---
title: "Nov 17: Chase banking delays, and more"
show: "RobotPaper's daily HN podcast"
canonical: "https://podcasts.robotpaper.ai/hn/ep/2026-11-17/"
date: "2026-11-17T03:36:00+00:00"
duration: "28:55"
duration_seconds: 1735
audio: "https://podcasts.robotpaper.ai/hn/ep/2026-11-17/2026-11-17-hn.mp3"
chapters: "https://podcasts.robotpaper.ai/hn/ep/2026-11-17/chapters.json"
transcript: "https://podcasts.robotpaper.ai/hn/ep/2026-11-17.txt"
feed: "https://podcasts.robotpaper.ai/hn/feed.xml"
language: "en"
author: "Roy Osherove"
ai_generated_voices: true
---

# Nov 17: Chase banking delays, and more

RobotPaper's daily HN podcast · Nov 17, 2026 · 29 min

- Listen: [MP3, 28 MB](https://podcasts.robotpaper.ai/hn/ep/2026-11-17/2026-11-17-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/2026-11-17/
- Chapters (JSON): https://podcasts.robotpaper.ai/hn/ep/2026-11-17/chapters.json
- Transcript (text): https://podcasts.robotpaper.ai/hn/ep/2026-11-17.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` Chase banking delays
- `00:03:07` Nvidia succession
- `00:08:01` Amazon sale prices
- `00:10:30` Qwen4-Max launch
- `00:12:52` Child safety tool
- `00:14:43` Solar grazing
- `00:16:42` Frame ePaper display
- `00:18:37` Software reliability
- `00:20:37` Crow decision speed
- `00:22:26` Bitly link hijacking
- `00:24:17` El Niño flight weather
- `00:24:55` BugMeNot closes
- `00:25:27` Windows tool telemetry
- `00:26:05` Microsoft gateway bug
- `00:26:42` uBlock picker flaw
- `00:27:14` Haskell technical plan
- `00:27:54` Apollo seismic tests
- `00:28:30` Outro

## Show notes

- [Technical issue at JPMorgan Chase triggers online banking delays for US banks](https://apnews.com/article/bank-outage-chase-wells-fargo-truist-50fee2693c22f597fa6ca51a97ef2cda) · [HN discussion](https://news.ycombinator.com/item?id=50521045) \| quoted: [Miraste](https://news.ycombinator.com/user?id=Miraste)

### Nvidia succession (3 stories)

From Steve Blank's blog, what's next for Nvidia?

