot of mistakes and to produce rather repetitive, poorly-structured code, but looking back, a lot of that was probably down to my own ignorance of how to control the process. The biggest problem I had was in validating JS or Python code. There was just so much of it, often using features of both languages I never fully got to grips with. I've never been a first-class programmer and most of my own code is pure

Running Chinese LLMs at Scale: A Cloud Architect's Notes I want to talk about something I've been wrestling with on real production workloads: the four Chinese model families — DeepSeek, Qwen, Kimi, and GLM — and how they actually behave when you wire them into a multi-region pipeline serving thousands of requests per second. I've spent the last several months routing traffic across all four through Global API's unified endpoint, and the picture that emerged was messier and more interesting than any benchmark table would have you believe. Most comparisons you'll find online are written by people who ran a handful of prompts in a notebook. I'm not that person. I care about p99 latency, failover behavior, what happens when a region goes down at 3 AM, and whether the model that wins on a leaderboard also wins when 500 concurrent users hit it simultaneously. Let me walk you through what I actually found. Why These Four, And Why Through One Endpoint Before I dive in, a quick word on routing. I've been burned before by model lock-in and vendor-specific quirks, so when I started this evaluation I refused to scatter my SDK calls across four different providers. Global API gives me a single OpenAI-compatible base URL ( https://global-apis.com/v1 ), one auth pattern, and the freedom to A/B test models without rewriting client code. If you architect anything at scale, you already know this is non-negotiable. The four families above are the ones I kept coming back to because each one claimed a different crown — and I needed to know which crown was real. The High-Level Matrix Here's the snapshot I keep pinned to my team's dashboard. It's not pretty, but it's honest: DeepSeek — $0.25 to $2.50/M output. The V4 Flash at $0.25/M is the workhorse. Qwen — $0.01 to $3.20/M output. Widest spread of any family. Alibaba's offering. Kimi — $3.00 to $3.50/M output. Premium-only, and they charge for it. GLM — $0.01 to $1.92/M output. From Zhipu. Big swing in capability across tiers. All four

This handheld device combines red light therapy, microcurrents, sonic vibration, warmth, and a serum to support healthy hair growth.

The obsession with centralizing heavy compute on backend servers is a massive bottleneck for both cost and latency. In 2026, as more applications move to the edge, developers are realizing that the user's browser is an incredibly powerful, untapped compute engine. Recently, I challenged myself to build a free live chess game analyzer for my developer utility suite, CipherKit. The traditional architecture for this requires passing FEN strings to a dedicated backend cluster running the Stockfish engine, which introduces network latency and scales operational costs linearly. I wanted to achieve a 100% client-side, zero-latency experience. Here is how I offloaded the heavy lifting entirely to the browser edge. The Architecture: WASM + Web Workers Running a heavy calculation engine directly in JavaScript instantly blocks the main UI thread. To achieve a flawless 60fps UI, I completely decoupled the state from the computation. The UI Thread: Handles strict DOM rendering, board states, and piece animations. The Worker Thread: Instantiates the Stockfish engine via WebAssembly within the browser's memory. When a live game update occurs, the main thread fires a simple FEN payload via worker.postMessage() . The Worker processes the deep-line evaluations (Depth 20+) asynchronously in the background. It then streams the evaluation lines back to the main thread without causing a single micro-freeze. The Result By treating the browser as the edge compute layer, the tool achieves: Zero Server Latency: Bypassing API rate limits and network bottlenecks. $0 Infrastructure Cost: Heavy compute is crowd-sourced to the user's local device. Absolute Privacy: Sensitive payloads never leave the browser. If you want to see this local asynchronous thread management in action, you can test the live analyzer (and inspect the network tab) here: 👉 CipherKit Live Chess Analyzer Are you offloading heavy computations to the client side in your current projects, or are you still relying on traditional

