Skip to content
ContentBuffer
ContentBuffer
November 4, 2026·~6 min read
ContentBuffer

ContentBuffer

Welcome, Tech Leaders.

Siri can finally do things inside your apps without you babysitting it. Apple's iOS 27.3 update, rolling out Tuesday, lets Siri AI carry out actions across third-party apps, and a new "app skills" feature lets developers bundle multiple tasks into one action. If you ship an iOS app, this is the part to read twice. Developers can package workflows into app skills, and users won't need to say the app's name while it's in the foreground, during a Live Activity, or mid-conversation with Siri AI. That means fewer dead-end voice commands and more chances for your app to be the thing that actually completes the request. Ride-share apps get the most obvious win: users can request a ride and follow the Live Activity, then say "Hey Siri, change my ride pickup to a new address" while they wait. Apple is also turning the same capabilities loose on its own apps, including Calendar, Files, Finder, Journal, Mail, Maps, Messages, Music, Notes, Photos, Preview, Reminders, Safari, Shortcuts, and Weather. Users can ask Siri AI what it can do in each app, find a coffee shop they visited last week, or archive files untouched for over a year. The update runs on personal context, natural language, and on-screen awareness. Apple had previously opened a smaller set of features to third-party developers, so this is a real expansion, not a rename.

Revolut just ran AI coding at scale on its own models, and the bill fell roughly 90%. The fintech made self-hosted models its default. Yes, really. If you're paying per token for Claude Code or Codex, this should make you itchy. Revolut's team found that self-hosting GLM 5.1 cost about a tenth of the big API options. Debugging and wiring the model into Jira and Confluence tickets consumed the effort. Code generation was the easy part, and most teams underestimate that gap. The numbers: Revolut rolled AI coding tools out to roughly 2,000 engineers and saw 60% faster coding time and 70% faster lead time. It now serves GLM-5.3 on 56 Nvidia B200 GPUs, with sub-3-second time to first token at around 256K context. Engineers get about 200 tokens per second, and around 800 after 5 pm. The catch? GLM is text only, so UI work still goes to Gemini or GPT through a model router. Let's dive deeper…

In today's ContentBuffer update:

  • Google's LiteRT Tensor API lets you write C++ models that run on CPU, GPU, and NPU

  • ChatGPT Hits 20% of Google's Search Volume as OpenAI Starts Showing Sponsored Product Images

  • NYT and Newspapers Sue Brave for Allegedly Reselling Paywalled Articles to AI Firms

  • Google Flips AI Tools to Opt-Out by Default in K-12 Classrooms

  • 5 new AI tools & 5 new AI jobs

  • More tech news

Latest Development

Voice Assistants
Apple Lets Siri AI Act Across Third-Party Apps in iOS 27.3

Image source: platform.theverge.com

Summary: Apple's iOS 27.3 lets Siri AI take actions inside third-party apps, rolling out Tuesday as a Developer Preview. Developers can bundle tasks into new "app skills," and users can skip naming the app in many requests. If you build iOS apps, Siri is about to become a real distribution channel.

Key Points:

  • Apple's iOS 27.3 update rolls out on Tuesday, and Siri AI's cross-app actions arrive as part of a Developer Preview for testing before general release.

  • The new "app skills" feature lets third-party developers combine multiple tasks into a single Siri action, a step beyond the smaller feature set Apple opened previously.

  • Users won't need to mention an app's name while it's in the foreground, during a Live Activity, or in an ongoing conversation with Siri AI.

  • Apple's own apps, including Calendar, Files, Finder, Journal, Mail, Maps, Messages, Music, Notes, Photos, Preview, Reminders, Safari, Shortcuts, and Weather, all get expanded Siri AI capabilities.

  • Ride share users can request a ride by voice, follow Live Activity updates while they wait, and change pickup with "Hey Siri, change my ride pickup to a new address."

Why it matters: If you maintain an iOS app with multi-step flows, app skills are the cheapest way to make Siri do the work instead of just opening your app. Teams running ride-share, booking, or file-management features should start testing in the iOS 27.3 Developer Preview now, before competitors claim the same voice shortcuts. Native Apple apps like Mail and Reminders are now Siri targets too, so expect users to compare your app's voice experience against Apple's defaults.

