---
title: "Oct 23: The eMudhra web certificate exit, and more"
show: "RobotPaper's daily HN podcast"
canonical: "https://podcasts.robotpaper.ai/hn/ep/2026-10-23/"
date: "2026-10-23T02:22:24+00:00"
duration: "28:19"
duration_seconds: 1699
audio: "https://podcasts.robotpaper.ai/hn/ep/2026-10-23/2026-10-23-hn.mp3"
chapters: "https://podcasts.robotpaper.ai/hn/ep/2026-10-23/chapters.json"
transcript: "https://podcasts.robotpaper.ai/hn/ep/2026-10-23.txt"
feed: "https://podcasts.robotpaper.ai/hn/feed.xml"
language: "en"
author: "Roy Osherove"
ai_generated_voices: true
---

# Oct 23: The eMudhra web certificate exit, and more

RobotPaper's daily HN podcast · Oct 23, 2026 · 28 min

- Listen: [MP3, 27 MB](https://podcasts.robotpaper.ai/hn/ep/2026-10-23/2026-10-23-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/2026-10-23/
- Chapters (JSON): https://podcasts.robotpaper.ai/hn/ep/2026-10-23/chapters.json
- Transcript (text): https://podcasts.robotpaper.ai/hn/ep/2026-10-23.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
- `00:00:35` eMudhra TLS exit
- `00:03:08` Cash App identity case
- `00:05:20` PHP exploit debate
- `00:07:40` Quad9 leaves France
- `00:09:52` AI views of humans
- `00:11:51` Multiplication proofs
- `00:13:51` Switch 2 jailbreak
- `00:15:45` Google forest clearing
- `00:17:38` Airbnb checkout chores
- `00:19:30` Arc shuts down
- `00:21:21` Xbox metered streaming
- `00:23:02` OpenShell sandboxes
- `00:23:43` PHP embargo apology
- `00:24:17` Security incident map
- `00:24:50` WorldTUI terminal
- `00:25:26` Kushner phone report
- `00:26:01` mitmproxy certificates
- `00:26:35` Epic versus Apple
- `00:27:09` iOS on Android
- `00:27:45` Outro

## Show notes

- eMudhra to exit global SSL/TLS certificate business
- Woman faces six years in prison over Cash App account she never opened
- There isn't an easy exploit for this PHP vulnerability
- Quad9 (DNS Resolver) to Exit France
- Every AI model has the same "controversial" take on humans
- The recently proposed o(n log n) integer multiplication algorithm \[pdf\]
- Jailbreaking the Switch 2
- Google cleared 300 hectares of Finnish forest
- Airbnb bans making guests do dishes and take out the trash
- A farewell to Arc browser
- Xbox wants you to pay 24 to 40 cents per hour to stream games you bought
- OpenShell – A safe, private runtime for fleets of autonomous AI agents
- Tell HN: I unknowingly broke the PHP RCE embargo. I am sorry.
- What’s getting pwned right now
- WorldTUI: I trained a 360M model to pretend to be a terminal
- Jared Kushner Hacked Phone Reveals Secret Negotiations, Millions in Transactions
- mitmproxy adds --emudhra option
- Epic Games Inc v Apple Inc
- Run full iOS 16.5 firmware on Android Devices

### Music

"Sneaky Snitch" Kevin MacLeod (incompetech.com)\
Licensed under Creative Commons: By Attribution 4.0\
http://creativecommons.org/licenses/by/4.0/

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 Friday, October twenty-third, 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, eMudhra is leaving web certificates; the trust debate continues.

\[Ada\] And I'm Ada. We also have Airbnb changing its checkout rules.

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

\[Dennis\] From eMudhra: the company is leaving the global SSL and TLS certificate business. The digital trust provider says it will withdraw after its certificate incident and focus on other services. eMudhra says it revoked the affected certificate authorities within fourteen hours of discovery, that they had issued no end entity certificates, and that its infrastructure and data were not breached. It says this business contributes about one percent of revenue. It cites rising compliance costs, competition from free certificates, and scrutiny it considers disproportionate. Customers are promised free replacement certificates from partners during the transition. The company says its G one RSA roots will keep serving customers through transition, while certificate status and revocation services continue until existing certificates expire or are replaced. It promises a root cause report to browser programs and says other signing and identity businesses are unaffected. This seems to be getting attention because the company's explanation focuses on the size of its business, while the discussion focuses on the consequences for everyone relying on its certificates.

