Understand

The other advances

1. AvailableOct 12, 2026Muse Wonder and Muse Wonder Full

   Meta introduces Muse Wonder and Muse Wonder Full, its most capable models, available via the Meta Model API and Muse Code (Wonder) and to Muse Code subscribers (Full).

   What “Available” means: Usable by the public or by customers, source read.

   See in the feedWritten by an AI · not reviewed

2. Not evaluatedOct 11, 2026Conditional integer-multiplication bound

   On 11 October 2026, a contributor to the integer-mult-bounds project proposed a conditionally improved theoretical bound for multiplying very large integers.

   Analogy

   Like a journey time worked out from a set of assumptions: the conclusion depends on those assumptions, and the calculation still needs checking.

   What it changes

   κ appears in a theoretical bound on computing time, of the form n (log n)^(1-κ) for n-bit numbers (binary digits): the larger κ, the lower this bound as n grows, with no promise of a practical speed-up. If the bound held, it would improve, under the same assumptions, on the one in family #109 of the OpenAI release, where κ = 2^-182, a far smaller number.

   What remains unclear

   The result only stands if the framework of OpenAI family #109 and the interfaces listed by the PR (a change request on GitHub) are sound. No published independent evaluation is recorded and the request is not merged. No AI use is declared in this PR: the entry is tracked because it builds on that OpenAI result.

   What “Not evaluated” means: No third-party evaluation has been found. This does not mean the result is wrong, nor that it is right: as far as we know, nobody has published a check.

   Read the full entryWritten by an AI · reviewed by a human

3. Not evaluatedOct 9, 2026TheoPhysEval: five AI agents on 50 unpublished theoretical physics problems

   Physicists post on arXiv 50 unpublished theoretical physics problems with numerical answers and test five AI agents on them. The best, GPT-6-Astra with Codex, gets 49 of 50 problems right with web access, 46 without. Preprint not peer-reviewed; one run per agent, no timed humans.

   What “Not evaluated” means: No third-party evaluation has been found. This does not mean the result is wrong, nor that it is right: as far as we know, nobody has published a check.

   See in the feedWritten by an AI · not reviewed

4. AvailableOct 8, 2026Claude Dashboards and Claude Motion in beta

   Claude can build a live dashboard connected to a data platform and explainer animations exportable as MP4.

   What “Available” means: Usable by the public or by customers, source read.

   See in the feedWritten by an AI · not reviewed

5. Demo onlyOct 8, 2026Midjourney: test of a "Thinking Mode" for image generation

   Midjourney is testing on its Alpha site a mode that reruns an image with additional reasoning.

   What “Demo only” means: Public demonstration (video, supervised trial), no open access.

   See in the feedWritten by an AI · not reviewed

6. AvailableOct 7, 2026ChatGPT: "Intelligent UI" answers as interactive interfaces

   With GPT-6, ChatGPT combines text, visuals and interactive elements (comparisons, calculators, games) and picks the format itself.

   What “Available” means: Usable by the public or by customers, source read.

   See in the feedWritten by an AI · not reviewed

7. AvailableOct 7, 2026Claude Haiku 5.5

   Anthropic releases Claude Haiku 5.5, its small model, the first Haiku with an adjustable effort setting and an average cost about 75% lower than Haiku 4.5.

   What “Available” means: Usable by the public or by customers, source read.

   See in the feedWritten by an AI · not reviewed

8. AvailableOct 7, 2026d1-3B and d1-omni-600M, open-weight decision models

   Liquid AI releases two open decision models that answer in a single forward pass without generating tokens, aimed at NVIDIA edge hardware.

   What “Available” means: Usable by the public or by customers, source read.

   See in the feedWritten by an AI · not reviewed

9. WithdrawnOct 7, 2026Three manuscripts withdrawn after a sign error

   On 7 October 2026, OpenAI withdrew three manuscripts from its mathematics release, on Weil classes, Kuga-Satake K3 and the rational Hodge conjecture, after a sign error. Status: withdrawn.

   Analogy

   It is like a book pulled from sale after a calculation error is found in a central chapter: the publisher sets it aside but keeps a record.

   What it changes

   According to OpenAI's change log, a sign error invalidates a cancellation argument, so the three results are no longer asserted. The same log also mentions 14 revised manuscripts and 6 added formalisations. OpenAI says it will keep updating the repository with any errata it notices.

   What remains unclear

   The person or tool that spotted the error is not identified in the sources read. The withdrawal says nothing about the other manuscripts in the release, whose status remains under review. Retraction Watch reports that roughly half of the results were released unconfirmed.

   What “Withdrawn” means: The result was withdrawn or abandoned by its author, or classified as incorrect by the reference source. It stays in the registry, with the date and the reason, so that the error remains visible.

   Read the full entryWritten by an AI · not reviewed