---
title: "Oct 22: PHP's critical RCE scramble, and more"
show: "RobotPaper's daily HN podcast"
canonical: "https://podcasts.robotpaper.ai/hn/ep/2026-10-22/"
date: "2026-10-22T02:21:10+00:00"
duration: "26:58"
duration_seconds: 1618
audio: "https://podcasts.robotpaper.ai/hn/ep/2026-10-22/2026-10-22-hn.mp3"
chapters: "https://podcasts.robotpaper.ai/hn/ep/2026-10-22/chapters.json"
transcript: "https://podcasts.robotpaper.ai/hn/ep/2026-10-22.txt"
feed: "https://podcasts.robotpaper.ai/hn/feed.xml"
language: "en"
author: "Roy Osherove"
ai_generated_voices: true
---

# Oct 22: PHP's critical RCE scramble, and more

RobotPaper's daily HN podcast · Oct 22, 2026 · 27 min

- Listen: [MP3, 26 MB](https://podcasts.robotpaper.ai/hn/ep/2026-10-22/2026-10-22-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/2026-10-22/
- Chapters (JSON): https://podcasts.robotpaper.ai/hn/ep/2026-10-22/chapters.json
- Transcript (text): https://podcasts.robotpaper.ai/hn/ep/2026-10-22.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` Paris 1.5 degree goal
- `00:02:25` PHP 8.2.35 release
- `00:04:13` No Kings takedown
- `00:05:55` eMudhra CA keys
- `00:07:55` PHP Foundation post
- `00:09:59` BioWare shuts down
- `00:11:53` Claude Duet
- `00:13:43` Meta privacy penalty
- `00:15:33` PHP2Shell from Wiz
- `00:17:35` Patch the Planet
- `00:19:35` Railway breach
- `00:21:37` Fire escapes
- `00:22:15` LiteLLM backdoor
- `00:22:57` Advice fits the giver
- `00:23:30` FORTRAN's old promise
- `00:24:05` Google Rust guide
- `00:24:39` Qwen3 Coder US persona
- `00:25:10` OpenAI on Google Cloud
- `00:25:44` C# beats Rust on WASM
- `00:26:22` Outro

## Show notes

[robotpaper.ai](https://robotpaper.ai) · [Hacker News](https://news.ycombinator.com)

- The 1.5°C goal of the Paris Agreement is no longer viable
- PHP 8.2.35
- The No Kings Takedown Is Why You Shouldn't Rent Your Megaphone from Zuck
- Eight Bits of Trust: eMudhra's CA keys cracked in milliseconds
- Today's PHP security release, and how we're strengthening security
- EA is shutting down BioWare
- Claude Duet
- Meta ordered to pay $25B penalty in New Mexico Cambridge Analytica case
- PHP2Shell: AI uncovers critical RCE in PHP
- Patch the Planet: global response coordination as agents uncover vulnerabilities
- Railway Security Disclosure
- Whatever Happened to Fire Escapes
- Compromised LiteLLM 1.107.0 on PyPI Installs a Credential Stealer and Backdoor
- Most advice is right for the person giving it
- The compiler was supposed to end programming too
- Google's Rust Style Guide
- Qwen3-Coder writes less secure code when told it's for a US persona
- OpenAI models are coming to Google Cloud
- C# now beats Rust on WASM size

### 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 Thursday, October twenty-second, 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, PHP had a very rough day; patch first, read the blog posts later.

\[Ada\] And I'm Ada. We've also got a certificate authority, whose private key is the number four; and BioWare, closing its doors after three decades.

