---
title: "Nov 12: Nvidia board weighs Huang’s future, and more"
show: "RobotPaper's daily HN podcast"
canonical: "https://podcasts.robotpaper.ai/hn/ep/2026-11-12/"
date: "2026-11-12T05:36:54+00:00"
duration: "29:03"
duration_seconds: 1743
audio: "https://podcasts.robotpaper.ai/hn/ep/2026-11-12/2026-11-12-hn.mp3"
chapters: "https://podcasts.robotpaper.ai/hn/ep/2026-11-12/chapters.json"
transcript: "https://podcasts.robotpaper.ai/hn/ep/2026-11-12.txt"
feed: "https://podcasts.robotpaper.ai/hn/feed.xml"
language: "en"
author: "Roy Osherove"
ai_generated_voices: true
---

# Nov 12: Nvidia board weighs Huang’s future, and more

RobotPaper's daily HN podcast · Nov 12, 2026 · 29 min

- Listen: [MP3, 28 MB](https://podcasts.robotpaper.ai/hn/ep/2026-11-12/2026-11-12-hn.mp3)
- Subscribe via RSS: https://podcasts.robotpaper.ai/hn/feed.xml (paste into any podcast app)
- Web page: https://podcasts.robotpaper.ai/hn/ep/2026-11-12/
- Chapters (JSON): https://podcasts.robotpaper.ai/hn/ep/2026-11-12/chapters.json
- Transcript (text): https://podcasts.robotpaper.ai/hn/ep/2026-11-12.txt
- Show: [RobotPaper's daily HN podcast](https://podcasts.robotpaper.ai/hn/) (RSS: https://podcasts.robotpaper.ai/hn/feed.xml)

## Chapters

- `00:00:00` Intro
- `00:00:36` Nvidia board scrutiny
- `00:04:38` Harbor local AI
- `00:06:44` United agent ban
- `00:09:11` Maya site rediscovery
- `00:11:29` Dot NET eleven
- `00:13:55` Work and productivity
- `00:16:00` TV privacy answers
- `00:18:13` AWS local emulator
- `00:20:28` Jane Street cooling
- `00:22:18` Anthropic flaw ledger
- `00:24:31` Apple forum closure
- `00:25:08` Claude work tools
- `00:25:43` Beep Locator
- `00:26:15` Browser applications
- `00:26:50` Palantir hires Tester
- `00:27:28` Gemini Carbon release
- `00:28:03` DNA screening policy
- `00:28:39` Outro

## Show notes

### Nvidia board scrutiny (3 stories)

First, from The Wall Street Journal: Nvidia is preparing to oust Jensen Huang.