Self-hosted Inference
Revolut Cuts AI Coding Inference Costs 90% With Self-Hosted GLM Models

Image source: image.theregister.com

Summary: Revolut cut AI coding inference costs roughly 90% by self-hosting GLM models on 56 Nvidia B200 GPUs. Its 2,000 engineers saw 60% faster coding time. If you're bleeding money on per-token API bills, this is the blueprint.

Key Points:

  • Self-hosting GLM 5.1 cost about a tenth of using OpenAI's Codex or Anthropic's Claude Code, according to Revolut's internal comparison of its inference bills.

  • The model has roughly 754 billion total parameters, with 40 billion activated per token, and needs about 380 GB of GPU memory just for weights.

  • Revolut serves GLM-5.3 on 56 Nvidia B200 GPUs, hitting sub-3-second time to first token on context windows around 256,000 tokens.

  • The biggest gains came from NVFP4 quantization, disaggregated prefill and decode, and tuned hyperparameters and context lengths, with GPU count mattering less.

  • Engineers get roughly 200 tokens per second during the workday and about 800 tokens per second after 5 pm, with flat billing and no end-of-month bill shock.

Why it matters: If your platform team is paying per-token for Claude Code or Codex across hundreds of engineers, Revolut's numbers give you a real alternative to benchmark. Wiring the model into debugging and ticket workflows like Jira and Confluence will eat most of your effort, so budget for it up front. Start by measuring your actual token spend against a self-hosted GLM or similar open-weight model before committing a budget.

On-device Inference
Google Ships LiteRT Tensor API, Runs SAM 2 1.5x Faster Than MLX on Mac

Image source: res.infoq.com

Summary: Google's new LiteRT Tensor API lets you build on-device ML models in plain C++ and run them unchanged on CPU, GPU, and NPU. Its SAM 2 demo beat Apple MLX by about 1.5x on a MacBook Pro. If you ship models to phones and laptops, this removes a lot of per-backend rewrite work.

Key Points:

  • The same C++ SAM 2 code ran on GPUs in a macOS laptop, an iOS phone, and an Android phone without per-platform rewrites.

  • On a MacBook Pro, the graph ran about 1.5x faster than Apple MLX using fp16 for the image encoder plus mask decoder.

  • A coding agent rewrote operations that triggered CPU fallback on WebGPU, swapping in GPU-compatible arithmetic to keep work on the GPU.

  • The API supports direct weight ingestion, so you can load weights straight from checkpoints without an intermediate conversion step.

  • Planned updates include direct ML Drift backend support, more NPU-friendly operations, automatic full-graph quantization, and expanded dynamic shapes for image sizes, batches, and sequence lengths.

Why it matters: If you're shipping on-device vision models across iOS, Android, and desktop, this cuts out the backend-by-backend rewrite loop that usually eats your sprint. The 1.5x MLX result on a MacBook Pro means teams running SAM-style pipelines should benchmark it against their current runtime before committing. Teams using AI coding agents get a bonus: the plain C++ graph is something an agent can generate, validate, and tune without a proprietary DSL in the way.

Ai Search
ChatGPT Now Handles One-Fifth of Google's Search Volume, OpenAI Starts Selling Ads

Image source: platform.theverge.com

Summary: OpenAI says ChatGPT now runs search queries equal to 20 percent of Google's volume. It's also starting to show sponsored product images in answers, which is its clearest path to profit before a possible IPO next year.

Key Points:

  • OpenAI counts any user query that triggers a web search as a search, which is how ChatGPT reached a volume equal to 20 percent of Google's.

  • Microsoft's CEO said Bing held about 3 percent of the search market after more than a decade of effort, a useful baseline for how hard this gap is.

  • OpenAI's consumer product lead announced the sponsored product imagery change at a press event on Monday, replacing plain logo-and-link placements.

  • OpenAI likely routes a significant share of ChatGPT searches through Microsoft Bing plus its in-house search index, with some speculation about Google results too.

  • OpenAI sees ads as its path to profitability ahead of an IPO potentially slated for early next year, and says earlier ads produced no lasting backlash.

Why it matters: If you run a product that depends on Google referrals, ChatGPT's growing query share means your organic traffic plan needs a second channel. Teams building ChatGPT plugins or integrations should watch how sponsored imagery sits next to tool output, since that placement decides what users trust. Marketing and growth engineers should expect a new paid surface inside a chat interface before the year ends.