- [What's Next for Nvidia?](https://steveblank.com/2026/11/15/whats-next-for-nvidia/) · [HN discussion](https://news.ycombinator.com/item?id=50512235) \| quoted: [areoform](https://news.ycombinator.com/user?id=areoform)
- [Nvidia was in trouble before the Jensen Huang mess](https://www.wired.com/story/nvidia-was-already-in-trouble-before-the-jensen-huang-mess/) · [HN discussion](https://news.ycombinator.com/item?id=50518567) \| quoted: [coreylane](https://news.ycombinator.com/user?id=coreylane)
- [Jensen Huang might return to Nvidia as CEO](https://www.reuters.com/world/china/jensen-huang-discusses-comeback-nvidia-ceo-with-allies-ft-reports-2026-11-15/) · [HN discussion](https://news.ycombinator.com/item?id=50511982) \| quoted: [waximabbax](https://news.ycombinator.com/user?id=waximabbax)
- [Many of Amazon’s Black Friday and Cyber Monday “deals” offered no savings](http://groundworkcollaborative.org/news/amazons-decoy-deals-groundwork-investigation-reveals-nearly-1-in-3-black-friday-and-cyber-monday-deals-advertised-by-amazon-offered-no-savings-to-shoppers-nearly-1-in-10-w/) · [HN discussion](https://news.ycombinator.com/item?id=50517488) \| quoted: [Aurornis](https://news.ycombinator.com/user?id=Aurornis), [Fogest](https://news.ycombinator.com/user?id=Fogest)
- [Qwen4-Max](https://qwen.ai/blog?id=qwen4-max) · [HN discussion](https://news.ycombinator.com/item?id=50513215) \| quoted: [Aurornis](https://news.ycombinator.com/user?id=Aurornis)
- [Strengthening the fight against online child sexual abuse](https://openai.com/index/strengthening-online-child-safety/) · [HN discussion](https://news.ycombinator.com/item?id=50520804)
- [Grazing sheep in fields with solar arrays](https://news.cornell.edu/stories/2026/10/sheep-grazing-solar-fields-offers-farmers-economic-ecological-gains) · [HN discussion](https://news.ycombinator.com/item?id=50520182)
- [Frame – 31.5" color ePaper display](https://trmnl.com/blog/introducing-frame) · [HN discussion](https://news.ycombinator.com/item?id=50519354)
- [Software is breaking down at a systemic level](https://www.breck-mckye.com/blog/2026/11/software-is-breaking-down-at-a-systemic-level/) · [HN discussion](https://news.ycombinator.com/item?id=50520866)
- [Crows make decisions faster than we can](https://www.nytimes.com/2026/11/15/science/birdbrain-crows-birds-research-decision.html) · [HN discussion](https://news.ycombinator.com/item?id=50509023) \| quoted: [Animats](https://news.ycombinator.com/user?id=Animats), [Vetch](https://news.ycombinator.com/user?id=Vetch)
- [Old Bitly links are being hijacked](https://ransomnews.com/burger-king-tweet-bitly-link-hijack-2026/) · [HN discussion](https://news.ycombinator.com/item?id=50511395)
- [A short explainer on how El Niño works \[video\]](https://www.youtube.com/watch?v=7_JF3etiAv8) · [HN discussion](https://news.ycombinator.com/item?id=50509308)
- [BugMeNot has closed](https://www.bugmenot.com/) · [HN discussion](https://news.ycombinator.com/item?id=50523601)
- [Windows 11 Is Collecting Data on Which Coding Agents and Developer Tools You Use](https://xusheng.dev/posts/windows-coding-agent-telemetry/) · [HN discussion](https://news.ycombinator.com/item?id=50524736)
- [Microsoft blames monster email outage on software defect introduced in July fix](https://www.theregister.com/saas/2026/11/16/microsoft-blames-monster-email-outage-on-software-defect-introduced-in-july-fix/5310026) · [HN discussion](https://news.ycombinator.com/item?id=50517321)
- [Universal XSS in uBlock Origin from website-driven element picker](https://iter.ca/post/ubo-uxss/) · [HN discussion](https://news.ycombinator.com/item?id=50513160)
- [A technical vision for the Haskell ecosystem](https://discourse.haskell.org/t/a-technical-vision-for-the-haskell-ecosystem/15078) · [HN discussion](https://news.ycombinator.com/item?id=50484235)
- [Why did some Apollo missions carry a grenade launcher? (2019)](https://space.stackexchange.com/questions/37775) · [HN discussion](https://news.ycombinator.com/item?id=50505271)

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 Tuesday, November seventeenth, 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, a Chase software bug delays banking transactions.

\[Ada\] And I'm Ada. Nvidia faces a leadership transition; Amazon's advertised discounts face scrutiny.

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

\[Dennis\] First up, from the AP: a technical issue at JPMorgan Chase triggers online banking delays for U.S. banks. A software bug at Chase disrupted access for its customers and slowed transactions involving other financial institutions. Chase spokesman Joe Evangelisti attributed the problem to an on-premise hardware system. The bank was working to restore services by the end of Monday. Bank of America and Citigroup were not experiencing problems with their own banking services. U.S. Bank said its systems were unaffected, although customers sending or receiving Zelle payments involving another institution could face delays. Early Warning Services, which operates Zelle, said the payment network was fully operational. Some Chase customers instead encountered a message saying that certain services were unavailable. The disruption followed separate outages affecting Microsoft services, air traffic control and United Airlines.