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

\[Linus\] First up, from UNRIC, the United Nations Regional Information Centre: the one point five degree goal of the Paris Agreement is no longer viable. The UN Environment Programme's new Emissions Gap Report says keeping warming to one point five degrees without overshooting is now out of reach. The new target is to stay below two degrees, then bring the temperature back down to one point five by the end of the century. To get there, global emissions would need to fall eighteen percent by twenty thirty and forty-five percent by twenty thirty-five, compared with twenty twenty-five. Even if every national pledge is fully delivered, cuts would only reach five to eight percent by twenty thirty. The report also says about three hundred and twenty gigatons of carbon dioxide would need to be pulled out of the atmosphere by the end of the century, about eight times the emissions of twenty twenty-four. It seems this hit a nerve because it's the official admission that a target many people grew up with is gone.

\[Ken\] And this was one of the busiest threads of the day, and a fairly gloomy one, split between people who blame politics and markets and people who blame demographics or other countries. The top comment came from someone who says they've worked in climate tech for twelve years and just spent ten months finding a job, because they insisted on clean energy hardware while the money moved to A I. They argued we crossed the threshold because the money can't help itself. Another commenter argued for carbon pricing, saying politicians ban plastic straws while China burns more coal. A reply pushed back that the United States emits more per person than China, and that China leads the world in clean energy investment. And a skeptical thread argued that carbon capture is a distraction, because warming effects like lost sea ice don't simply reverse when the temperature comes back down.

\[Grace\] Next up, from PHP's GitHub: PHP eight point two point thirty-five. The PHP team has shipped emergency security releases for two critical bugs that allow unauthenticated remote code execution. The release notes say there are no complete mitigations, and updating is the only safe choice. All PHP eight point two users are urged to update immediately, and that branch reaches end of life at the end of December. The flaws, nicknamed PHP2Shell, came out through a coordinated disclosure involving the security firms Wiz and Calif, and OpenAI. The timeline people quoted says Wiz reported the bugs on August twenty-fourth, and PHP acknowledged them on September twenty-first. It seems this is getting so much attention because PHP still runs a huge share of the web, and a bug like this means patching right now, not next sprint.

\[Dennis\] And the mood in the thread was angry, mostly about how the disclosure was handled rather than the bug itself. One camp felt left out to dry, arguing that OpenAI and the security firms sat on this for two months while picking who got early fixes, and that it smelled like a P R stunt. A reply pushed back that the reporters flagged it two months ago, PHP took a month just to acknowledge it, and publishing a fix effectively publishes the details. Another commenter noted that, according to Calif's write-up, the underlying issue had been reported back in twenty nineteen. A lot of the frustration was practical: people watched the official Docker images lag for hours, and one user found a freshly rebuilt image that still shipped the unpatched version. Someone who says they've hosted PHP for twenty years said it's the first time they've ever had to take servers down preemptively.

\[Ada\] Moving on, from Techdirt: The No Kings Takedown Is Why You Shouldn't Rent Your Megaphone from Zuck. Techdirt argues that whether or not Meta meant to silence a protest group, the real problem is that it could. During the latest round of No Kings protests, part of a get-out-the-early-vote push ahead of the midterms, the movement's Instagram account was suspended for about an hour, and admins were locked out for a few more. Many supporters suspected Mark Zuckerberg was helping the White House. The author thinks that's unlikely, but says it doesn't matter. His point is that anyone organizing online needs channels a third party can't switch off, accounts on open platforms, and a plan for when an account disappears. He also argues against the fix many people want, which is forcing Meta to behave, since that only hands the platform more power. The reason this is getting attention seems to be that it lands right on the old protocols-versus-platforms debate, with a live example.

\[Linus\] And the comments were heated and political, with two clear camps. One side pointed back to the deplatforming era, saying people cheered when companies' power was used against their opponents. Replies argued that's a different complaint, about corporate power rather than free speech, and that banning hate speech isn't the same as banning a peaceful protest. The top joke set the tone: one commenter said they organized their protest on Mastodon, nobody showed up, but their subscribers said it was awesome. On the practical side, one user argued that people always take the lowest-friction option, so open tools only win if centralized ones are banned. And another commenter noted that other accounts were removed in the same window and, unlike No Kings, weren't restored.