\[Linus\] And the thread is sharply critical, with arguments over both the company's response and what the underlying failure says about certificate trust. One commenter objects that the announcement complains about scrutiny without adequately explaining the failure or apologizing. Another cites the company's earlier account of a script error during key handling, followed by staff selecting the wrong keys. A reply argues that describing those keys as merely weaker understates how unsuitable they were. That distinction matters to the debate: some comments ridicule an apparently deliberate choice of a tiny key, while others point back to the reported process failure. There is also disagreement over whether browsers have finished blocking the affected certificates. One person reports an updated iPhone still accepting them. A reply says Apple has revoked them, while further responses describe mixed results on different devices and browsers. Those are commenters' reports, rather than a comprehensive verification of protection. Another participant describes the burden of updating separate certificate stores across software stacks. The frustration extends beyond eMudhra's exit to the practical difficulty of repairing trust once it fails.

\[Ken\] TechSpot reports that a woman faces up to six years in prison over a Cash App account she says she never opened. Amanda Smith says identity theft has left her defending criminal charges in a state she had never visited. She reportedly spent three nights in jail in Pennsylvania before a court appearance in Arkansas. A scammer allegedly directed a Facebook Marketplace buyer to send one thousand dollars to an account bearing Smith's name. The account used her Social Security number and birth date. Smith denies ever using Cash App. The report describes an unfamiliar account address and network connections inconsistent with the mobile provider she says she uses. Cash App declined to give her account information because she could not prove ownership. She later obtained it through discovery. Her lawyer is seeking dismissal, and she has rejected a plea offer while maintaining her innocence. TechSpot says neither Cash App nor the prosecutor responded to requests for comment. The case seems to be drawing attention because an account associated with someone's identity is being challenged as evidence that they controlled it.

\[Ada\] The comments are angry about the prosecution, but they divide over what the reported technical discrepancies actually establish. One commenter questions why investigators did not pursue the connection information more thoroughly. A reply emphasizes the mismatched address and mobile provider. Another pushes back that network differences and an operating system version cannot, by themselves, establish innocence: a person might connect elsewhere or update a phone. A separate argument centers on plea bargaining. Commenters describe the cost and uncertainty of fighting charges as pressure to accept a deal, and some allege that career incentives discourage prosecutors from abandoning weak cases. Those are their interpretations of the process, not established motives in Smith's case. Others debate Cash App's identity checks, with one demanding stronger safeguards and a reply asking what additional verification would prevent this kind of fraud. Across the discussion, the central concern is how evidence connects an account to the person actually operating it.

\[Grace\] On Elias Taylor's blog: there isn't an easy exploit for this PHP vulnerability. The post challenges claims that the flaw is immediately exploitable across ordinary PHP installations while still urging administrators to patch. Taylor argues that demonstrations either fail against an unmodified server or depend on adding a separate vulnerability to the test application. He criticizes examples requiring local file inclusion, an existing application flaw. The post blames repeated mistakes on relying on language models to design and validate demonstrations. It argues that models can alter test servers until apparent success no longer proves the claim. Taylor compares the situation with earlier vulnerabilities that were much easier to exploit, and says this case should not be described in the same terms. That is the author's argument when posting, rather than proof that working attacks cannot exist. This seems to attract interest because it combines urgent patching decisions with uncertainty about what demonstrations show.

\[Dennis\] And the thread is split between criticism of exaggerated demonstrations and concern that this framing could weaken the response to a serious vulnerability. One commenter says research teams now report working remote code execution against unmodified PHP applications. Replies ask for evidence, and another points to claims from Wiz and Calif. The exchange distinguishes researchers reporting success from a public exploit that readers can independently run. Someone who says they tested demonstrations describes one requiring a separate file inclusion flaw and another triggering a memory corruption bug without demonstrating remote code execution. That supports the criticism of those examples, without resolving the vulnerability's full potential. Another participant, who says they helped remediate the issue before disclosure, describes exploitation as requiring knowledge of the PHP configuration and adjustment. On the other side, a commenter praises operators who temporarily took sites down to protect users and questions the benefit of dialing back concern. The author's patching advice remains explicit despite the disagreement over exploitability.

\[Linus\] Quad9 announces its departure from France. The nonprofit DNS resolver says it will withdraw service on November twenty-fourth after a court ruling requiring domain blocking. Quad9 says the case concerns domains beIN alleges infringe its sports broadcasting rights. Quad9 says the court reduced a financial penalty to one thousand euros per day, but that other pending cases make continued exposure unsustainable. Its objection also concerns privacy and a resolver's role. It says it has no registration or database reliably tying users to jurisdictions. Blocking the domains worldwide could satisfy the order, it says, but would extend a French decision to users everywhere. Quad9 distinguishes its optional protection against malicious domains from mandatory restrictions on content. The withdrawal seems to be getting attention because it combines technical limits on geographic restrictions with a dispute over who should enforce copyright claims.

