---
title: "Oct 27: Anthropic's S-1 filing, and more"
show: "RobotPaper's daily HN podcast"
canonical: "https://podcasts.robotpaper.ai/hn/ep/2026-10-27/"
date: "2026-10-27T03:21:59+00:00"
duration: "15:52"
duration_seconds: 952
audio: "https://podcasts.robotpaper.ai/hn/ep/2026-10-27/2026-10-27-hn.mp3"
chapters: "https://podcasts.robotpaper.ai/hn/ep/2026-10-27/chapters.json"
transcript: "https://podcasts.robotpaper.ai/hn/ep/2026-10-27.txt"
feed: "https://podcasts.robotpaper.ai/hn/feed.xml"
language: "en"
author: "Roy Osherove"
ai_generated_voices: true
---

# Oct 27: Anthropic's S-1 filing, and more

RobotPaper's daily HN podcast · Oct 27, 2026 · 16 min

- Listen: [MP3, 15 MB](https://podcasts.robotpaper.ai/hn/ep/2026-10-27/2026-10-27-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/2026-10-27/
- Chapters (JSON): https://podcasts.robotpaper.ai/hn/ep/2026-10-27/chapters.json
- Transcript (text): https://podcasts.robotpaper.ai/hn/ep/2026-10-27.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
- `00:00:38` Anthropic filing
- `00:03:13` Microsoft buyouts
- `00:05:47` iPad mini update
- `00:08:10` GPT-6.1 Terra
- `00:10:21` Open models on-prem
- `00:12:42` Solar passes nuclear
- `00:13:22` Scientific discoveries
- `00:14:04` Switzerland Minecraft
- `00:14:42` Peter Thiel profile
- `00:15:17` Outro

## Show notes

- Anthropic S-1
- Microsoft offers mass buyouts for roughly a quarter of its workforce
- New iPad mini
- GPT-6.1 Terra
- IBM CEO says open weight models are now good and cheap on-prem
- Solar power is becoming bigger than nuclear
- Three new discoveries in mathematics, computer science, and physics
- Switzerland. Block by block. At 1:1 scale
- Peter Thiel, the tech oligarch who believes in the Antichrist

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, October twenty-seventh, 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, Anthropic's filing prompts debate over demand; Microsoft offers voluntary exits.

\[Ada\] I'm Ada, and Apple updates the iPad mini; OpenAI introduces Terra.

\[Grace\] 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 the Securities and Exchange Commission, Anthropic's S-one filing, filed October twenty-sixth. The registration statement describes Anthropic as a public benefit corporation. A Long-Term Benefit Trust of five trustees elects a majority of the board, and a single founder share, held through an LLC of the seven co-founders, carries fifty point one percent of the general voting power. The company says that makes it a controlled company under Nasdaq rules. On customers, it reports more than eleven thousand nine hundred accounts with at least one hundred thousand dollars in trailing revenue, and about two hundred fifty above ten million. It describes compute as a single pool across AWS Trainium, Google TPUs, Nvidia and AMD chips, with purchase commitments that run as far as twenty forty-five. The Hacker News discussion centers on those commitments. Commenters engaged because the filing put actual revenue and cost figures on the table for the first time.

\[Grace\] And the thread is divided between confidence in demand and concern about commitments made in a rapidly changing market. Trjordan argues that Anthropic has to secure capacity ahead of demand or risk losing business through poor availability and expensive leases arranged at short notice. Jacquesm counters that today's prices should be treated as a ceiling, with competition putting future margins under pressure. Maxatar challenges the time frame used in that debate, emphasizing that the commitments are multiyear rather than annual. Dyslexit adds a breakdown with different end dates for different suppliers, illustrating why treating all those obligations as one year's bill distorts the discussion. On valuation, jmuguy questions how much future growth investors would already be paying for. Thomasahle replies that an obviously attractive growth opportunity would itself be reflected in the price. The disagreement over product differentiation is equally direct. Cloverich says recent Claude use has become so compelling that switching models no longer feels worthwhile. Rbehrends reports a different experience: modern models are largely interchangeable for common software engineering tasks, while the strongest models matter more for unusually demanding work. Josephg frames the opportunity around two questions: whether the value of AI keeps expanding, and how much of that value Anthropic can retain if users can choose competing or self-hosted models.