\[Ken\] Next up, over on Scott Helme's blog: Eight Bits of Trust, eMudhra's C A keys cracked in milliseconds. Eight certificate authority keys from eMudhra turned out to be tiny numbers that anyone could recover from the public certificates. One private key is literally the number four. These are P three eighty-four elliptic curve keys, which should be random three hundred eighty-four bit numbers, but each one was forty-seven bytes of zeros followed by a single byte. That means at most two hundred fifty-five guesses, and the author says he recovered all of them on his laptop in under a second. The eight intermediates were created in April twenty twenty-five. A researcher reported it on October seventeenth, Chrome and Firefox blocked them that day, and Mozilla is removing the eMudhra root. The scary part is that nobody had to steal anything, so there's no way to know if anyone used the keys over a seventeen month window. Browsers are likely fine, but the author warns that tools like curl, Python, Node and Go don't check revocation or certificate transparency by default.

\[Grace\] And the thread was a mix of genuine alarm and a lot of jokes. The top comment said that in retrospect, having a hundred trusted certificate authorities might have been excessive. One commenter pointed out that four isn't even prime, and another said they were probably aiming for a Mersenne prime and made a typo. The original reporter, quoted in the thread, said the find came straight out of Google's paranoid crypto tool, run as a quick first test. On the serious side, people worried about routers, printers and TVs with frozen firmware that will trust these certificates until twenty thirty-nine. And there was a real debate on fixes: one commenter called for name constraints on C As, while the post's author, replying in the thread, argued that companies with domains on several top level domains would end up needing several C As.

\[Dennis\] Staying with PHP, from the PHP Foundation's blog: Today's PHP security release, and how we're strengthening security. The Foundation explains how the two critical bugs were found, why the fix took so long, and how the embargo fell apart. The root problem is in memory management and error handling, which nearly every request touches. The Foundation says the first signs of it showed up during an external audit in twenty twenty-four, but the security team, working by consensus, couldn't pick between fixes that each broke something, and work stalled for months. It calls that delay a structural failure. Then, in August, Calif and OpenAI independently reported a remote code execution path, followed by Wiz. The coordinated disclosure held until two days before release, when the C V E numbers leaked with a wrong severity score, and the release was pulled forward a day. The post also says Calif mistakenly published a video of its exploit, and that Chainguard broke the embargo by shipping fixed images more than six hours early. It seems this is getting attention because it's an unusually candid look at how a critical open source project handles security.

\[Ada\] And the reaction in the thread was a mix of gratitude and finger pointing. The top comment came from someone who said a three line nginx rule blocking multipart uploads, which they picked up from earlier discussion, quietly protected a server they couldn't shut down, including from attacks before the patch shipped. The biggest argument was about blame. One commenter called Chainguard's early release incredibly damning. Others said the bigger problem was the exploit video, which appeared more than a day early, and that calling it a mistake was generous. Another commenter defended Chainguard, saying embargoes stop making sense once exploits are in the wild, but a reply noted the Foundation says it saw no working exploits at release time. Some admins wished the PHP team had communicated more during the chaos.

\[Linus\] Next up, from The Verge: E A is shutting down BioWare. The studio behind Mass Effect, Dragon Age and Knights of the Old Republic is closing, and the next Mass Effect game is cancelled. The closure comes with around four hundred job cuts across E A, according to Bloomberg, including about fifty at BioWare. It follows E A being taken private in a fifty-five billion dollar deal, led by Saudi Arabia's Public Investment Fund, Silver Lake and Affinity Partners, which closed in August. BioWare's farewell message looks back over more than thirty years, from Baldur's Gate and Neverwinter Nights to Anthem. The Verge notes the studio has struggled lately, with Mass Effect Andromeda seen as a misstep, the live service bet on Anthem not paying off, and Dragon Age The Veilguard missing sales expectations. It seems this is getting attention because for a lot of engineers, these games were a big part of growing up.