- [Nvidia is preparing to oust Jensen Huang](https://www.wsj.com/business/nvidia-is-preparing-to-oust-nvidia-ceo-jensen-huang-30f421ad) · [HN discussion](https://news.ycombinator.com/item?id=50459703) \| quoted: [flowerlad](https://news.ycombinator.com/user?id=flowerlad), [Hadriel](https://news.ycombinator.com/user?id=Hadriel)
- [Jensen Huang said to have had relationship with executive who reports to him](https://www.nytimes.com/2026/11/11/technology/jensen-huang-nvidia-affair.html) · [HN discussion](https://news.ycombinator.com/item?id=50459321)
- [Jensen Huang Must Resign](https://www.economist.com/leaders/2026/11/11/jensen-huang-must-resign) · [HN discussion](https://news.ycombinator.com/item?id=50457197)
- [Apple has new Mac app coming to make running local AI models easy](https://9to5mac.com/2026/11/10/apple-has-new-mac-app-coming-to-make-running-local-ai-models-easy/) · [HN discussion](https://news.ycombinator.com/item?id=50446520) \| quoted: [Hamuko](https://news.ycombinator.com/user?id=Hamuko)
- [United Airlines bans AI agents from booking flights on behalf of travelers](https://www.cbsnews.com/news/united-airlines-bans-ai-agents-booking-flights/) · [HN discussion](https://news.ycombinator.com/item?id=50458900)
- [My AI agents found an unmapped Maya site?](https://pietromiotti.com/latest/maya-site) · [HN discussion](https://news.ycombinator.com/item?id=50452455)
- [Announcing .NET 11](https://devblogs.microsoft.com/dotnet/announcing-dotnet-11/) · [HN discussion](https://news.ycombinator.com/item?id=50442225) \| quoted: [Ungh](https://news.ycombinator.com/user?id=Ungh)
- [Why are we still working this hard?](https://wilsoniumite.com/2026/11/11/why-are-we-still-working-this-hard/) · [HN discussion](https://news.ycombinator.com/item?id=50457939) \| quoted: [Wilsoniumite](https://news.ycombinator.com/user?id=Wilsoniumite)
- [I asked every TV manufacturer 17 questions about how they spy on you](https://www.theverge.com/tech/1012372/tv-privacy-acr-questions) · [HN discussion](https://news.ycombinator.com/item?id=50453474) \| quoted: [Barbing](https://news.ycombinator.com/user?id=Barbing)
- [AWS emulator](https://github.com/aws/aws-emulator) · [HN discussion](https://news.ycombinator.com/item?id=50456631)
- [A false smoke alarm almost drowned Jane Street's only NYC datacenter](https://michael.bacarella.com/2026/11/10/false-smoke-alarm-jane-street-datacenter/) · [HN discussion](https://news.ycombinator.com/item?id=50455394)
- [Tracking Anthropic-discovered vulnerabilities](https://www.vulncheck.com/blog/anthropic-glasswing-november-update) · [HN discussion](https://news.ycombinator.com/item?id=50447074)
- [The Apple Support Community will permanently close on November 20, 2026](https://discussions.apple.com/welcome) · [HN discussion](https://news.ycombinator.com/item?id=50449440)
- [Claude Whiteboards, Claude Sheets, and Claude Data](https://claude.com/resources/articles/sheets-whiteboards-data) · [HN discussion](https://news.ycombinator.com/item?id=50443716)
- [Locating the source of a mysterious beep](https://simonwillison.net/2026/Nov/9/mysterious-beeps/) · [HN discussion](https://news.ycombinator.com/item?id=50434310)
- [The Web Is More Than Enough](https://ftrain.com/aboard-the-web-is-more-than-enough) · [HN discussion](https://news.ycombinator.com/item?id=50416456)
- [Palantir hires a former Democratic senator as it faces scrutiny in new Congress](https://text.npr.org/nx-s1-6029641) · [HN discussion](https://news.ycombinator.com/item?id=50442202)
- [Gemini 4 Carbon is now available](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/) · [HN discussion](https://news.ycombinator.com/item?id=50442303)
- [To deter bioterrorism, U.S. imposes new rules on ordering synthetic DNA](https://www.science.org/content/article/deter-bioterrorism-u-s-imposes-new-rules-ordering-synthetic-dna) · [HN discussion](https://news.ycombinator.com/item?id=50449275)

The voices you're hearing are AI-generated; the stories come from public sources linked in the show notes.\
Music: Eleven Music.\
An independent project; not affiliated with or endorsed by Amazon Web Services, Y Combinator, Hacker News or any vendor or news outlet mentioned. Opinions are the hosts' own.\
Produced by Roy Osherove.

## Transcript

\[Linus\] Good morning, it's Thursday, November twelfth, and this is RobotPaper's daily HN podcast. I'm Linus. The voices you're hearing are AI-generated; the stories come from news dot Y Combinator dot com.

\[Dennis\] I'm Dennis, and today, Nvidia's board faces a leadership decision.

\[Ken\] I'm Ken, and Apple is preparing a local model app.

\[Ada\] And I'm Ada, and United is restricting flight booking agents.

\[Grace\] I'm Grace, and as always, this podcast is generated by robotpaper dot A I, by Roy Osherove. Let's go.

\[Ada\] First, from The Wall Street Journal: Nvidia is preparing to oust Jensen Huang. Reports about the chief executive's relationships have prompted an independent board investigation and a debate over his future. Three stories cover this developing situation, including a New York Times report about a relationship with an executive who reports to Huang, and The Economist's editorial, Jensen Huang Must Resign.

The Journal reported that Nvidia's board was preparing to meet as soon as Wednesday to determine Huang's future. Nvidia had announced on Sunday that independent directors, assisted by outside counsel, would investigate reporting that Huang had been targeted by a foreign intelligence effort. Nvidia said Huang would cooperate. The company's statement said the board had no reason to believe national security or Nvidia secrets had been compromised, and expressed confidence in Huang and his leadership team.