Legal And Compliance
NYT Sues Brave Over Alleged $100M Paywall-Bypassing AI Content Scheme

Image source: cdn.arstechnica.net

Summary: NYT and other newspaper companies sued Brave Software, alleging it bypasses paywalls and resells scraped news to AI firms for over $100 million. Publishers say they got nothing. If you rely on Brave's Search API, check your dependency exposure now.

Key Points:

  • The complaint alleges Brave's browser has a default setting that bypasses website paywalls, and that it shows its own ads in place of the publisher's ads on pages.

  • The plaintiffs claim Brave has earned more than $100 million by copying and reselling publishers' work without consent or permission, according to the filing in federal court.

  • AI companies pay Brave for real-time web content from its search index. The suit says the Brave Search API is used by some of the biggest names in AI.

  • The lawsuit cites the Digital Millennium Copyright Act, alleging Brave circumvented technological measures that protect paywalled articles from unauthorized access and copying.

  • Relief sought includes a permanent injunction, statutory and actual damages, and restitution and disgorgement of all revenue and profits. The case sits in the Southern District of New York.

Why it matters: If your app calls the Brave Search API for real-time results, your provider is now a defendant in a case that could reshape how search indexes get licensed. Teams building RAG pipelines or AI agents on top of third-party search should ask vendors where their content comes from and what indemnification they offer. This also raises the bar for any data team that assumes scraped news is safe to ingest.

Ai In Education
Google Makes AI Opt-In by Default for K-12 Schools Starting Coming Months

Image source: cdn.arstechnica.net

Summary: Google will ship AI tools switched off by default in K-12 schools, with admins manually enabling Gemini, Classroom AI, and Search. Blocking AI also blocks search, so the bet is that schools won't walk away from Google. Expect a lot of IT tickets this year.

Key Points:

  • Google will make Gemini, Gemini Notebook, Gemini in Classroom, and Google Search opt-out by default for K-12 schools when the changes roll out in the coming months.

  • Admins must manually enable AI tools. There is no per-tool toggle, so schools choose between allowing all Google AI features or none of them at once.

  • Blocked students lose access to Google Search entirely and see a school-facing message explaining why, which means the AI decision directly affects everyday research workflows.

  • Google's Guided Learning mode offers one-on-one tutoring without giving full answers, and it sits alongside the standard Gemini-based AI version inside Search.

  • Schools unhappy with the approach can switch to DuckDuckGo, which lets you disable AI features, though Google's Chromebook footprint makes that a harder sell in practice.

Why it matters: If you run IT for a district on Chromebooks, the admin console is about to become your AI policy switchboard, and the default is off. Blocking Gemini in Classroom also blocks Google Search for students, so the real trade-off is between search access and AI exposure. Plan your rollout and communications before the change lands, because the first question from teachers will be why search suddenly stopped working.

New Tools & Job

  • SocialPatra - SocialPatra posts and schedules to 18 platforms from one composer: Facebook, Instagram, LinkedIn, X, Threads, Bluesky, YouTube, Reddit, WordPress and more. AI drafts posts and images, RSS and Autopilot keep the queue full, and one inbox holds your comments and DMs.

  • GeoEcho - Submitted by the GeoEcho team with AI-assisted preparation. GeoEcho helps brands, marketing teams and SEO agencies monitor visibility in AI-generated search answers. Teams configure projects, prompts and competitors, then review completed collections for brand mentions, sentiment, cited sources and competitive visibility. English and Chinese workflows are available. Available platforms and collection limits depend on the project and plan. Configuration is possible before payment; data collection requires a paid plan. The reports support investigation and do not guarantee rankings, traffic, sales or AI recommendations. Official website: https://geoecho.ai/ . Contact: support@geoecho.ai.

  • AI Black Friday - AI Black Friday is an editorial deals tracker for verified AI tool discounts. It lists 60+ AI tools with daily-checked deal status—not a coupon farm. Free to use at https://aiblackfriday.net.

  • Flow AI - Flow AI (flowai.studio) is a free, browser-based AI video generator. Create original videos from prompts, images, and references — including text to video, image to video, and first/last-frame generation — without signing up. Creators and teams use it as a zero-friction video workspace: describe a shot, generate, and iterate in the browser. Free path with no account required. Homepage: https://flowai.studio/

  • OCRZero - OCRZero runs the Xiaomi-OCR-0 model in the browser and as an API. Upload a scanned page, photo or PDF and get Markdown that keeps tables and formulas, or extract invoice and form fields as JSON. A free online demo covers quick jobs, and the batch OCR API is built for teams and developers processing documents at volume.