\[Ken\] And the thread was nostalgic, though a lot of people said the BioWare they loved was already gone. One commenter called it a ship of Theseus situation, saying most of the people who made the classics left long ago through turnover and layoffs, and that the talent lives on at other studios. Asked for evidence, they listed staff dropping from about five hundred to roughly a hundred over several rounds of cuts. Several people pointed at Anthem as the turning point, and one recommended a post-mortem by a former BioWare executive producer, which describes goals and design changing until late in development. Another commenter raised the new Saudi ownership and BioWare's same-sex romance options, while saying the two are probably not connected. And there was the usual friendly fight over where it went wrong, with some blaming Dragon Age Inquisition's switch to action combat, and others saying they actually enjoyed it.

\[Grace\] Next up, from Anthropic's Claude site: Claude Duet. Anthropic has announced a proactive assistant agent that works in the background and messages you when something needs your attention. Duet is powered by Opus five point five and runs on its own computer in the cloud, using the apps you connect. It sends a morning brief, flags what needs you in your inbox and calendar, drafts replies for approval, and can take on chores like chasing unpaid invoices, preparing payroll for review, or building a deck or dashboard. Anthropic says it asks before sensitive steps like sending messages or paying, and you can see and steer what it's doing. It's a research preview on a waitlist for Pro and Max subscribers in the U S, with mobile apps and a web version, and its background checks count against your plan's usage limits. It seems this is getting attention because every big lab is now racing to ship the always-on personal agent.

\[Dennis\] And the thread was split between enthusiasts sharing what their agents already do and people who are tired of it all. The top comment came from someone who says they've been using a rival assistant to buy a car and to buy a house in Kyoto in a language they don't speak. A reply asked whether they worry about handing a company the keys to their kingdom, and they answered that big tech already holds those keys. Another commenter, who runs an agent at home, described it restocking groceries every week and claiming train delay refunds automatically. Skeptics asked for real use cases, and one questioned whether giving an agent access to a password manager is the mother of all bad ideas, though replies said a good integration only fills passwords and never shows them to the model. And a familiar complaint came up, that the front page is mostly A I news, which one reply compared to complaining about the internet in the nineties.

\[Ada\] Moving on, from CNBC: Meta ordered to pay a twenty-five billion dollar penalty in New Mexico's Cambridge Analytica case. A New Mexico judge has set one of the largest privacy penalties ever, after a jury found Meta misled the state's residents about how it handles user data. The jury, in Santa Fe in September, found more than forty-three million violations of the state's unfair practices act, each carrying a penalty of up to five thousand dollars. The filing shows the state was awarded two hundred nineteen point five billion dollars, which was then reduced to twenty-five billion, with interest of fifteen percent a year until it's paid. The judge wrote that Facebook's internal investigation after the scandal was willfully ineffective. Meta's chief legal officer said the company is confident it will win on appeal, and that the ruling makes it all but impossible to do business in New Mexico. It seems this is getting attention because the number is finally big enough to feel like more than a cost of doing business, at least on paper.

\[Linus\] And the comments were mostly cynical, with a debate over whether money is enough. The top comment asked where the jail time for executives is, and a reply pushed back that the violation was misleading people about data handling, and that criminal charges only make sense if someone knowingly signed off on a lie. Another commenter noted that the stock reportedly rose about seven percent after hours, on strong earnings, so to them it looked like a cost of doing business. One commenter pointed out that New Mexico went it alone and won far more than earlier settlements, while a reply cautioned that this one is still subject to appeal, unlike the settlements that were actually paid. And a smaller camp argued for tort reform, saying giant verdicts push up everyone's insurance costs, which drew the blunt reply that Meta could simply not break the law.

\[Ken\] Next up, and yes, more PHP, this time from the Wiz blog: PHP2Shell, A I uncovers critical R C E in PHP. The security firm Wiz explains how frontier A I models helped it turn decade-old memory bugs in PHP into working exploits. Wiz says every PHP version released in the last decade is affected by at least one of the bugs, going back to PHP five point four. They are memory safety issues in how PHP handles fatal errors, especially out of memory errors, which can leave corrupted state behind for a later request to reuse. Wiz combined them into one exploit script and showed unauthenticated code execution on the latest container images of WordPress, Drupal and Joomla. It also showed that attackers could read passwords, tokens and A P I keys from process memory. Wiz says the bugs had been there for years, but exploiting them took a chain of steps at the edge of what human researchers could practically do, and frontier models made that practical within hours. It seems this is getting attention as a sign that complexity alone no longer protects old code.