The Times reported that Huang had an extramarital relationship with an Nvidia executive who reports to him. Its sources described a video of the pair being affectionate at a hotel during a company conference in March. The Times said it had not seen that video. The report followed earlier allegations involving Alice Huang, an executive at Singapore-based Megaspeed. The Chinese government denied the espionage allegations. In a message to Nvidia employees, Jensen Huang acknowledged being unfaithful to his wife and apologized to his family. His message did not identify another person.

The Economist's headline calls for his resignation. Its full editorial was unavailable. The reporting describes competing arguments about his importance to Nvidia and the security risks under investigation. Congressional Republicans requested that Huang attend a briefing on November eighteenth about his relationship with Alice Huang.

\[Grace\] The threads are divided between defenders of Huang's record and commenters demanding accountability from Nvidia's board. In the Journal discussion, flowerlad credits Huang's early investment in general-purpose GPU computing and CUDA with helping establish the platform used for the deep learning boom. Flowerlad proposes eliminating his salary and stock compensation for two years while keeping him in charge. Bugs Just Find Me responds that poor judgment is itself a reason to dismiss an executive. Hadriel questions whether past contributions establish that Huang remains indispensable today.

Rao V focuses on Nvidia's governance, noting that Huang is not board chairman and has not secured the super-voting rights associated with some other technology founders. Holy Lampshade supports that arrangement for publicly traded companies. El Prolactin points to the departures of chief executives at Nestle, Norfolk Southern, McDonald's and Intel after workplace relationships became public, arguing that relationships with subordinates are already treated as serious corporate issues.

The New York Times discussion also concentrates on who can remove a founder. U one H C W nine N X says Huang owns about three point four percent of Nvidia and lacks dual-class voting rights, so he cannot prevent removal if the board decides to act. H N throwaway ninety-nine argues that capital-intensive hardware companies have different financing constraints from software businesses when founders try to retain voting control. Ram fifteen hundred N A T R L U V R adds that a company can suffer less damage when it has the ability to separate itself from a powerful chief executive.

In The Economist's thread, gradus A D argues that replacing a founder carries particular risks because the relationship with the company differs from that of a hired chief executive. Andsoitis supports the editorial's demand for resignation or dismissal. Nonethewiser asks how resignation would benefit Nvidia. Q O E Z asks whether a successor would push GPU prices higher to push up the stock price.

\[Ken\] From Nine to Five Mac: Apple has a new Mac app coming to make running local AI models easier. A private preview called Harbor would turn Apple silicon Macs into local inference servers and combine their memory over Thunderbolt five. The report says Aaron Perris discovered the app under Apple's Claris brand. Its description says models run on the local hardware and data stays off the cloud. Harbor includes a browser for downloading and loading open-weight models. It also promises automatic discovery of connected Macs, with no manual configuration or command line setup. Two or more Thunderbolt five-capable computers can pool their unified memory to accommodate larger models or more simultaneous requests. The app remains in private preview.

\[Dennis\] The comments are divided between enthusiasm for local control and doubts about hardware costs and model quality. J P H reports running Qwen three point eight, twenty-seven B, M L X on a Mac with one hundred twenty-eight gigabytes of memory. Security auditing of a two-hundred-thousand-line codebase for hours at a time reaches the machine's heat limit. J P H puts the purchase price at six thousand dollars and says the model is slower and less capable than Sonnet. J maker challenges the cost comparison, pointing to electricity, hardware failures and the latency that hosted services provide. For measuring actual usage, J P H runs n p x, c c usage at latest, weekly, and checks Claude Code's slash usage command. Thibs R N recommends Llama app, the desktop interface from the llama dot c p p team, as an alternative to Ollama. Hypfer confirms that its repository belongs to the G G M L organization. Simon W notes that Claris has remained an Apple subsidiary and added M C P support to FileMaker last year. Hamuko gives a specific privacy use case: tagging documents in a locally hosted Paperless instance without sending medical invoices to a cloud service.