\[Grace\] The discussion was anxious, with disagreement over whether the recent outages reflected shared operational problems. Bobson three eighty-one described how nontechnical colleagues saw the Microsoft outage, the air traffic ground stop and banking delays as interconnected. Legend twenty-four forty argued that the outages had coincided within the same news cycle. A D E P T compared the banking disruption with Revolut's earlier Google Cloud outage. Zenexer pointed to Chase's statement identifying a software bug in its own hardware system. The comments also examined how outage reports should be interpreted. Ceejayoz questioned DownDetector spikes for services including Steam, American Express and Visa, arguing that attention to one outage could generate reports about other companies. Miraste defended the service as noisy but useful and recommended checking the highest-ranked failing sites on its homepage. Ceejayoz challenged that method by pointing to Google's high position in the list. On software testing, J gilias recalled a workplace that had employed a capable quality-assurance team before laying the entire team off. Lanthissa described a different experience on trading and banking applications, where insufficient testing was accepted as a reason to delay a release. K C C Q Z Y said Zelle exposes their legal name to transaction partners, and contrasted that with Venmo's option to use a pseudonym for secondhand purchases.

\[Linus\] From Steve Blank's blog, what's next for Nvidia? The company's leadership transition coincides with growing competition from customers developing their own chips. Three stories examine that transition. Blank says Nvidia's board announced Huang's resignation thirty-three years after he cofounded the company. Blank describes how Jensen Huang moved Nvidia from graphics boards into a software platform with CUDA, and then into integrated data center systems through Mellanox. He argues that a successor focused on operating the existing business could cut investments needed for the next transition. Blank argues that operating executives prioritize stability and repeatable processes, and that promoting them can drive creative employees to leave. Huang had about sixty direct reports, discouraged one-on-one meetings, and had not named a successor. Blank identifies Google, Amazon, Microsoft, Meta, OpenAI and Tesla as customers building competing chips. Blank uses Microsoft as a comparison. Steve Ballmer tripled sales to seventy-eight billion dollars and more than doubled profits, while the company missed major shifts in search and mobile computing. Blank credits Satya Nadella with reorganizing around Azure and freeing the Office and Azure teams from dependence on Windows. Wired's report, Nvidia was in trouble before the Jensen Huang mess, examines that competition in detail. It says chief financial officer Colette Kress has taken charge. Nvidia holds about eighty percent of AI accelerator sales, according to Silicon Analysis. AMD's share rose from two percent in twenty twenty-two to between seven and ten percent this year. Wired also describes Taalas, acquired by AMD, which builds model weights into silicon, and Nvidia's acquisition of Groq. Wafer chief executive Emilio Andere says his company can put newly released open-weight models onto AMD hardware within hours, at least eighty percent as efficiently. From Reuters, Jensen Huang might return to Nvidia as chief executive. Reuters, citing the Financial Times, says Huang has discussed a comeback with confidants. Reuters says he stepped down Thursday following revelations of extramarital affairs. An ally told the Financial Times about efforts to organize Silicon Valley executives to lobby for his return and seek support from the White House.

\[Ada\] And the discussions divide between confidence in Nvidia's engineering platform and concern about its dependence on one leader. In the thread on Steve Blank's post, areoform argued that Nvidia should acquire companies that build model weights into silicon and sell that capability as a service, even if it reduces sales of existing products. Coolest Beans disputed that strategy, arguing that specialized inference chips would give Nvidia less control over demand and less ability to retain customers through its software. Mike Hearn replied that software polish was itself an important part of CUDA's appeal. Madaxe again emphasized that CUDA had been developed for twenty years, and pointed to Jetson Nano and its surrounding software as evidence that Nvidia had already invested in lower-power edge inference. Amarshall raised a different concern about customers' private chips. They argued that alternatives such as Google's tensor processing units and Amazon's Inferentia do not offer the same publicly purchasable hardware market as Nvidia's products. In the Wired discussion, coreylane compared AI accelerators with cryptocurrency mining, where specialized chips displaced graphics processors. Jack P P answered that Bitcoin's algorithm is static, while AI model designs are still changing quickly. J P G V M, who said they previously worked at Groq and now work on Nvidia's L P X, described its language processing units as a compromise between specialization and flexibility when architectures change. Liz Knope, who said they have designed complex chips for thirty years, emphasized the production timeline. They said the process from architecture through manufacturing, testing and mass production takes years, and that fabrication at T S M C alone takes four months from design delivery to finished chips. Lenerdenator focused on governance, arguing that boards should prepare for a founder's departure and establish a succession plan before a crisis. In the Reuters thread, waximabbax argued that a public-company board cannot simply reverse its decision under pressure from Silicon Valley executives. Trebligdivad suggested Huang could start another company with departing Nvidia employees.