\[Ken\] The thread is sympathetic, with disagreement about whether its technical explanation fits its existing service. One commenter asks why a resolver that filters malicious domains cannot also filter piracy domains. Someone identifying themselves as Quad9's chief technology officer replies that threat filtering is optional and matches users' intentions, while a third party's mandatory content restrictions do not. They say Quad9 will probably use geographic IP filters to block French users after withdrawal. They acknowledge the apparent conflict, explaining that accepting targeted content blocking would violate the service's mission without providing legally perfect coverage. Another discussion focuses on Quad9's account of an earlier German case, where a VPN bypassed geographic blocking and the company says it was penalized. Commenters argue over where responsibility should shift from a service to someone circumventing restrictions. Suggestions that Quad9 retaliate by disrupting its opponents' domains get pushback too: replies say that would harm users and weaken its claim to neutrality.

\[Ada\] From the ModelCompare blog: AI models converge on a supposedly controversial view of humans. A prompting exercise found answers about people putting belonging or self-justification ahead of truth. ModelCompare says it asked twelve text models for an unhedged opinion, without a system prompt or retries, and grouped answers by hand. It reports that every model produced that theme at least once, with exceptions on other attempts. One Claude response instead argued that people are more decent than their cynical self-image suggests. The blog also asked image models to choose their own criticism. Those pictures mostly portrayed consumers absorbed in phones and damaging the planet. ModelCompare interprets the difference as reflecting familiar arguments in text and familiar editorial cartoons in images. This seems to be attracting interest because different systems produced similar answers, raising questions about their training and how the experiment classifies agreement.

\[Grace\] And the comments divide between people who find that convergence interesting and those who think the question wrongly treats generated language as a personal belief. One commenter argues that the models reproduce patterns rather than hold opinions. A reply agrees that belief is not required, but still asks why this particular answer is so probable. Another suggests that science fiction and writing about AI may influence the response to a request for an AI's judgment of humanity. There is also a concrete challenge to the grouping. One reader notes that an answer about disguising convenience as principle does not necessarily say anything about group conformity. Replies concede that some matches are approximate while arguing that the pattern remains. The disagreement is therefore partly about the models and partly about whether a broad category makes distinct answers appear more alike than they are.

\[Dennis\] From the University of New South Wales mathematics site, a recently proposed little o of n log n integer multiplication algorithm. The discussion concerns both a claimed theoretical advance and how mathematicians can assess difficult proofs with AI assistance. We could not fetch the paper itself, so this account comes from the fetched comments, including passages they quote. Commenters distinguish little o from big O: the proposed bound would grow strictly more slowly than n log n, rather than merely staying within a constant multiple of it. That distinction is the mathematical claim under discussion, not a verified performance result from our source material. A commenter quoting David Harvey says he initially struggled with the papers, then questioned Fable repeatedly and gradually understood parts of the argument. The quoted assessment is tentative: the main result might be correct. This seems to be attracting attention because the discussion connects a possible advance in multiplication with a harder question about who understands the proof.

\[Linus\] The thread is divided between excitement about assistance and concern about mistaking convincing explanations for verification. One commenter argues that mathematics requires communicating why a proof establishes its intended claim, rather than merely producing valid symbol manipulation. A reply says questioning a model can help an expert examine smaller arguments independently. Another commenter, identifying themselves as a contributor, says the result has been formalized in Lean; that remains a claim in the discussion here. Several replies also separate asymptotic improvement from practical speed. Their view is that the result is theoretical, with no demonstrated everyday computing benefit established in the fetched discussion.

\[Ken\] Over on the Pepsipu blog, jailbreaking the Switch two. A software-only, tethered exploit is presented as a route to homebrew, game modifications and save-file backups. The author says it can execute native code in ordinary system modules and the built-in filesystem module. The demonstration includes modifications to Mario Kart World. AI agents helped with the research, but the post explicitly corrects the idea that a single prompt produced the jailbreak. The initial autonomous goal failed, and the exploit took substantial effort over about two weeks. The Mario Kart modification, by contrast, was created with AI assistance in a day. The setup gave Codex controller input and video output through JetKVM devices. A Bluetooth switch mechanism physically pressed the console's power button after crashes. The investigation moved through a memory bug in SmileBASIC and a filesystem vulnerability. This seems to be getting attention because the account shows how software agents and physical testing equipment can support hands-on security research.