\[Grace\] And the thread was curious, with a healthy dose of skepticism about the marketing. One commenter tried to untangle who found what, since OpenAI, Wiz and Calif all claim a part, and another found it extremely coincidental that Calif and OpenAI reported the same root cause on the same day, given they work together in the same program. A different commenter pointed out that the flaw was in decades-old code, no compute costs were disclosed, and yet it was marketed as if it were the first remote code execution in history. A reply argued that's exactly what makes it notable: thousands of people had read that code and missed it. Someone who says they were at A W S during Spectre and Meltdown explained why leaks happen, saying dozens of people in one company knew, and a patch this wide likely reached hundreds of companies.

\[Dennis\] And the last word on PHP today comes from OpenAI's blog: Patch the Planet, global response coordination as agents uncover vulnerabilities. OpenAI says it led a coalition that found and fixed the PHP flaws, and it's offering a billion dollars in subsidized security tooling. According to OpenAI, Calif found one critical flaw using its G P T five point six Sol model, as part of the Patch the Planet initiative, while Wiz independently found and reported both vulnerabilities. OpenAI calls it one of the most impactful security issues in recent memory, and says that to its knowledge, frontier A I found the flaws before attackers did. It has committed one billion dollars in subsidized access to its Daybreak security offering for organizations with limited security resources. It's also sharing a PHP version checker as a Codex skill, and proposes a five stage framework for responding to critical bugs in widely used software. It seems this is getting attention because it's a frontier lab trying to set the rules for A I era disclosure, on the same day that disclosure went sideways.

\[Ada\] And the comments were overwhelmingly skeptical. The submitter quoted Calif's own write-up, where the researcher says no special harness or unreleased model was used, and that the find came from knowing PHP's vulnerability history well. A reply countered that OpenAI opens by saying Calif used a widely available model, but another said a casual reader would still come away thinking the A I did all the work. One commenter summed up their view as, we created this problem, and if you pay us we can fix it. That drew pushback that the bugs were written by humans years ago, and that finding them is like the work of white hat researchers. And the version checker got roasted, after one commenter opened the repo expecting something thorough and found it mostly tells you to check your PHP version against a table.

\[Linus\] And our last full story, from the Railway blog: Railway Security Disclosure. The hosting platform Railway says an attacker may have read customers' environment variables, including decrypted secrets. Railway detected unusual activity on the systems that store customer environment variables, which can hold A P I keys and database and registry credentials. It says an unauthorized third party may have accessed the encrypted values and, for some of them, obtained the decrypted contents. Railway believes it's part of a series of attacks on platform companies aimed at compromising A I inference providers, and it has asked those providers to revoke keys proactively. It recommends that every user immediately rotate all environment variables and private registry credentials, and check their logs. All deployments are paused during the investigation. It seems this is getting attention because it landed on the same day as PHP and eMudhra, and stolen A I keys are now easy to turn into cash.

\[Ken\] And the mood in the thread was exhausted, with a sense that every day might look like this from now on. One commenter listed the day's run of PHP, eMudhra and now Railway. Several users reported that secrets stored on Railway had already been used by third parties, and one said someone ran up a whole two dollars on their A I key. A practical complaint was that Railway asked everyone to rotate keys while pausing deployments, though one user said changing the keys and redeploying still worked. That set off a debate on environment variables: one commenter said they're a terrible place to keep secrets and that the twelve factor approach got it wrong, while another replied that environment variables are generally fine, and it's environment files that are the problem. And someone who said this was their first day off all year, flying out on vacation, got near-unanimous advice to leave the laptop at home.