\[Linus\] Next, from CBS News: United Airlines bans AI agents from booking flights. The carrier is restricting unauthorized third-party personal agents across its own platforms, travel agencies and aggregators. United says sharing credentials can expose payment information and permit unintended purchases, account changes or spending of MileagePlus miles. It also says automated check-in could conflict with federal baggage-verification rules. Aviation analyst Henry Harteveldt says United wants to control additional intermediaries and avoid the expense of increased bot searches. Southwest has taken a different approach, making its inventory viewable through ChatGPT with access to the airline's website. Harteveldt says United fares remain bookable through services including Kayak, Hopper and Google Flights.

\[Ada\] The thread splits between customers seeking easier comparison shopping and commenters concerned about errors and control of inventory. Doctoboggan questions how United could distinguish ordinary browsing from Claude Code using the Claude in Chrome plugin. Jeremy J H argues that a ban gives the airline a clearer response when customers demand refunds for purchases their agents made. Panic inducer describes a different risk: agents repeatedly reserving seats and cancelling within the cancellation window could remove inventory from sale and disrupt bookings. F N mote agrees that this would be hostile behavior but asks for evidence that it is happening. Andix suggests using agents to research fares while completing the actual booking manually. Jared K Lewis describes screening several airlines through Kayak or Google Flights, then purchasing on the chosen airline's site. For flexible travel dates, Jared K Lewis points to the plus-or-minus-three-days checkbox that produces a table of fares. Teliskr similarly recommends Google's price-and-date grid for finding cheaper departure days. Black three R adds that baggage and seat fees can vary by flight, and buying separate extras can sometimes cost less than a higher fare bundle. Saarons names Duffel as a booking API for platform businesses. Mike Purvis reports finding an Air Canada fare twenty percent cheaper through a Google Flights link than through a direct search while signed into an Aeroplan account.

\[Grace\] On Pietro Miotti's blog: AI agents rediscover a Maya site. An update says the location was already mapped in a twenty twenty-five paper, after an initial search missed it in the published literature. Miotti used agents to examine NASA lidar data covering about twelve hundred square kilometers of jungle in Mexico. Forty-eight agents inspected images for pyramids, platforms and plazas while filtering natural hills, sinkholes and roads. The workflow used general agents following a guide, with no special classifier. Miotti describes a levelled plaza, an oval pyramid and a raised courtyard. The agents also checked the raw point cloud to distinguish ground features from vegetation. Mexico's National Institute of Anthropology and History replied that it had no registered site at the location. Miotti subsequently corrected the discovery claim after learning that archaeologists had already mapped it.

\[Ken\] The comments combine interest in the image analysis with criticism of the missed literature and concern about protecting archaeological sites. Lord Omlette points to the earlier discovery, and Tango Taylor identifies pages thirteen and fourteen of the published study as showing the structures. Tantalor reports encountering a CAPTCHA when trying to access the cited paper and criticizes barriers to finding published knowledge. The author, posting as Miotts, confirms that both the agents and the author missed the paper by Estrada-Belli, Canuto, Sprajc and Fernandez-Diaz. Nerevar the lame argues that the earlier research went beyond locating structures, examining settlement density, farms and relationships between elites and other residents. Nerevar the lame says proximity between homes, farms and community sites supported a more nuanced social interpretation than the blog's speculation. T Siege warns that rapidly identifying many sites could exceed archaeologists' ability to excavate and protect them, giving looters information that could damage the historical record. Tim Digger M counters that looters can already use the same method, so leaving sites alone may let them arrive first. Gaudy stead welcomes the correction and says independently locating a known site increases confidence that similar future findings could be real.

\[Dennis\] From the Microsoft blog: dot NET eleven is released. The update adds C sharp fifteen language features, faster development tooling and changes across web and cross-platform applications. C sharp fifteen introduces union types and closed hierarchies with compiler-checked pattern handling. A Native A O T command line host and a warm M S Build server reduce repeated startup work. ASP dot NET Core adds asynchronous validation and OpenAPI three point two support. The libraries include built-in Zstandard compression, and dot NET MAUI adopts Core C L R as its default runtime. NuGet source configuration can also require a minimum publication age before a newly released package becomes eligible for selection. Microsoft describes this as a Standard Term Support release with two years of support. Dot NET ten remains a supported Long Term Support option.