\[Ken\] From Groundwork Collaborative's blog, many of Amazon's Black Friday and Cyber Monday deals offered no savings. An investigation compared advertised holiday discounts with prices available during the preceding month. Groundwork examined more than three hundred and fifty products featured on Amazon's own deal pages for the twenty twenty-four and twenty twenty-five events. It found that nearly one in three had previously sold for the same price or less, and nearly one in ten cost more during the sale. Yellow Hunter women's short rain boots advertised for one hundred twenty-six dollars and ninety-nine cents on Black Friday twenty twenty-five had cost seventy-nine dollars and ninety-seven cents nine days earlier. Among the products with worse deals, the median sale price was nineteen percent above the earlier price. The report calls for retailers advertising discounts to display the lowest price from the previous thirty days.

\[Dennis\] And the comments are skeptical of sale labels, with disagreement over whether Amazon's delivery convenience still justifies using it. Ilamont said canceling Prime reduced impulse purchases and encouraged batching necessary orders until they qualified for free shipping. Strong like Dan replied that Amazon's shipping speed was the deciding benefit, with Walmart covering some remaining purchases. Teh like recommended Camel Camel Camel for tracking Amazon prices over several weeks. Fogest said Amazon's own price-history display in Canada matched both Camel Camel Camel and Keepa in their comparisons, while observing that the built-in tool appeared to omit third-party sellers and used goods. Rancar two raised limitations in third-party tracking as well, saying Amazon's data feed can miss short-lived promotions and does not include checkout codes or clippable coupons. Yathern described another problem while tracking mini computers on AliExpress: sellers removed listings and brought back the same specifications under different names, disrupting trackers tied to a product link or stock identifier. Aurornis said their comparison method was to put each vendor's item in the cart and continue to checkout to see the actual price after discounts. Madibo thirty-one fifty-six described a first-item promotion where the discount depended on the order products entered the cart. They removed cheaper items, kept the one with the largest discount, and then added the other products back.

\[Grace\] From the Qwen blog: Qwen four Max. The new model combines native vision and a one-million-token context window with an architecture designed for extended professional tasks. Qwen reports a two-point-four-trillion-parameter backbone, with seventy-one billion parameters activated per token and an additional three hundred twenty-eight billion N-gram embedding parameters. Its examples include creating scenes in Blender and developing games in Unity, Unreal Engine and Godot. In Qwen's Work three sixty evaluation, the reported average cost per task falls from four dollars forty-three to one dollar sixty-seven compared with Qwen three point eight Max. The model is available through QwenCloud's Token Plan and API, and through QwenWork, Qoder and QwenStudio. Qwen also released a plugin for delivering successive usable versions of a long-running task. The announcement says smaller models in the same family are coming to open source.

\[Linus\] The comments were divided over benchmark comparisons, inference speed and the prospects for open models. Maplet cited an Artificial Analysis score of forty-three at an average cost of three dollars twenty-seven, comparing that with a score of forty-six at thirty-six cents for GPT six point one Sol at medium effort. Esperent objected to comparing different reasoning settings and cited a cost of one dollar twenty-seven for Sol at maximum effort. A throwaway three Z argued that comparisons should use equal resources or measure the cost of completing the same task, because reasoning labels differ between models. Aurornis recommended the cost-per-task Pareto chart, which includes multiple effort settings, instead of the default maximum-effort ranking. Seize the cheese described an agent-comparison service built around OpenCode. In that service, Qwen three point eight generated about thirty tokens per second against roughly eighty for Opus, and early voting made the slower model perform worse. Seize the cheese said the models produced similar amounts of output per task and wanted more third-party serving options. Simon W reported that Qwen four Max adds a setting that disables reasoning, while the earlier Max required choosing a reasoning level.