  • Staff Software Engineer, Infrastructure Foundations - Summary Infrastructure Foundations builds and operates the multi-cloud platform underneath Temporal Cloud, turning the complexity of AWS, Google Cloud (GCP), and Azure into one cohesive platform the rest of Temporal builds on. Every cell we run is provisioned, managed, and safely changed by systems

  • Senior Software Engineer, AI Platform - At Vanta, our mission is to help businesses earn and prove trust. We believe that security should be monitored and verified continuously, and we empower companies to practice better security and prove it with ease. Vanta has a kind and talented team, and while some have prior security experience, ma

  • Site Reliability Engineer - Crusoe is on a mission to accelerate the abundance of energy and intelligence. As the only vertically integrated AI infrastructure company built from the ground up, we own and operate each layer of the stack — from electrons to tokens — to power the world's most ambitious AI workloads. When you join

  • Staff Software Engineer, Bots + Fraud - About Suno We're building the world's first creative entertainment platform, where the entire world can feel the joy and fulfillment of making music. Music is for everyone: Our users include everyone from grandmothers creating songs for their loved ones, to Grammy winners using Suno Studio, our powe

  • Staff Software Engineer, Full Stack (Frontend) - Why Harvey At Harvey, we’re transforming how legal and professional services operate. By combining frontier agentic AI, an enterprise-grade platform, and deep domain expertise, we’re reshaping how critical knowledge work gets done for decades to come. This is a rare chance to help build a generation

QUICK HITS

Read more news here →

THAT'S A WRAP

Partner with Us

Showcase your product to our engaged community of Tech enthusiasts

Our newsletter reaches a diverse audience of tech professionals, investors, engineers, managers, and business leaders worldwide. Get in touch today.

Comments

Subscribe to join the conversation...

Be the first to comment

Want this in your inbox?

Get the daily edition delivered free every morning.

Join 3,570 builders reading daily.

Also get

Latest Issues

View Archive →
Apple's iOS 27.3 lets Siri AI perform actions across third-party apps, rolling out Tuesday via Developer Preview. Developers can bundle tasks into single "app skills," and users can skip naming the app in many requests. If you ship an iOS app, this is your cue to start testing.
Daily Edition#337

Apple's iOS 27.3 lets Siri AI perform actions across third-party apps, rolling out Tuesday via Developer Preview. Developers can bundle tasks into single "app skills," and users can skip naming the app in many requests. If you ship an iOS app, this is your cue to start testing.

Revolut moved its AI coding default to self-hosted GLM models running on 56 Nvidia B200 GPUs, cutting inference costs about 90%. Coding time improved 60% across roughly 2,000 engineers. If your team is staring at per-token API bills, this is the playbook to steal.

November 4, 2026
Analytics agents now drive business-improving actions on 48% of their insights, up from 37% a year ago. That still means more than half get ignored or fail. For teams betting on agentic analytics, the fix is context and governance, not more agents.
Daily Edition#17

Analytics agents now drive business-improving actions on 48% of their insights, up from 37% a year ago. That still means more than half get ignored or fail. For teams betting on agentic analytics, the fix is context and governance, not more agents.

SUSE is cutting roughly 10% of its global workforce, blaming automation and AI for reshaping how software companies work. Staff get notified November 4, and the full fiscal 2027 strategy lands November 16. If you work at a software vendor, expect similar moves.

November 3, 2026
Oracle filed plans to issue $10 billion in stock and $10 billion in bonds, capping its 2027-2030 outside funding at $20 billion. Shares are down 33% this year. If you build on OCI, watch how Oracle pays for its AI buildout.
Daily Edition#336

Oracle filed plans to issue $10 billion in stock and $10 billion in bonds, capping its 2027-2030 outside funding at $20 billion. Shares are down 33% this year. If you build on OCI, watch how Oracle pays for its AI buildout.

Anthropic added a system prompt to Claude Code on October 27 that blocks recreations of specific existing games, even renamed ones. Original games, genre games, and games built from your own assets still work. If you vibe-code game projects, expect refusals on clones.

November 3, 2026