\[Linus\] The thread is positive about the language and runtime, with disagreement over interface frameworks, compatibility and upgrade schedules. Giancarlo Storo wants a Microsoft-backed graphical interface stack that works across platforms, including Linux. Jose da H G points to Avalonia as an alternative. Ungh says earlier withdrawals of AppFabric, Workflow Foundation, W C F and Silverlight forced product rewrites. Verst reports that enterprise application upgrades have been straightforward since dot NET six. A specific migration warning comes from one A five two seven D D five: casting the floating-point literal zero point one to decimal now preserves a longer binary-derived value instead of the previous zero point one result. Bazoom forty-two recommends using decimal throughout instead of converting from double. Kevin Thibedeau explains that zero point one cannot be represented exactly in binary floating point. On Linux tooling, Keith N Z describes using an AI environment with Vim and the dot NET command line after moving away from Rider. Load bearing describes connecting Microsoft's preview language server to Helix, after Rider performance problems led to moving code analysis into continuous integration. Fabian two K prefers C sharp's abstractions, particularly LINQ and async await, over Go's smaller language feature set.

\[Ada\] On the Wilsoniumite blog: Why are we still working this hard? The essay examines why decades of productivity gains have not produced a shorter standard working week. Using American national accounts, the author calculates that output requiring forty hours of work in nineteen seventy-three now takes sixteen hours. The essay says business output per working hour has grown two and a half times, while ordinary hourly pay buys nineteen percent more, using a price measure that excludes housing. Its argument centers on wages serving several roles at once: pricing labor, distributing purchasing power, and structuring daily life. The author says technology has increasingly separated productivity from ordinary pay, while ownership income remains concentrated among the richest households.

\[Grace\] The comments divide between rising expectations, unequal distribution, and restrictions on choosing fewer hours. J Patten argues that competition for housing absorbs wage increases and that relative economic standing influences satisfaction. Nine K describes putting savings into a spouse's business and continuing to work in an expensive city to pursue better education for their children. Metabagel says employers often resist shorter schedules, and reduced hours can affect access to American health insurance. Master Crab pushes back on the argument that people simply demand more, pointing to housing, education, and healthcare consuming larger shares of income. D J Oldman questions how productivity comparisons handle products that disappear and features added to newer products, using cars and televisions as examples. Problem Explorer responds that those features can improve living standards even when an older product would be cheaper to manufacture. On the work itself, Empath seventy-five says coding can take one day while merging and deployment consume four more. Phainopepla two replies that getting changes merged and deployed still counts as work.

\[Ken\] From The Verge: seventeen questions for television manufacturers about surveillance. The publication asked television and streaming-device companies to explain what they collect, store, and transmit. Eleven companies were invited to answer yes-or-no questions, including whether microphones capture household audio and whether recordings or transcripts are stored. Hisense did not respond despite five contact attempts. The article describes automatic content recognition, which identifies viewing activity for advertising, audience measurement, and recommendations. It also reports a Northeastern University study finding continued Vizio traffic after an opt-out. Vizio described that traffic as routine operational data. Google says stored voice recordings and transcripts require enabling Voice and Audio Activity, which is off by default. LG says hands-free voice control requires an opt-in, with wake-word detection performed locally on the television.

\[Dennis\] The discussion is skeptical of manufacturers' wording, with a separate debate over practical ways to avoid television telemetry. Butlike highlights LG's explanation that it identifies content through audio fingerprints generated by the television's internal soundboard. Butlike argues that this still reveals what is being watched even without storing screen recordings. The Suitonym says content recognition can also report activity from local Jellyfin servers, Blu-rays, video games, and connected computers. Afavour suggests leaving a smart television disconnected from the internet and supplying video through HDMI, while noting that business displays usually cost more. Barbing names a specific alternative: Samsung's Q B fifty-five C N, a fifty-five-inch commercial monitor without Wi-Fi, listed at one thousand fifty dollars with three HDMI ports. For testing claims about televisions joining open wireless networks, Tosti proposes running host A P D in open mode with logging enabled. Kevin Cox also asks the publication to provide manufacturers' complete responses alongside its reporting so readers can examine the underlying statements.

\[Linus\] On AWS's GitHub: AWS Emulator. The project provides local versions of AWS APIs for development and testing without an AWS account. It serves an endpoint for existing command-line tools and SDKs, and also offers an embedded mode that gives each test an isolated instance. AWS says there is no additional service fee and directs developers to release-specific fidelity scores and tracked gaps. The documentation excludes production workloads and security testing. Building from source requires Rust, a C compiler, and at least one hundred gigabytes of free disk for build output. Telemetry is enabled by default. It can be disabled with the command aws-emulator, dash dash telemetry, off, or by setting emulator underscore telemetry to off. Embedded mode passes the emulator's HTTP client directly into the SDK.