\[Ada\] From OpenAI, strengthening the fight against online child sexual abuse. The company is offering vetted organizations a free image classifier through its API to help identify potentially abusive material. OpenAI says the tool can flag previously unidentified images and indicate whether material may be AI-generated. Its scores help human reviewers prioritize cases. The company developed and evaluated the classifier with assistance from the National Center for Missing and Exploited Children. In an internal comparison with its existing protections, OpenAI estimated that the new classifier would flag forty-eight percent fewer images while finding roughly the same amount of abusive material. In testing by early partner fal, human reviewers confirmed seventy-two percent of high-confidence flags as previously unidentified child sexual abuse material.

\[Ken\] And the thread combines support for reducing investigators' workload with concern about account bans and opaque review processes. Max gee said an autonomous agent switched an image-classification task from a local model to OpenAI, and their account was banned after processing more than one hundred thousand user-uploaded images. Max gee said requests to identify the flagged material and report its uploader received no usable information, and the appeal was denied. None the wiser replied that a failed appeal process does not establish that the company is indifferent to child protection. Eutropia welcomed a model that could reduce investigators' exposure to disturbing material and help them work through the additional volume created by image generators. Delichon warned that sending unreviewed user uploads to a service that refuses this material can put an account at risk, even when the intended task is classification.

\[Dennis\] From Cornell's newsroom, grazing sheep in fields with solar arrays. Two studies examine how flock size affects beneficial insects and the economics of maintaining solar sites. In a three-year experiment at Cornell's Cascadilla solar farm, two sheep per hectare increased the combined abundance of wild bees, honey bees and lady beetles by thirty-eight percent compared with mowing. Four sheep per hectare controlled vegetation more effectively, with insect abundance similar to the mowed plots. At eight sheep per hectare, honey bee abundance fell forty-five percent. A separate Cornell team developed a spreadsheet model for sheep purchases and sales, incorporating flock targets, costs, expected revenue and vegetation. United Agrivoltaics has incorporated that tool into its onboarding process. The grazing experiment used Cornell's teaching flock at a university-owned site in Dryden, New York.

\[Grace\] The comments were interested in combining farming and electricity production, with debate over land use and practical installation costs. M L monkey asked why solar arrays should occupy grassland when deserts are available. Noddingham pointed to transmission and the value of generating electricity near demand. Heimdall, who keeps around twenty cattle of smaller breeds, described a different obstacle: finding affordable racks that could withstand thousand-pound steers rubbing against them. Heimdall also reported difficulty feeding electricity back to the local grid. Andrew F L N R suggested protecting the rack posts with bollards made from pipes filled with cement. Heimdall said inexpensive structural steel pipe sometimes appears on local Marketplace listings and welcomed the filling idea. Blue Gill added that cattle need strong posts, while goats require structures that resist chewing and support animals climbing onto the panels.

\[Linus\] On the T R M N L blog, Frame is a thirty-one-point-five-inch color ePaper display. The device, now available for preorder, combines a handmade wooden enclosure with the company's existing display platform and several new hardware options. Its microSD slot supports completely offline operation. Clock Sync schedules updates at particular times of day, and an additional rear USB-C port supports continuous external power. T R M N L is developing Lazarus, a color conversion engine it plans to open-source. The first batch is roughly four hundred seventy units. The company intends to ship a few before the end of this year, with most deliveries expected in early twenty twenty-seven.