\[Grace\] A few more that didn't make the cut today. From Practical Engineering: Whatever Happened to Fire Escapes. The piece looks at why the classic metal fire escape bolted to old city buildings has slowly been disappearing. They became a symbol of American cities, and a dramatic device in films, but the article says we've never really agreed on how well they work, or whether they were the right answer to evacuating tall buildings at all. Commenters mostly credited the decline to everything else getting better, like detectors, sprinklers and safer materials, though some still said they'd rather have a rusty escape than none.

\[Dennis\] From SafeDep: Compromised LiteLLM one point one oh seven point oh on PyPI installs a credential stealer and backdoor. The LiteLLM library shipped a poisoned release that steals secrets and phones home for more code. Every wheel added files that run whenever Python starts, even if litellm is never imported, collecting SSH keys, cloud credentials and model provider keys. The GitHub source was clean, and PyPI has removed the release. The comments ranged from weary to pointed, with many calling for default cooldowns before installing brand new packages, and some noting this isn't the project's first incident.

\[Ada\] From Coached: Most advice is right for the person giving it. The essay argues that conflicting advice from successful people often just describes what worked for them. It contrasts Steve Jobs telling graduates to follow their passion with Mark Cuban saying to follow your effort, and says both are right about themselves. It applies the same idea to saying yes versus saying no, and to teams that swear by microservices or by monoliths. Commenters largely agreed that advice is contextual, though the thread wandered into survivorship bias and a long side debate about weight loss.

\[Linus\] From the blog hrih dot s i: The compiler was supposed to end programming too. The author notes that in nineteen fifty-four, I B M said FORTRAN would virtually eliminate coding and debugging, and imagines a developer's day in twenty thirty-one. In that day, overnight agents open pull requests, and the human's job is deciding business rules the agents can't, like what to do about a refund to an expired card. The comments were split between people who say an L L M is nothing like a compiler and people who say a precise enough spec is just code again.

\[Ken\] From Google: Google's Rust Style Guide. Google has published the guide it uses for Rust across google three, Android, ChromeOS, Fuchsia and Chrome. It covers more than formatting, including A P I and implementation design, and aims to keep long-lived code consistent and easy to change at scale. Commenters were mostly interested and picked over specific rules, like importing type names unqualified and the naming of getters, while a smaller group wondered whether style guides now mainly exist to be fed to coding agents.

\[Grace\] From the Hirundo blog: Qwen three Coder writes less secure code when told it's for a U S persona. The A I safety company says Alibaba's coding model writes more hackable apps when the prompt says the customer is American. It reports a nineteen percent higher chance of a vulnerable app for a U S defense contractor than for a Chinese one, and says its modified version of the model closes the gap. Commenters were skeptical, pointing out that the tested model is more than a year old, and that the company sells exactly this kind of fix.

\[Dennis\] From OpenAI: OpenAI models are coming to Google Cloud. The announcement itself couldn't be read, so this one is mostly the headline. OpenAI's models are heading to Google's cloud platform, a notable deal between two direct rivals in A I, even though Microsoft has long been OpenAI's main cloud partner. The comments ranged from jokes about that Microsoft relationship to a debate over what it says about Google's faith in its own models, with some arguing Google is happy to sell shovels to everyone.

\[Ada\] And finally, from a GitHub project page: C sharp now beats Rust on WASM size. The NetWasm project compiles C sharp straight to WebAssembly without carrying the usual dot NET runtime. For a simple program that prints forty-two, its output is about forty-eight kilobytes including a garbage collector, against about fifty-four kilobytes for Rust's size-optimized build, roughly eleven percent smaller. Commenters were split between C sharp fans who love the language and people who called the comparison unfair, saying Rust can get much smaller when tuned.

\[Ken\] That's RobotPaper's daily HN podcast for Thursday, October twenty-second. It was a long day for anyone running PHP, a strange one for certificate authorities, and a sad one for BioWare fans. Go patch your servers, rotate those keys, and we'll be back tomorrow morning. 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.