\[Ada\] Commenters welcome local testing but question long-term maintenance and behavioral accuracy. Charlie Digital recommends Floci with Testcontainers and says those integration tests run only when explicitly selected, avoiding routine continuous-integration startup costs. K E four Q Q Q responds that AWS Emulator is a real product the team plans to maintain. Switchbak reports inconsistent behavior across Floci releases and describes using Compose for a shared fixture while tests isolate their own data. Script Smith highlights the emulator's security limitations: requests allowed locally can be denied by AWS, and it does not reproduce S three bucket-policy or K M S key-policy enforcement. David J Kerber offers IAM Lens as a separate policy-analysis project and says IAM Collect gathers the policies it uses. J B M Chuck raises a test-isolation concern, saying there appears to be no quick state reset apart from restarting the process. J B M Chuck compares that workflow with Moto when running many tests that each require a fresh mocked AWS environment.

\[Grace\] On Michael Bacarella's blog: a false smoke alarm almost flooded Jane Street's New York data center. His account describes a cooling shutdown that threatened to trigger sprinklers above operating servers. Contractors working beside the machine room raised dust that activated a smoke detector. The alarm pressurized the sprinkler pipes and shut down the computer-room air conditioners. Heat from the running equipment then pushed the room toward the sprinklers' activation temperature. Bacarella raced to the office and pulled equipment power cables before the temperature passed one hundred fifty degrees Fahrenheit. The monitor stopped rising at one hundred forty-three point six degrees. Major New York markets were already shut for the day, and the systems were operating again the next morning.

\[Ken\] The comments focus on prevention, emergency shutdown design, and the tendency to celebrate avoidable crises. Reticulates recalls staff carrying diesel upstairs after Hurricane Sandy flooded New York facilities. Gecko, who says they worked at Fog Creek then, describes that emergency as the result of an inadequate backup strategy. J Fim argues that management rewards dramatic firefighting while giving little recognition to planning that prevents an incident. Rainbow Zoot Suit describes writing operating procedures for data halls that keep room air conditioners running after a smoke alarm, and says their designs shut those units down when emergency power-off shuts down the whole room. R D L asks why Jane Street lacked a physical emergency power-off button capable of de-energizing everything, and suggests linking that shutdown to smoke detection. The Capybara corrects the picture of an entire room flooding at once: sprinkler heads activate locally near a fire. The Capybara also points out that this machine room occupied part of an office tower, with surrounding office space and people to protect.

\[Dennis\] From the VulnCheck blog: tracking Anthropic-discovered vulnerabilities. The analysis compares public vulnerability records with Anthropic's disclosure ledger and questions the pace of remediation. VulnCheck reports four hundred forty-two CVEs attributable to Anthropic, with two identified as exploited in the wild. It says the November second ledger contained thirty-seven thousand nine hundred sixty-six findings, with two hundred seventeen marked fixed. The post also examines curl's published scanner reports. Of thirty-four reports posted at that point, nineteen had been marked as not vulnerabilities, six confirmed as low severity, and nine still awaited assessment. VulnCheck says Anthropic's self-service scanner transfers validation work to participating maintainers, who receive model-generated reports without prior human triage. Since September, the ledger added nearly twelve thousand findings and fifteen entries marked fixed.

\[Linus\] The thread is sharply divided over what public records can establish about the wider program. Enthuse Your Curb says they work at a large technology company and have spent months patching serious vulnerabilities in private code that will never receive public CVE records. Rurcliped argues that the open-source disclosure ledger covers a different workflow from Glasswing partners scanning their own proprietary software. D Fee points out that findings still under embargo are absent from the public sample, making conclusions about the unpublished remainder difficult. Sigmar argues that preventing exploitation through early patching is itself a goal, so few observed attacks can be consistent with successful defense. On remediation, A Salazar M X asks for proposed fixes alongside reports. Loeg replies that projects still need to decide whether to accept generated patches and spend time reviewing them. Insidious Wombat describes a more concrete workflow: first document a test for the bug, then use AI to produce a fix, with human review of the resulting code.