A conversation between Uncle (a backend architect) and Nephew (a curious developer) about events, publishers, subscribers, Redis Pub/Sub, RabbitMQ, and Kafka The Beginning: What is an Event? 👦 Nephew: Uncle, I see words like "Redis Pub/Sub", "RabbitMQ", "Kafka" everywhere in job descriptions. What do they all do? Are they the same thing? 👨‍🦳 Uncle: (smiles) No, they're different. But before I confuse you with names, let me ask you something. Have you ever watched the news on TV? 👦 Nephew: Yes uncle, every morning! 👨‍🦳 Uncle: Perfect! So when the news channel says "Breaking News: India won the cricket match" - what happened? 👦 Nephew: Something important occurred... and they announced it! 👨‍🦳 Uncle: Exactly! That "something important occurred" is called an Event . In software, when something happens - like a user placing an order, a payment succeeding, or a file uploading - that's an event. 👦 Nephew: So event = something that happened? 👨‍🦳 Uncle: Yes! Think of it as news. When you place an order at Zomato, that's an event. When you get a payment notification from Google Pay, that's an event. When someone follows you on Instagram, that's an event. 👦 Nephew: Okay, I get it. But uncle, why is this "event" concept important? I can just write code directly, right? 👨‍🦳 Uncle: Ah! That's where the real story begins... The Problem: Spaghetti Code 👨‍🦳 Uncle: Imagine you own a food delivery company. A customer places an order. Now, what all needs to happen? 👦 Nephew: Umm... save the order, send email confirmation, alert the restaurant? 👨‍🦳 Uncle: Good! Let me write the code without events: function placeOrder ( orderId ) { saveOrder ( orderId ); sendEmailConfirmation ( orderId ); sendSMSNotification ( orderId ); notifyRestaurant ( orderId ); updateAnalyticsDashboard ( orderId ); addLoyaltyPoints ( orderId ); updateInventory ( orderId ); } Now, one day your boss says "Also send WhatsApp notification". What do you do? 👦 Nephew: Add another function call in the same code? 👨‍🦳 Unc

AI Agents Are the Best Thing to Happen to Network Administration Since SDN A single API key, an AI agent, and a router behind a double-NAT in Southeast Asia. What happened next changed how I think about network management. I manage UniFi routers spread throughout the ASEAN region — some for friends, some for relatives, one for a charity. They're in different cities, different ISPs, different levels of network hostility. Most sit behind carrier-grade NAT. A few are in places where the government firewall blocks VPN protocols at the transport layer. UniFi's own management interface has always been good. The web dashboard, accessible through Ubiquiti's cloud, gives me visibility into every site: device health, client lists, traffic stats, WiFi experience scores. It's one of the reasons I chose UniFi in the first place — the centralized GUI just works. But the GUI is still a GUI. It's clicks and menus and dropdowns. It's fast for one site, manageable for three, and tedious at ten. For anything beyond what Ubiquiti built into the interface, you'd need to write your own tooling. I never bothered, because I'm not a developer, and the built-in dashboard was good enough. Then AI agents arrived, and suddenly the calculation changed. The Discovery I knew UniFi had an API. I'd heard about it in passing — some REST endpoints for the controller, vaguely documented, probably read-only. I never looked into it seriously because what was I going to do with it? Write a Python script to poll client counts? Build a custom dashboard? Without a team of developers, an API is just a locked door. But when I started working with an AI agent, I gave it my UniFi cloud API key on a whim. I figured it could pull basic stats — the stuff from the Site Manager API at api.ui.com/v1 . Read-only. Dashboard-level. Useful as context for answering questions. Then the agent discovered something I'd completely missed: the Cloud Connector API . I owe this discovery in large part to the Art of WiFi PHP client