\[Ada\] The comments combine admiration for the engineering with disputes over ownership and the risks of modification. A commenter working on an internet-connected device says they have been acting as an agent's hands during update and recovery tests. Replies describe automating power cycling and other physical interactions so debugging can continue unattended. Elsewhere, commenters argue about game keycards: supporters value resale and lending, while critics worry that downloads depend on servers remaining available. Several also warn about account or console bans after tampering. Those are experiences and expectations in the thread, not a guarantee about this exploit's consequences. The practical excitement is tempered by uncertainty about what console owners could lose.

\[Grace\] TechSpot reports that Google cleared three hundred hectares of Finnish forest. Regulators are examining whether work for a major data center project moved ahead before the required environmental assessment was complete. The report says Finland's Permit and Supervision Agency asked Google's local project company to explain its actions. Google is investing thirteen billion euros in new AI infrastructure. It says the clearing complied with Finnish law and that surveys protected areas of high environmental value. The chair of the Finnish Association for Nature Conservation disputes that account, saying satellite images show areas that should have been preserved were cleared. She is seeking a halt to construction while authorities establish whether the project complied with the rules. The fetched report describes an unresolved dispute, rather than a final regulatory finding. This seems to be drawing attention because the pace of data center construction is colliding with questions about local environmental oversight.

\[Dennis\] The thread is heated, with disagreement about both the scale of the loss and the significance of the approval process. One commenter compares the clearing with Finland's wider forestry activity and asks why this project receives particular attention. Replies say the relevant question is whether the rules were followed, regardless of how much forest remains elsewhere. Other commenters object that the article has not established illegality, pointing to Google's denial. A separate discussion asks whether data centers could be built on less sensitive sites. A reply notes that moving farther from Finnish cities can still mean moving into forest. Several commenters also question who benefits from the electricity use and investment, extending the argument beyond the cleared land.

\[Linus\] From The Verge, Airbnb bans demanding checkout chores. New house rules distinguish ordinary tidying from work that would normally be done by a cleaner. The policy applies to reservations booked on or after November twentieth. Hosts cannot require guests to wash and put away dishes, strip beds or start laundry. They can ask for dishes in the sink and used towels gathered together. Trash requests can cover sorting into designated indoor bins, but generally cannot require taking garbage off-site. Airbnb also says hosts cannot prohibit use of a listed amenity such as air conditioning, although they can request that it be switched off at departure. Requests will appear more prominently in listings, and hosts cannot add extra rules through messages or signs. Exceptions can include animal-proof trash storage in bear country. This seems to be getting attention because it puts explicit limits on how much checkout labor a booking can demand.

\[Ken\] Commenters welcome clearer limits, but disagree about where courtesy ends and paid cleaning begins. One says leaving used kitchenware dirty is inconsiderate. A reply distinguishes putting rubbish in a bin from carrying it downstairs to locate a dumpster, and putting sheets in a machine from doing the laundry. Another argues that self-cleaning can make sense for a cabin rental when the lower price reflects the work. The debate also broadens into reliability. A commenter recounts holiday cancellations that drove them back to hotels, while others describe frustration with reviews and support. The reaction suggests that chore restrictions address a specific complaint within a wider argument about predictable service.

\[Ada\] On the Arc website, a farewell to the Arc browser. The team is ending desktop access and directing users toward its replacement, Dia. New Arc accounts stopped being available on October twenty-second. On December twenty-second, Arc for macOS and Windows will stop working and accounts will be deactivated, with associated personal data deleted within thirty days. The company says its engineering and security teams will focus on Dia. It offers imports for passwords, history, extensions and browser organization, and says the familiar browser features remain free with AI optional. Arc also provides export options for people moving elsewhere. Mobile changes begin November twenty-third, when AI features, account sign-in and tab syncing stop. This seems significant because a browser shutdown affects an accumulated working setup, not simply the availability of future features.

\[Grace\] The comments mix disappointment with detailed discussion of replacements and trust. One former user says Zen preserves useful ideas from Arc, including sidebar organization and spaces. Replies debate whether Firefox already supplies the features they need, and whether adding another browser developer is worthwhile. Another commenter remains optimistic about Dia as a browser for organizations, while a reply says important Arc features are still missing. Several ask for Arc to become open source. A broader argument concerns products funded by venture capital: critics say users need an exit plan, while others ask how ambitious new software should otherwise be financed. Across those exchanges, the recurring concern is how much workflow to entrust to a product whose direction can change.