\[Dennis\] Next, from The Seattle Times, Microsoft offers buyouts to roughly a quarter of its workforce. The company is extending voluntary exit packages while expecting only a small share of eligible employees to accept. About sixty thousand employees received offers on Monday. A spokesperson said Microsoft anticipates acceptance as low as three to five percent, based on comparable programs elsewhere. Eligible employees at vice president level and below in the United States, Canada, India and United Kingdom have until early December to decide. The standard severance package depends on location. The paper reports that these deals lack the additional incentives offered in an earlier retirement program. Participation also differs by organization, with some divisions fully included, others partly included and some excluded. Microsoft says it is controlling labor costs and adapting its workforce, while continuing to hire and move people into other roles. It seems the significance is the breadth of the offer, alongside the much smaller departure rate the company expects. It drew attention because a large share of the thread works at or near companies that send these memos.

\[Ken\] The comments mix concern about job security with disagreement over how much work large technology companies need people to do. AbbeFaria, who says they spent eight years at Microsoft across two periods, describes constant reorganizations and an unproductive final eighteen months before resigning. Frieren pushes back on the broader inference: maintenance, regulatory changes and integrating acquisitions may feel less exciting than creating a product, but established companies still depend on that work. Ecshafer focuses on the employee's dilemma, weighing a payout against the uncertainty of finding another job and the fear of later layoffs. Hypfer rejects blanket advice to accept every offer, arguing that the decision depends on individual circumstances. Christophilus recalls a former Microsoft boss warning that voluntary programs can encourage the most employable staff to leave. Elsewhere, tandr stresses that offering packages to a quarter of the workforce does not mean that everyone offered one will depart. That distinction is also central to the newspaper's reporting: the number receiving offers is much larger than the number Microsoft expects to accept.

\[Ada\] Over on Apple's newsroom, a new iPad mini. The smaller tablet gains an OLED display, an A-twenty Pro chip and water and dust resistance. Apple says the eight-point-five-inch screen fits within the previous model's footprint through thinner bezels. The enclosure is five-point-four millimeters thick, and carries an I-P sixty-seven rating. A nano-texture screen option is available to reduce glare. The front camera moves to the landscape edge, while Apple's N-one wireless chip brings Wi-Fi seven, Bluetooth six and Thread. Preorders begin October twenty-eighth, with availability from November sixth. Apple also describes performance improvements for graphics and on-device AI, although the expanded dual Neural Engine is specified for the five-hundred-and-twelve-gigabyte configuration. The announcement presents portability, display quality and work away from a desk as central uses. It seems the update is drawing attention because it changes several parts of the small tablet at once, while leaving buyers to judge which improvements matter for their own use. The thread picked it up mainly because Apple product changes tend to produce strong reactions.

\[Linus\] The reaction is mixed, with the discussion concentrating on display smoothness, physical design and price. Sheepscreek objects to Apple's non-Pro displays remaining at sixty hertz. Mtoner23 questions whether most ordinary users notice higher refresh rates, while benley describes a clear improvement in scrolling and input response on faster displays. MarkusWandel would prioritize battery life over smoother animation. Another branch examines thinness: micw argues that a projecting camera undermines the benefit of a thin enclosure and would prefer the space used for a larger battery. When paxy suggests removing the rear camera, dylan604 replies that people do use it, including themselves. On cost, jmull cites a seven-hundred-dollar starting price, and ProZsolt says the combination of price, refresh rate and starting storage has put them off upgrading. Jerojero offers a different weighting of the features, saying OLED is a substantial improvement for watching films even without a faster refresh rate.

\[Grace\] From OpenAI, GPT-six-point-one Terra. OpenAI calls it a fast, efficient model that combines detailed visual reasoning and precise computer use with clearer writing, aimed at workday tasks such as drafting proposals, interpreting technical drawings, filling out forms and editing video. Standard API prices are fifty percent lower than those of GPT-five-point-six Terra and GPT-six-point-one Sol: one dollar per million input tokens and five dollars per million output tokens. OpenAI reports that on its benchmark of error-inducing prompts, the error rate at low reasoning effort fell from fifty-three point six percent to sixteen point seven percent, and that at maximum effort the model solves fifty point three percent of Terminal-Bench four point zero tasks, up from twenty-one point five. It is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and through the API. People weighed in because many of them choose between these models every day, and the names, versions and prices keep changing under them.

