Skip to content
AI Newsletter

Siri is about to stop just answering and start acting

ContentBuffer
ContentBuffer
November 20, 2026·~6 min read
ContentBuffer

ContentBuffer

Welcome, Tech Leaders.

Apple is hiring for an AI agent that can browse the web, move across apps, and call services on its own. The job listing says it will plan multi-step tasks and coordinate specialized sub-agents, with the engineer working across the Siri, UI Frameworks, and Apple Foundation Models teams. Siri is the obvious front end. If you ship iOS software, pay attention. Apple's iOS 27.3 update will let developers integrate Siri AI into all apps, so the assistant will be able to take actions inside your product, not just answer questions about it. Users will move fluidly between tools, which means the apps that expose clean actions and APIs will get used and the rest will get routed around. The timing matters. The AI-upgraded Siri lands in September, and the job listing hints at small but telling features like rotating and cropping images in Photos or organizing Safari tabs. Apple is also racing rivals: Meta's Muse, OpenAI's Dots, Anthropic's Duet, and Google's Tasks are all chasing the same agent role. Muse, ChatGPT, and Perplexity already control Mac apps, and Google just let Gemini do the same. Apple is late to this particular race, and it knows it.

Anthropic just pulled off the largest IPO in history, raising about $81 billion at a valuation near $1.5 trillion, and it now has roughly $200 billion in the bank. Revenue is the headline: the latest quarter came in at $17 billion, up more than 1,100 percent year over year, while operating losses are shrinking. Why should you care? Because that money is going somewhere. Anthropic has committed to paying vendors over $900 billion across the next 19 years, which means cloud capacity, chips, and data center buildouts are about to get a lot of demand behind them. If you're building on Claude, expect the platform to keep scaling fast, and expect the vendor ecosystem around it to tighten. The numbers that matter: customers spending six figures a year are up 12 times from a year ago, and the company logged over $104 billion in losses through September. The company is organized as a public benefit corporation, which lets it weigh societal impact alongside shareholder returns. Worth watching: the company has published no environmental impact report and has no commitment to renewable power for its data centers. Watchdogs have also flagged concentrated voting control. Those are the parts to track as the stock starts trading. Let's dive deeper…

In today's ContentBuffer update:

  • Microsoft Releases .NET 11 With C# 15 Union Types and OpenAPI 3.2 Default

  • Google's Gemini 4 Neon Ranks #3 Overall and Undercuts on Price

  • Meta's Standard Model Templates Cut Per-Model Engineering Time by 92%

  • AI Gun Detection Mistakes Doritos for a Weapon, Teens Handcuffed

  • 5 new AI tools & 5 new AI jobs

  • More tech news

Latest Development

Ai Agents

Apple Hires Engineers to Build Its Own AI Agent for Siri Across 3 Divisions

Image source: platform.theverge.com

Summary: Apple is hiring engineers to build an AI agent that browses the web, works across apps, and calls services. It ties into the Siri upgrade shipping in September and iOS 27.3's developer integration, so if you ship on iOS, your app's actions are about to matter a lot more.

Key Points:

  • Apple's AI-upgraded Siri rolls out in September, letting users ask questions about on-screen content and perform actions across apps without leaving the current view.

  • The upcoming iOS 27.3 update gives developers the ability to integrate Siri AI into all apps, widening the range of tasks the assistant can complete on a user's behalf.

  • The job listing describes an agent working across apps, browsers, command lines, code, and APIs, able to coordinate specialized sub-agents while carrying out multi-step tasks end to end.

  • Apple's job listing points to small features like rotating and cropping images in Photos or organizing Safari tabs, suggesting the agent targets everyday actions people already perform.

  • Competitors Meta's Muse, OpenAI's Dots, Anthropic's Duet, and Google's Tasks are racing for the same assistant role, while Google recently let Gemini take control of computers like Muse and ChatGPT already do.

Why it matters: If you ship an iOS app, iOS 27.3's Siri integration means the assistant may start calling your app's actions directly, so your intents and APIs become the new product surface. Teams working on Siri, UI Frameworks, and Apple Foundation Models will set how deep that access goes, which determines how much of your app users reach without opening it. Mac developers should watch the same pattern, since Muse, ChatGPT, and Perplexity already drive desktop apps and Apple is clearly building toward matching that.

Ai Business And Funding

Anthropic Hauls In $81 Billion in Record IPO at $1.5 Trillion Valuation

Image source: media.wired.com

Summary: Anthropic raised about $81 billion in the biggest IPO ever for a public benefit corporation, at a roughly $1.5 trillion valuation. Quarterly revenue hit $17 billion, up over 1,100 percent, so the AI infrastructure spending wave just got a lot bigger.

Key Points:

  • The IPO raised about $81 billion, the largest initial public offering in history, leaving Anthropic with roughly $200 billion in cash after closing.

  • Anthropic has committed to paying vendors more than $900 billion over the next 19 years, a figure that will ripple through cloud, chip, and data center suppliers.

  • Quarterly sales reached $17 billion, up more than 1,100 percent from the prior year, while operating losses are declining as revenue accelerates.