\[Ada\] The thread mixed enthusiasm for the display with objections to price and confidence in delivery. Calvin zero two said earlier T R M N L purchases had been followed by declining quality assurance and late shipments, and also objected to the founder's politics. Vunderba put Frame's price at twelve hundred dollars, with an increase to thirteen hundred planned. Vunderba described an alternative already operating at home: a secondhand Samsung Frame television connected to an old laptop. The laptop generates images and transfers them through Samsung's WebSocket API, with the picture changing every thirty minutes. Vunderba also connected the television to the local network while blocking its outbound internet requests at the router. For smaller displays, Asenna pointed to the seven-point-five-inch reTerminal E one thousand one, with microphones and a beeper, at sixty-nine dollars. S one N added that the color E one thousand two costs about one hundred dollars, while the thirteen-point-three-inch color E one thousand four costs two hundred eighty dollars.

\[Ken\] From the Breck McKye blog, software is breaking down at a systemic level. The essay argues that growing technical complexity and changes in software production are weakening organizations' ability to manage failures. It examines recent Microsoft service disruption, the PHP two Shell vulnerability and the Shai Hulud supply-chain attack. The author describes cloud infrastructure and dependency-heavy applications as systems whose individual components can be difficult for any team to understand completely. AI appears in the argument both as a tool for discovering complicated vulnerabilities and as a way to accelerate development. The essay also traces how cloud services and reusable libraries changed the division of work between developers, operations staff and managers.

\[Dennis\] The comments were divided over that diagnosis, with substantial disagreement about the historical timeline. Doug Durham said production cloud computing and the open-source movement predated the essay's account, and argued that software's scale and ambition had grown alongside computing capacity. Charlie Digital focused on technical writing, recommending explicit patterns, specifications and test plans to guide coding agents. Charlie Digital also described zeeq-biome, an open-source library for improving agents' unit and integration tests through denser tests and type-checked setup. Dotdi countered with an account of rushed reviews after layoffs. Dotdi said a DevOps manager nearly accepted Opus five point five's recommendation to disable JIT compilation across a performance-sensitive product to address a HotSpot C two optimizer bug. According to Dotdi, the actual fix was a one-line adjustment in a library. C miles seventy-four recommended questioning a model's assumptions through conversation instead of expecting a complete solution from one prompt.

\[Grace\] From The New York Times: crows make decisions faster than humans. Researchers studied how two carrion crows accumulated visual evidence before choosing a direction. At the University of Tübingen, the birds watched moving dots on a touchscreen and pecked a target indicating whether most dots traveled left or right. Their average response took two hundred thirty-five milliseconds, and accuracy approached one hundred percent when the signal was clear. Recordings from individual neurons showed distinct activity for the possible choices after about seventy milliseconds. The findings appeared in Science. The researchers described decision-making dynamics resembling established models for humans and other primates, despite the different organization of bird brains.

\[Linus\] The comments were interested in alternative brain architectures and divided over explanations for the speed. Animats pointed to mammals, corvids and octopuses as examples of intelligence emerging with different nervous-system designs. Vetch added parrots, citing high neural density and similarities between the pallial neuron counts of keas and ravens. Walter Bright suggested that shorter connections in smaller brains could help explain faster decisions, and compared the issue with the feedback involved in piano playing. Bob ten twenty-nine disputed that account of musical learning, arguing that predictive models let people anticipate events and compensate for latency. Seba dos one offered a practical example using Jamulus for playing music over the internet. Seba dos one recommended monitoring the player's own sound after the network round trip, allowing the player to adjust to the delay and synchronize with the other musicians.

\[Ada\] From Ransomnews, old Bitly links are being hijacked. Reissued short-link endings redirected traffic from years-old posts to destinations chosen by new owners. The report verified affected links in Burger King, McDonald's and Donald Trump posts. A restaurant-locator link in Burger King's twenty eleven tweet was registered again on November fifteenth and redirected to Wikipedia's list of Nazi concentration camps. Bitly subsequently confirmed that an issue had allowed old, inactive links to be repurposed. It suspended the accounts involved and began disabling affected links. The company said its Trust and Safety team was developing additional protections. By the report's latest check, the tracked links pointing to new destinations had stopped redirecting. Forty-nine re-created links in the Trump sample still worked and pointed to what appeared to be the original destinations.

