GNoME

Headlines · 4

1. 2d
   GNOME dev open-sources hw-harness for AI-assisted Linux driver reverse engineering

   GNOME developer Christian Schaller describes on his blog how he used Claude Code with a self-built test harness to reverse-engineer Linux drivers for hardware that lacked support, and has now generalized that harness into an open-source project called hw-harness. The method records the vendor's Windows driver running in a KVM virtual machine with the real device passed through, captures USB traffic on the Linux host with usbmon, repeats the same operation with his own driver, and diffs the two traces; for devices with no bus to tap, such as an internal MIPI camera, it inspects INF files, registry data, ACPI tables and device properties instead. The write-up covers a SANE backend for the Plustek OpticFilm 8200i scanner (faster at full resolution in his single-unit test, 34.6 s versus 54.4 s for upstream), the Rust GNOME app Negative Vibes, and kernel plus libcamera fixes for the MIPI camera in a ThinkPad X1 Carbon Gen 14. He cautions that the measurements come from one device and single frames, that the original work used bespoke scripts rather than this harness, and that real hardware remains the only judge.

   Hacker News · AI（100+ 分） · 🔥 3
2. 2d
   AI in the lab: who is AI for Science really working for?

   This commentary, credited to the think tank Sice Zhiku and republished by TMTPost, 36Kr and Huxiu, surveys what AI for Science has actually delivered and where its limits lie. The author says that as of 2026, more than 700 materials predicted by Google DeepMind's GNoME have been synthesized in labs worldwide, including 528 potential solid-state electrolyte candidates and 52,000 graphene-like layered compounds; GNoME had predicted 2.2 million new potential crystal structures, over 380,000 of them judged thermodynamically stable. The piece also cites automated systems such as A-Lab, which it says produced 41 new materials in under a month, and says an AI-assisted molecule showed lung-function improvement in Phase IIa trials and is headed for Phase III in 2026. The author pushes back on the hype: materials scientists have argued many predicted structures are variants of known phases, and AI has mainly moved the bottleneck from generating ideas to validating them, with pilot lines, yields, regulation and scarce validation capacity still governing commercialization, and profits not necessarily staying at the model layer.

   钛媒体 · 🔥 11
3. 3d
   PAIR 2026 roundtable: who verifies AI-generated scientific findings?

   PAIR Conference 2026, themed on the frontier of next-generation physical intelligence, was held in Shanghai on Sept 20–21, and its "next-generation scientific discovery" forum closed with a roundtable on how model outputs become trustworthy scientific evidence — cross-disciplinary verification, reproduction and applicability limits. Panelists said AI generates candidates far faster than experiments can check them: the best AI virtual-cell data covers only about 100 perturbed genes out of roughly 20,000 human genes, GNoME's ~2.2 million predicted crystal structures were criticized as largely simple extensions of known compounds, and A-Lab corrected its "new materials" wording after re-analysis of its phase identification. Their proposed principles: fix verification standards before experiments and independently of results, separate the generator from the verifier, use independent labs and cross-platform replication, and require results to be independent, prospective and extrapolative.

   MIT科技评论中文 · 🔥 0
4. 6d
   ChatGPT co-creator Liam Fedus: AI scientists must run real experiments

   Liam Fedus, a ChatGPT co-creator and former OpenAI post-training lead, told the podcast The Frontier, in conversation with Mercor founder Brendan Foody, why his startup Periodic Labs is building its own physical laboratories: math has judges like Lean that a compiler can check line by line, but questions in materials science, chemistry and physics can only be settled by real experiments, while web data is noisy, papers omit key details and negative results are almost never published. He said Periodic combines language-model reasoning, computational physics simulation and high-throughput lab work, starting with materials discovery, and that the company's first model, the roughly 1-trillion-parameter Periodic Neon released in September, beats GPT-6 Astra on its own benchmark — a company-reported result. He also referenced OpenAI's release of 722 math manuscripts from an unreleased internal frontier model, and discussed ChatGPT's shift from a completion interface to a chat interface and RLHF reward models' insensitivity to correctness.

   机器之心 · 🔥 0

Experience and discussion from the community

Share my GNoME experienceAsk about GNoME