  • Customers spending six figures annually grew 12-fold in a year, and the company posted over $104 billion in losses through September this year.

  • Anthropic has no renewable power commitment for its data centers and has never published an environmental impact report, a gap investors and regulators will press on.

Why it matters: If your team builds on Claude, this IPO signals the platform will keep scaling hard, which means better capacity and more pricing leverage for you. The $900 billion vendor commitment will pull cloud and chip suppliers toward Anthropic's workloads, so expect infrastructure partners to prioritize its needs. Platform and infra teams should watch how the public benefit structure and any sustainability pressure affect where Anthropic builds data centers.

Web-frameworks

Microsoft Ships .NET 11 With C# 15 Union Types and OpenAPI 3.2 by Default

Image source: res.infoq.com

Summary: Microsoft shipped .NET 11 with C# 15 union types for JSON APIs and OpenAPI 3.2 as the default document version. It's a Standard Term Support release with two years of patches. Check your OpenAPI tooling before upgrading, since the new default is a breaking change.

Key Points:

  • Minimal API validation now runs asynchronous validators, so a rule waiting on a database lookup or remote call no longer blocks a thread while it waits.

  • Endpoints can skip the rest of the middleware pipeline through the new [ShortCircuit] attribute, useful for hot paths like health checks and static responses.

  • OpenAPI documents default to version 3.2, backed by the Microsoft.OpenApi 3.3.1 dependency, a breaking change for projects pinned to older generator behavior.

  • SignalR adds authentication refresh, so long-lived connections stay open, and the .NET client refreshes automatically by default while the TypeScript client can do the same.

  • The experimental Microsoft.AspNetCore.Components.AI package stays in prerelease throughout .NET 11, while Zstandard compression is on by default for responses and requests.

Why it matters: If you maintain Minimal APIs or MVC controllers returning JSON, C# 15 unions let you model polymorphic responses without wrapper hierarchies, but only for bodies, so route and query parameters stay as they are. Teams running OpenAPI generators or Swagger UI should test against 3.2 output now, because the default flip can break clients silently. Blazor teams on static SSR forms get client-side validation and CacheView for free, which may let you drop custom caching middleware.

Frontier Models

Google Ships Gemini 4 Neon at $3 Per Million Input Tokens With 1M Context

Image source: storage.googleapis.com

Summary: Google released Gemini 4 Neon, ranking #3 overall on the Artificial Analysis Intelligence Index. It's priced at $3 per million input tokens with a 1M-token context window, aimed at coding agents and enterprise workflows. Rolling out today in the API, AI Studio, and the Gemini app.

Key Points:

  • Neon scores 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level, placing it third overall and ahead of comparable models in its class.

  • Cached input tokens cost $0.15 per million, a 95% discount versus the $3 standard input rate, which matters for agents that resend long system prompts on every call.

  • On SEC-bench Pro, Gemini 4 Neon for cyber hits 90.5% state-of-the-art while costing roughly half as much per challenge as the prior generation.

  • Neon posts 70.5% on Terminal-Bench Science, 84.0% on RiemannBench, and 71.3% on HLE-Diamond, all best-in-class results across hard reasoning evaluations.

  • On SWE-sweep, Neon leads the mini-SWE harness at 12.8% and climbs to a state-of-the-art 39.0% when running inside Google Antigravity with parallel subagents.

Why it matters: If you run coding agents on Antigravity or the Gemini API, Neon's cached input discount and 1M-token window change how you architect long-context workflows. The 32% lower cost per task versus Carbon at default settings makes it a serious candidate for high-volume automation. Teams running cyber defense pipelines should test Neon for cyber against their current scanners, since it outperforms Carbon across every AutoBaxBench difficulty tier.

Ml Infrastructure

Meta Cuts Per-Model ML Engineering Time 92% With Standard Model Templates

Image source: engineering.fb.com

Summary: Meta cut per-model ML engineering time by 92% by standardizing its ads ranking models into reusable templates. The approach shipped 5000+ technique-model combinations and delivered 0.86% online lift. If you maintain many models, templates may matter more than new architectures.

Key Points:

  • Meta's Standard Model Templates reduced per-model engineering time by 92%, shifting effort from repetitive per-model adaptation toward tuning shared template versions across the ecosystem.

  • The approach shipped more than 5000 technique-model combinations, a volume that would have been impractical if every idea had to be evaluated and adapted by hand for each model.

  • Offline model quality improved by 0.63% on average at neutral serving capacity, meaning the gains came without requiring extra compute to serve the models.

  • Cumulative online lift reached 0.86%, covering roughly 70% of models by count, which translates into measurable gains across Meta's billions of daily ads ranking requests.

  • The development loop tunes techniques across representative models, adds them to a new template version, and rolls them out to all dependent models, reducing unique setups and raising reliability.

Why it matters: If you maintain a fleet of recommendation or ranking models, Meta's results show that per-model tuning doesn't scale, since each new idea otherwise grows engineering work linearly with model count. Teams running similar setups should audit how many unique configurations they actually have, because that number drives both cost and reliability risk. Start by identifying the two or three model families that could share a template before investing in any new architecture.