\[Ken\] The comments were sharply critical of outsourced shorteners, while discussing the practical reasons organizations still use them. X P eighty-four argued that a direct link depends on fewer organizations surviving and maintaining their systems over time. Frio said their company uses shorter addresses to make QR codes work better in poor lighting and on weaker phones, but operates the redirect service itself. Silver Slash offered a simpler option for creating a QR code: right-click a page in Chrome and choose Create QR code for this page. G B R raised a related ownership problem, describing spam sent through reputable organizations' old CNAME records after the destination domains changed hands. G B R recommended tracking those records, asking users to report when they stop using them, and periodically checking whether their target domains have new owners.

\[Dennis\] A few more that didn't make the cut today. From YouTube, a short explainer on how El Niño works. The video description outlines the winter twenty twenty-six to twenty twenty-seven outlook for general aviation. It follows Pacific trade winds, ocean temperatures and shifting thunderstorms into changes in the winter jet stream, with implications for icing and mountain visibility. Commenters welcomed accessible weather explanations and climate data, while others criticized the video's United States focus and debated how warming changes familiar weather patterns.

\[Grace\] On BugMeNot's website, BugMeNot has closed. The site now displays a short shutdown notice. It thanks everyone who envisioned a better internet and follows that message with a block of uppercase letters arranged in groups. The notice gives no explanation for the closure. Commenters mixed nostalgia for the service with frustration about compulsory website accounts. The discussion also reflected disagreement over how reliably shared logins had continued to work and whether alternative approaches were more convenient.

\[Linus\] From Xusheng's blog, Windows eleven tracks coding agents and developer tools. An investigation found a Windows diagnostics component recording tool usage on the author's test machine. Its executable-name checks include Codex, Claude and Cursor, alongside tools such as Python. The report says this usage data is uploaded when optional diagnostics are enabled. Commenters expressed frustration with Windows telemetry and enthusiasm for switching to Linux, while others emphasized the applications and gaming requirements that still complicate a switch.

\[Ada\] From The Register, Microsoft's email outage traced to a July software change. A latent defect exhausted gateway resources when increased traffic coincided with reduced server capacity. Microsoft's preliminary report describes Windows lock contention that worsened on machines with more processor cores, while testing had concentrated on weaker hardware. Engineers restored the earlier code path through a new update. Commenters questioned the mismatch between testing and production equipment and debated how large services should prepare to recover from cascading failures.

\[Ken\] On Iter's site, a universal cross-site scripting flaw in uBlock Origin. The now-fixed issue allowed a malicious website to trick the element picker into adding trusted filters that could execute JavaScript on other sites. Firefox's uBlock Origin was affected, with a less severe related issue in uBO Lite. The researcher verified the finding after assistance from an AI model. Commenters debated the security risks of embedding extension controls inside web pages and the trade-offs involved in restricting browser extensions.

\[Dennis\] On the Haskell community forum, a technical vision for the Haskell ecosystem. The Haskell Foundation's technical leadership committee is seeking initiatives, funding and contributors to improve the language's practical tooling. Its goals include faster iteration, better debugging, easier onboarding and stronger compatibility. One target is cutting GHC's continuous integration turnaround from six hours to two. Commenters welcomed attention to developer experience but disagreed about installation complexity, the benefits of separate tools, and Haskell's prospects in AI-assisted programming.

\[Grace\] From Space Exploration Stack Exchange, why some Apollo missions carried a grenade launcher. A question from twenty nineteen asks about a launcher and four grenades listed in an Apollo experiment manual. The answer explains that explosions of known size, measured with a seismometer, reveal how shock waves travel underground and help identify subsurface structure. Commenters showed interest in lunar experiments and spacecraft equipment, while the discussion also split over Stack Exchange's moderation practices and the value of carefully documented questions and answers.

\[Dennis\] That's RobotPaper's daily HN podcast for Tuesday, November seventeenth. 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.