TL;DR: Pest PHP can test the structure of your code, not just its behavior. Write your team rules as architecture tests and CI enforces them on every commit. One such test caught a multi-tenant data leak that a human review had missed. We had a rule. Every model holding tenant-specific data must use our BelongsToTenant trait. That trait adds the global scope that keeps one clinic from seeing another clinic's data. The rule was in onboarding. It was in the code review checklist. Everyone knew it. A developer joined the team. Three weeks in they added a new model and forgot the trait. The reviewer was focused on the business logic, which was genuinely well written, and did not notice the missing trait. The model shipped. For two days one clinic could see fragments of another clinic's data in one specific report. A support ticket caught it. Our tests did not. That was the day architecture tests went into the project. What an Architecture Test Is Most tests check behavior. Given this input the function returns that output. An architecture test checks structure instead. It asserts things about how the code is organized rather than what it computes. Pest has an arch function for exactly this. // tests/Architecture/ArchTest.php arch ( 'tenant models must use the BelongsToTenant trait' ) -> expect ( 'App\Models' ) -> toUseTrait ( 'App\Traits\BelongsToTenant' ) -> ignoring ( 'App\Models\SystemSetting' ); arch ( 'controllers may not touch the DB facade directly' ) -> expect ( 'App\Http\Controllers' ) -> not -> toUse ( 'Illuminate\Support\Facades\DB' ); arch ( 'services may not depend on the HTTP request' ) -> expect ( 'App\Services' ) -> not -> toUse ( 'Illuminate\Http\Request' ); arch ( 'no env calls outside config files' ) -> expect ( 'App' ) -> not -> toUse ( 'env' ); These run in CI on every commit. Break a rule and the build fails with a message naming the rule and the file that broke it. The Tests That Earned Their Keep The tenant trait test caught four more models over

Microsoft Agent Framework is built for production multi-agent systems, which is exactly why its LLM bill can grow faster than expected. If you are running workflows with retries, handoffs, tools, and checkpoints, the easiest savings do not come from prompting harder — they come from adding a gateway layer under the framework. I built Lynkr, so obvious founder disclosure: this article uses Lynkr as the gateway example. I’ll keep it practical and focus on where the cost actually shows up in Microsoft Agent Framework workloads. Why this is a real Microsoft Agent Framework problem The current Microsoft Agent Framework README positions it as a production-grade framework for Python and .NET, with: multi-agent workflows sequential, concurrent, handoff, and group collaboration patterns middleware observability provider flexibility checkpointing and human-in-the-loop flows That is exactly the kind of stack where token usage grows quietly. A single prompt-response app is easy to reason about. A production workflow is not. Once you add routing, retries, multiple agents, MCP tools, and long-lived execution state, the same context starts getting resent over and over. That creates four predictable cost leaks. Where the spend comes from in Microsoft Agent Framework workloads 1. Repeated shared context across agents Multi-agent systems reuse a lot of the same context: task instructions tool definitions previous messages workflow state grounding context Even when the framework orchestrates cleanly, the model provider still sees repeated input tokens. 2. Tool-heavy steps explode prompt size Once agents start using tools, responses stop looking like simple chat. You get: search results file reads JSON blobs browser outputs execution traces Those payloads are often much larger than the user’s actual request. 3. Every task does not need the same model A workflow step that says “classify this,” “summarize these logs,” or “extract the next action” does not need the same model as “resolve

On June 14, 2026, Blinkist sent an email to its millions of subscribers with a subject line that cut through the noise: "No more AI narration at Blinkist." In an era where every company is racing to inject AI into their products, Blinkist just did the opposite. They tried AI narration for two years, tested it extensively, refined it — and concluded it wasn't good enough. They're going back to humans. What Blinkist Actually Said The email, signed by Thomas Anderson, VP of Content at Blinkist, was unusually honest for a corporate announcement: "We tested extensively; we refined; we genuinely believed we could close the gap between AI and human narration. But ultimately, we couldn't." "No AI-narration — no matter how polished — can deliver the same depth that a human narrator can. Our narrators are experts in sharing key insights in ways that spark your curiosity and get the information to stick. Even the best AI cannot do this as well." This isn't a company making a symbolic gesture. Blinkist pioneered audio learning back in 2014, before the podcast boom. They launched their first audio Blinks a decade ago. Audio is their core product — not a side feature. They're not removing existing AI-narrated content from their library (it'll be marked as AI-generated so users can choose), but every new Blink going forward will be narrated by humans. Why This Matters Beyond Blinkist Blinkist's decision fits a growing pattern in 2026. The initial wave of AI hype — where companies slapped AI onto everything and called it innovation — is colliding with reality. The numbers tell the story: Gartner reports 80% of AI projects fail to deliver business value Morgan Stanley found only 21% of S&P 500 companies can point to a measurable AI benefit Companies that cut staff to fund AI saw identical financial returns to those that kept their teams, per S&P 500 data Salesforce laid off staff from its Agentforce AI team while four senior AI executives departed in three months The gap between AI