Skip to main content

Posts

Showing posts from November, 2026

StateBench: Can AI Follow Every Change in a Long File?

Image
StateBench: Can AI Follow Every Change in a Long File? OPEN SOURCE FACTORY · NOVEMBER 6, 2026 Watch the English motion graphic Follow the Change — a 42-second English conceptual animation. Original 3D motion, newly composed music, and sound effects: Open Source Factory. The ledger scene is our design proposal; this is not footage of a model test. A model finds a name buried in a long file. Then it loses track of who holds the key after a few handovers. Both questions use the same document, but the second answer has to be computed from a changing state. That is the gap I wanted to understand in StateBench. Ai2 introduced the benchmark on November 3, following a November 2 preprint. Its useful idea is to separate different kinds of updates instead of treating every long-context question as the same test. A bigger context window gives the model more text to work with; following all the changes remains a separate capability. Ai2’s announcement and the research paper make that distincti...

Kimi K3.1 Pro: 104B Active, 1.56 TB to Download

Image
Kimi K3.1 Pro: 104B Active, 1.56 TB to Download Open Source Factory · 2026-11-05 Watch the English video English 3D explainer. Motion, newly composed original music and effects: Open Source Factory. Conceptual animation, not model-execution footage. The number I checked first was the download size. Kimi K3.1 Pro’s official repository lists 96 weight shards totaling about 1.56 TB . The weights are publicly accessible, and there is a license to read. This has moved beyond a promised release date. That is a meaningful change for anyone who wants to deploy or adapt the model. It also needs some scale: “104B active parameters” does not describe a workstation-sized download. I checked the official files, configuration, license and serving recipes on November 5, 2026, Korea time . I have not run the model. Image: Moonshot AI, from the official checkpoint repository. Brand artwork, not a generated model sample. What can you actually download? The destination is moonshotai/Kimi-K3.1-Pro o...

Bambu Lab H3C: What the $2,499 Combo Really Includes

Image
Bambu Lab H3C: What the $2,499 Combo Really Includes Open Source Factory · November 4, 2026 · Source-based launch analysis Watch the English motion graphic English 3D overview. Motion graphics and newly composed music, “Loaded Paths”: Open Source Factory. Conceptual animation, not hardware footage or a printer test. Prices and features follow the official announcement. I went from the H3C’s nine-tool headline straight to its box-contents table. That table changes how I would plan a first print: the standard Combo contains eight tools for 1.75 mm filament and one for 2.85 mm flexible filament. It is a useful allocation, especially for rigid-and-soft parts. A job using nine ordinary PLA paths needs another compatible 1.75 mm PrintTool. Bambu Lab launched the H3C on November 3, 2026. The Combo with AMS 3H is $2,499 in the US before tax ; the official Korean launch announcement lists ₩3,999,000 including tax and shipping . This article follows our earlier P3C/H3C leak and certification r...

Microduck: What You Can Build With a $399 Robot Duck

Image
Microduck: What You Can Build With a $399 Robot Duck Open Source Factory · Research report · November 2, 2026 Your browser does not support this video. A Small Step Into Robotics — original conceptual 3D animation, music and sound design by Open Source Factory. Illustrative geometry; this is not a recording of a robot or a reproduced experiment. English, 40 seconds. The beak is a gripper. The waddle comes from a learned control policy. Microduck makes those two facts look like a character introduction. That is a good way to get me interested in a robot. Pollen Robotics, Hugging Face’s robotics team, has built a small biped around a useful idea: make experimenting with physical movement approachable. You can play with its existing behaviors, then work on your own in simulation. The appeal is the path between those two experiences. The team’s introduction explains that intention. My take: Microduck is interesting as a movement-learning platform. I would judge it by how easily a new...

Mage-VL-Cowork: A 4B Agent That Chooses Clicks or Code

Image
Mage-VL-Cowork: A 4B Agent That Chooses Clicks or Code Open Source Factory · 2 November 2026 · Release analysis Watch the English motion graphic English 3D motion graphic: Open Source Factory. Newly composed original music and sound design: “Click. Code. Verify.”, 128 BPM. Concept animation, not a model run. Data sources: Microsoft Mage Team; checked 2 November 2026. A desktop agent can spend twenty clicks fixing a spreadsheet, or edit the file with a short script. Mage-VL-Cowork is interesting because it learns to choose the route. Microsoft’s 4B-class model combines screenshot-based interaction with terminal execution. The useful question is whether that choice produces a file you can actually reopen and use. My answer from the release is cautiously positive: the design makes sense, and the team supplies concrete desktop demonstrations. But the headline 66% is a reported benchmark result under a specific setup. It is not a promise that two out of three tasks on your PC will work. ...