\[Dennis\] The thread combines interest in lower costs with frustration about how to choose among models. Someone1234 says changes across model names and versions make the lineup confusing. Tedsanders, who says they work at OpenAI, responds that larger models remain a useful starting point for thinking about capability and price, but performance varies across tasks and newer training can change the comparison. Atombender asks for clearer guidance that maps models to activities such as writing, coding and mathematics. Silagi questions whether a middle option is useful when a larger model at lower reasoning effort, or a smaller one at higher effort, can work better. H14h answers that developers paying API prices may find a useful cost and speed tradeoff when their own evaluations show a smaller model is sufficient. Wyre argues that this release is aimed more at computer work, automation and technical writing than software development.

\[Ken\] Next up, from The Register: IBM's chief executive says open weight models are now good and cheap to run on premises. Arvind Krishna told investors on IBM's third-quarter earnings call that in twenty twenty-four, closed frontier models were plainly superior. On a broad benchmark, the top model might score seventy and an open model twenty or thirty. In late twenty twenty-five that gap shrank to about seventy against fifty. He said open weight models now cost between ten and twenty-five percent of the frontier models' token price, while being ninety percent as effective. Most tasks, such as summarizing an email or examining a contract, do not need the frontier model. Krishna said IBM is seeing booming interest in on-premises AI, and that most businesses can build something useful in two or three racks. He added that many enterprises will struggle to run them. IBM's answer is software to ease deployments, and he predicted demand for its Spyre accelerator cards. People weighed in because many of them already self-host models like these.

\[Ada\] And the mood is enthusiastic about independence, but divided over cost and the value of the strongest hosted models. Jacquesm describes an eight-GPU rig and two DGX Sparks, and says they have not used a paid model in months. Asked about the economics by otterley, jacquesm puts the rig at about twenty-five thousand dollars and says they don't think it is cheaper than a frontier lab. Independence, not saving money, is the priority. Bitexploder counters that running these models does not take twenty thousand dollars of hardware, describing a server with four V one hundred S GPUs and a MacBook Pro. Mattlondon argues that a model can be good enough for most coding without being the most capable. Roncesvalles disagrees for work with a person in the loop, asking why anyone would use less than the best model and waste their own time. Liorn says on-premises inference needs continuous monitoring and tuning, responding to trollbridge's suggestion that an AI agent could handle the setup. Awesan uses self-hosted open weight models to check cheaper models' output, avoiding dependence on one or two providers. Collinmcnulty cautions that IBM has a commercial interest in the argument.

\[Linus\] A few more that didn't make the cut for full coverage. From the BBC, solar power is becoming bigger than nuclear. Global solar generation has moved ahead of nuclear generation. The BBC's analysis of Ember data puts solar thirteen percent ahead over the twelve months through September twenty twenty-six. That comparison concerns electricity generated, rather than installed capacity. The comments are enthusiastic about solar's expansion, while questioning installation costs and how storage can support electricity supply through nights and longer periods of cloudy weather.

\[Grace\] From OpenAI, new discoveries in mathematics, computer science and physics. OpenAI says its models produced a proof of the local smoothing conjecture in every dimension, which also resolves the Kakeya conjecture, and a refutation of the Exponential Time Hypothesis that was verified in Lean. The third result concerns quantum field theory. External expert groups will shepherd the results, and the physics group is holding back the AI output. Commenters show excitement about mathematical progress and interest in involving academic researchers, alongside questions about how results in formally verifiable fields might translate into progress in experimental science.

\[Dennis\] On Manifold three D, Switzerland at one-to-one scale. A Minecraft world uses open swisstopo geographical data, with each block representing one meter. The site offers a free Jungfrau-Aletsch region download, including terrain and glaciers, while the whole country starts at fifteen Swiss francs. It asks people to test the free region with the same setup before buying the larger world. Commenters debate the value of charging for the project and its presentation, with some welcoming lower barriers to making such maps and others finding the marketing unconvincing.

\[Ken\] Finally, from El País, Peter Thiel and his belief in the Antichrist. The profile examines the technology investor's political and religious influence. It describes his support for Donald Trump and J. D. Vance, Palantir's government contracts, and his funding of Hulk Hogan's case against Gawker. Those connections place the profile at the intersection of technology, wealth and politics. The comments are sharply divided between criticism of his influence and interest in his ideas, with additional disagreement over how his religious views are being interpreted.

\[Dennis\] That's RobotPaper's daily HN podcast for Tuesday, October twenty-seventh. Today's discussions examined the economics of AI, Microsoft's voluntary exits and Apple's smaller tablet, alongside developments in energy and research. 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.