\[Dennis\] Ars Technica reports Xbox's hourly pricing for streaming purchased games. Microsoft is testing prepaid Cloud Playtime packs with Xbox Insiders, separate from buying an eligible game. Introductory pricing is four dollars for ten hours, eight dollars for twenty-five hours, or twelve dollars for fifty hours. That works out to between twenty-four and forty cents an hour. Microsoft says purchased time remains available until used. The packs require a cloud-playable game purchase, except for selected free-to-play titles. The report follows limits introduced to Game Pass streaming in September. This seems to be attracting attention because a familiar game library is being paired with a more visible meter for the remote hardware that runs it.

\[Linus\] The thread is split between people who see reasonable hardware rental and those worried about a shift away from ownership. One commenter challenges the headline's emphasis on games already bought, pointing out that local play does not incur this cloud charge. A reply says the frustration comes from streaming previously being bundled into Game Pass. Other commenters see a useful option for occasional gaming without purchasing a console or powerful computer. A commenter who previously rented a remote gaming machine describes that calculation working for a short burst of play. Skeptics focus on latency, future price increases and dependence on a service remaining available. Several distinguish accepting today's hourly price from wanting cloud gaming to become the only practical option.

\[Ken\] A few more that didn't make the cut today. From NVIDIA's GitHub, OpenShell, a private runtime for autonomous agents. The project uses sandbox policies to limit access to files, credentials, and networks. Its documentation says kernel controls enforce those limits, while formal verification checks policy changes before approval. Credentials are added only to requests headed for approved destinations. Commenters welcomed stronger isolation, but debated whether containers alone offer enough protection against data theft. Others raised questions about telemetry and the practical friction of restricting agents.

\[Ada\] Posted on Hacker News, an apology for unknowingly breaking the PHP security embargo. The author says an early post about upcoming remote code execution vulnerabilities disrupted coordinated disclosure and created extra work. They say they had been told there was no embargo, and take responsibility for their actions. Commenters were divided over where that responsibility belongs. Many thought an undisclosed embargo could not bind the recipient, while others emphasized the consequences of passing along sensitive security information.

\[Grace\] From Pwn Watch, what's getting pwned right now. This dashboard gathers security incidents and indicates whether fixes are available. It groups incidents by severity and patch status, offering a way to follow a crowded stream of vulnerability announcements. Commenters were concerned about the pace of security failures, but skeptical about the site's reliability. The discussion questioned its incident classifications and broken features, alongside broader disagreements over how much of the security surge reflects new threats and how much reflects the surrounding hype.

\[Dennis\] Over on the Simedw blog, WorldTUI, a model trained to pretend to be a terminal. A language model predicts screen changes from keypresses, without a real shell or filesystem underneath. The author describes a three hundred sixty million parameter model that can imitate editors and terminal games, with training aimed at repairing mistakes and maintaining consistency. Commenters were impressed by the compact demonstration and curious about learned interfaces. Others debated whether a learned interface really counts as a computer, or just imitates one.

\[Linus\] From Vanity Fair, a report on Jared Kushner's allegedly hacked phone and financial transactions. The account describes claims that Dutch intelligence monitored his access to a Dubai bank account. The writer attributes the account to sources and describes subsequent reporting into purported transfers and secret negotiations. These are allegations reported by the publication. Commenters reacted with anger and concern about foreign influence, while debating political accountability and why different allegations receive such different levels of media attention.

\[Ken\] On mitmproxy's GitHub, an eMudhra option joins the project. The code change incorporates a compromised eMudhra certificate chain. That connects the certificate controversy to a concrete tool change, with the published patch identifying an intermediate certificate authority and a cross-signed root. Commenters reacted with dark humor about the key failure and curiosity about its implications for inspecting device traffic. Others discussed how trust stores and revocation affect whether the technique works, and questioned what practical value it offers for reverse engineering.

\[Ada\] From the Federal Court of Australia's judgments site, Epic Games versus Apple. The case is prompting a debate about software distribution and control over devices. We couldn't access the judgment text, so we aren't summarizing its orders here. In the comments, supporters of alternative app stores welcomed greater freedom to obtain software outside Apple's store. Skeptics worried about security and trusted app review, while the wider discussion questioned how much discretion Apple would retain and how any changes would work in practice.

\[Grace\] From GitHub, a project that claims to run full iOS sixteen point five firmware on Android devices. The submission presents an experiment in bringing Apple's phone software to different hardware. The submission links to the project website, but provides no technical description, so the headline alone cannot establish how complete it is. Commenters found the experiment intriguing, while debating its usefulness and limitations. Their reactions also ranged from enthusiasm for control over phone hardware to concern about imitation iPhones and the reliability of the result.

\[Dennis\] That's RobotPaper's daily HN podcast for Friday, October twenty-third. Today's discussions covered web certificates, browser changes, and questions of trust and control. We'll be back with the next daily update. 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.