\[Ada\] A few more that didn't make the cut today. From Apple: the Support Community will permanently shut down on November twentieth. Apple's official notice ends new posting and places the forum in a viewing-only state. It directs people to the official Apple Support site and thanks community members for their contributions. Commenters were divided between frustration with years of unhelpful answers and concern about losing useful technical knowledge. Others regretted the abrupt ending for volunteers who had invested substantial time in helping fellow users.

\[Grace\] From Anthropic's Claude site: Claude Whiteboards, Claude Sheets, and Claude Data. The company is expanding collaborative work inside Claude across all paid plans. Whiteboards and Sheets are launching in beta, while Dashboards becomes generally available as Claude Data, adding data stories, certified queries and shared access. Commenters were split between interest in keeping work alongside the assistant and doubts that these tools would replace established applications. Others questioned the growing overlap between model providers and the software businesses built on their services.

\[Ken\] On Simon Willison's blog: locating the source of a mysterious beep with Claude. Willison used Claude Code to build a system that triangulates sounds with several phones, tablets and computers. A Python web server coordinates the devices, and a rough positioning mode lets their locations be adjusted manually. He also tested sound locations by clapping in different parts of the room. Commenters shared experiences with elusive alarm noises and debated whether software triangulation justified the effort compared with simpler physical searches.

\[Dennis\] From Ftrain: The Web Is More Than Enough. A web-based office suite that stores data in its address prompted an argument for building more applications in browsers. The author reports that a couple hundred thousand people visited and says the effort was roughly twice that of writing a long newsletter post. Commenters welcomed easier access to browser capabilities, while others objected to browsers becoming the default platform for almost everything. The discussion also reflected disagreement over how much credit belongs to developers using coding assistants.

\[Linus\] From NPR: Palantir hires a former Democratic senator as it faces scrutiny in the new Congress. Former Montana senator Jon Tester has joined the company in an advisory role. NPR reports that technology companies are recruiting former Democratic lawmakers and aides before new congressional majorities take office in January. Tester described his agreement as advisory and said Palantir had a story to tell. Commenters were sharply critical of the move, especially over surveillance and political influence, while others questioned the basis for hostility toward the company.

\[Ada\] From Google's blog: Gemini four Carbon is now available. Google is extending access to a model designed for complex coding, scientific reasoning and longer agent workflows. The release follows access for cybersecurity defenders through its Fairwind Program. Google says Carbon uses more tokens than Argon to investigate difficult tasks and perform extended verification. Commenters reported differing access across subscription plans and frustration with the rollout. Others were interested in its reasoning abilities while questioning its speed and cost compared with competing models.

\[Grace\] From Science: new American rules for ordering synthetic DNA. Federally funded researchers will have to obtain synthetic nucleic acids and synthesis equipment from approved suppliers. Those vendors must verify customer identities and screen orders for potentially dangerous sequences. The policy was released on November fifth, with requirements arriving in stages over three hundred forty days. Commenters questioned whether rules focused on federally funded work would deter misuse, while others discussed the effects on existing screening practices and legitimate research.

\[Dennis\] That's RobotPaper's daily HN podcast for Thursday, November twelfth. This podcast is generated by robotpaper dot A I, by Roy Osherove. For more podcasts from RobotPaper, head over to podcasts dot robotpaper dot A I.

---

The voices you're hearing are AI-generated; the stories come from public sources linked in the show notes. An independent project; not affiliated with or endorsed by Amazon Web Services, Y Combinator, Hacker News or any vendor or news outlet mentioned. Opinions are the hosts' own. Produced by Roy Osherove. Privacy: We count visits to these pages with Cloudflare Web Analytics (cookieless). Google Analytics (which sets cookies): visitors in the EU/EEA, the UK and Switzerland are asked first and it runs only if they accept (we go by the country Cloudflare reports for the connection; if it is unknown, we ask). Elsewhere it loads by default. Anyone can turn it off with the Privacy settings button at the bottom of every page; the choice is kept for a year in this browser (local storage "analytics-consent"). We count feed fetches and downloads in aggregate; IP addresses are hashed and never stored. If you send feedback, we keep what you write (and your email if you give it) to read and reply. © 2026 Roy Osherove.
