TL;DR
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
Astral Codex Ten published an October 2026 links roundup spanning AI-generated art and translation, a bot’s win in a forecasting competition, and commentary on marriage, suburbs and other subjects. The author warns that individual links have not all been independently verified and may contain errors or misleading claims.
Astral Codex Ten has published its October 2026 links roundup, a collection of outside articles, online claims and the author’s commentary that includes a reported AI bot victory in the Metaculus Cup and examples of AI-assisted art and literary translation. The author says the links have not all been independently verified, so the post is a curated collection rather than a vetted account of every claim it cites.
The roundup says a bot named Laertes placed first in the Metaculus Cup forecasting competition in September, marking the first time a bot won the competition, according to the post. It attributes details about the system to The Economist: Laertes was developed by Jeffrey Liang, who reportedly spent fewer than 150 hours and a few thousand dollars on computing power and data. The post also names systems from Mantic, FutureSearch and Preseen among other high performers, and says ACX reader Ben Shindel placed first among human competitors.
The author connects the result to a previous prediction of AI-human forecasting parity by the end of 2026. That is the author’s forecast, not a finding established by the competition result alone. The post also notes a separate development: AI forecasting startup Mantic reportedly raised $25 million in seed funding. The source material does not provide further terms or independently confirm the financing.
Other entries range from a historical note about Jeremy Bentham’s cookbook to criticism of an argument about marriage and reported happiness. The author says a post claiming that only 18% of marriages are non-miserable appears to classify “pretty happy” marriages as miserable, and says their reading of the cited General Social Survey indicates that 61% of respondents describe their marriages as “very happy.” The author reports checking the marriage-category point but says other alleged errors have not been confirmed.
AI Systems Enter Forecasting Contests
The reported Metaculus Cup result offers a concrete competition outcome in a debate about whether AI can make forecasts at a level comparable to people. It is a single contest result, not by itself proof of general forecasting parity: the roundup does not provide the contest’s full scoring details, the margin between competitors or evidence about how the systems perform on other forecasting tasks.
The account also draws attention to the reported development effort behind Laertes. If the figures attributed to The Economist are accurate, the system was built with limited time and computing expenditure compared with the scale often associated with major AI development. But the roundup does not describe the bot’s methods in enough detail to assess how much of its performance came from its design, data, tools or contest conditions.
For readers, the post’s wider relevance is partly methodological: it gathers claims and commentary from across different subjects while openly warning that some items may be wrong. Its discussion of the marriage statistics illustrates why checking definitions and source data matters when a striking number is used to support a broad argument.
Popular right now
TCL T7 Series 85-Inch 4K 144Hz QLED Smart Google TV
Dyson V8 Cyclone Cordless Vacuum
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
A Curated Monthly Link Collection
Astral Codex Ten’s post is organized as a numbered set of links with short reactions, rather than as one report about a single event. The October collection moves among AI, culture, social issues and history. The author’s notes sometimes distinguish personal opinion from reported information—for example, describing an AI-generated interpretation of Hieronymus Bosch as a favorite while acknowledging its reliance on a human artist’s work.
Several entries concern language and creative work. The author says they used a newer AI model to produce an English version of the opening of Romance of the Three Kingdoms in rhyming pentameter, while acknowledging they do not speak Chinese. The post also links to an English rendering of the first 100 lines of the Odyssey in heroic hexameter. These are examples offered by the author, not a formal assessment of translation quality.
The roundup also discusses an online claim that an older version of Claude might continue text resembling an internal company email. The author says the possibility that this would expose actual internal emails is probably false and cautions that this is not how training works. That assessment is commentary in the post; the supplied material does not document a technical investigation of the claim.
“I haven’t independently verified each link.”
— Astral Codex Ten author
Verification Limits and Open Questions
The author explicitly says the links have not each been independently verified and estimates that commenters find roughly two or three wrong or misleading links per roundup, on average. The source text does not identify which other entries may later be corrected, and it warns that corrections may not be in place by the time a reader sees the post.
For the forecasting result, the supplied material does not include contest rules, detailed scores, the number of bots and human participants, or a comparison across different forecasting periods. The claims about Mantic’s funding and other systems’ performance are also presented without supporting documents in the supplied excerpt. The excerpt ends mid-entry, so it may not represent the complete set of October links.
The discussion of Claude’s text completion is framed as a claim about an older model. The author rejects the most alarming interpretation as probably false, but the excerpt gives no test results or independent technical analysis that would settle how the behavior arose.
Corrections and Further Evidence
The author says they correct errors as they notice them and will highlight important corrections later. Readers should check the original post for updates and consult the linked sources directly when relying on a particular statistic, funding report or competition result.
The next evidence relevant to the forecasting claim would include complete Metaculus Cup results and information about how Laertes and the other systems were evaluated. Those details could help show whether the reported win reflects performance that carries over beyond one competition. The roundup itself does not announce a follow-up test or provide a date for additional results.
Key Questions
What is the October 2026 links post?
It is an Astral Codex Ten roundup of links and short commentary covering AI, culture, social issues and other topics. It is not a single, independently reported news investigation.
What AI forecasting result does it report?
The post says that Laertes won the Metaculus Cup in September, the first bot to do so. It attributes details about the system’s developer and development effort to The Economist.
Does the competition prove AI and human forecasters are equal?
No. The author connects the result to a prediction of AI-human forecasting parity by the end of 2026, but the supplied material gives no full scores or broader evaluation. A single reported competition win does not establish general parity.
Are all the links and claims verified?
No. The author says they have not independently verified every link and that commenters often find errors or misleading material. The post says corrections are made as they are identified.
Source: astralcodexten.com