Ai-safety

Omnilert AI Flags Chips as Gun, Police Handcuff Four Black Teens

Image source: cdn.arstechnica.net

Summary: Four Black teens sharing Doritos after football practice were handcuffed by police when a school AI gun detector flagged a bag of chips. The alert was cancelled within two minutes, but officers still arrived in at least six cars. The students are suing, and the case is a warning for anyone trusting AI alerts without a human check.

Key Points:

  • The Omnilert system flagged a student holding a bag of Cool Ranch Doritos as a potential gunman at Kenwood High School in Baltimore County in October 2025.

  • Officials confirmed no threat from security camera footage within two minutes, yet the principal still texted a school resource officer, who called in police.

  • At least six patrol cars arrived with lights flashing, and some officers emerged with guns drawn, forcing the teens to kneel on concrete and be handcuffed.

  • The students were released only after officers searched them and found no gun, leaving them with deep red marks on their wrists from the handcuffs.

  • The Office of Inspector General for Education concluded the Omnilert system showed no bias and recommended bi-annual school trainings and a protocol review by BCPS.

Why it matters: If your district or company runs an AI alerting tool, the Omnilert case shows a cancelled alert can still trigger a live police response when a human forwards it on autopilot. Security teams should audit who receives alerts and require a cancellation check before any escalation. The OIG's recommendation to cut the recipient list is the practical takeaway for anyone wiring vendor AI into safety workflows.

New Tools & Job

  • AnswerDojo - AnswerDojo builds a mock interview around the job you're applying for. Choose one of 2,247 real postings from 28 companies, including Amazon, Google, and Netflix, or paste your own JD to see what the role tests. Then answer out loud as an AI interviewer asks questions, listens, follows up, and scores the session. Two voice mocks are free. No credit card required.

  • FullCastBook - FullCastBook turns a manuscript into an AI audiobook with a full cast. Upload an EPUB, PDF, DOCX, TXT or Markdown file, and it finds the chapters, works out who is speaking in each line and gives every character a distinct voice alongside an AI narrator. Regenerate a single line without redoing the chapter, then export the finished audiobook as one M4B with chapter markers, MP3 files or a ZIP. Freemium, paid plans from $9.90 a month.

  • Roomvana - Roomvana is an AI design tool for your home's exterior. Upload one photo of your house and try new siding or exterior paint colors, a new front door or garage door, new shutters or a new roof — or redesign the front yard, backyard or patio. The result keeps your layout, walls and windows, so what you see is still your house, just with the updates. It works indoors, too. Free to try on the web and as a mobile app.

  • British Accent Generator - British Accent Generator is an AI British/UK accent TTS workspace with 315 regionally labelled voices — RP and standard British, young British, Scottish, Welsh, and Northern Irish. Browse and preview speakers, paste a script, and download natural UK English audio for videos, lessons, podcasts, demos, and training. Also includes Voice Clone, Voice Design, Voice Effects, and AI Lip Sync. Free daily credits; paid credit packs and subscriptions unlock longer scripts and premium features.

  • Nano Banana (nanobananana.com) - Nano Banana at nanobananana.com is an independent AI image generation and editing tool for designers, content creators and marketers. Create images from text prompts, edit reference images with natural-language instructions and blend multiple images. A Chinese-language interface is available. It is separate from nanobanana.im and is not an official Google website.

  • Senior Systems Engineer, Virtualization - CoreWeave is The Essential Cloud for AI™. Built for pioneers by pioneers, CoreWeave delivers a platform of technology, tools, and teams that enables innovators to build and scale AI with confidence. Trusted by leading AI labs, startups, and global enterprises, CoreWeave combines superior infrastruct

  • Software Engineer, Production Engineering, FinOps - About Ramp Ramp is building the smart infrastructure for finance teams, embedded in the transaction flow of every dollar a business spends. We automate how over $275B in annualized spend flows in and out of 85,000+ companies: authorizing payments, flagging risk, categorizing spend, and closing books

  • Software Engineer - Postgres Tooling - About Supabase Supabase is the Postgres development platform, built by developers for developers. We provide a complete backend solution including Database, Auth, Storage, Edge Functions, Realtime, and Vector Search. All services are deeply integrated and designed for growth. About the Role We’re lo

  • Senior AI Development Lead - Airbnb was born in 2007 when two hosts welcomed three guests to their San Francisco home, and has since grown to over 5 million hosts who have welcomed over 2 billion guest arrivals in almost every country across the globe. Every day, hosts offer unique stays and experiences that make it possible fo

  • Principal Software Engineer - Lambda, The Superintelligence Cloud, is a leader in AI cloud infrastructure serving tens of thousands of customers. Our customers range from AI researchers to enterprises and hyperscalers. Lambda's mission is to make compute as ubiquitous as electricity and give everyone the power of superintelligen

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.

More from the AI Newsletter

AI Newsletter

Get the next issue

One email a day. Free. Unsubscribe anytime.

How we handle your email: Privacy policy

Comments

Subscribe to join the conversation...

Be the first to comment