Semantic review is behind discovery. Evidence and Pulse may be incomplete until the backlog is cleared.
2 material first-party/frontier/open-problem candidate(s) have waited more than 6 hours for semantic review.
EVIDENCE REGISTER
A chronological register of published signals. Source class, verification state, evidence quality and relevance remain visible before interpretation.
A preprint proves rapid mixing for proper colorings and broader spin systems on graphs of girth at least five near the q=(1+δ)Δ threshold; the authors say the main proof ideas were developed through interactions with GPT-5.6 Sol Ultra.
Nature reports AGENTEX, a robotic cell-free workflow that prototypes alternative genetic codes using engineered ribosomes and tRNAs; the study demonstrates compressed codes, non-standard amino-acid incorporation and reassignment of up to three codons, while remaining in vitro rather than a demonstrated living-organism implementation.
Z.ai released GLM-5.3-Flash, a 320B MoE with 18B active parameters and publicly available weights. The model had been anonymously tested as ox-alpha. Artificial Analysis independently reports Intelligence Index 57 at very low task cost; Z.ai reports strong coding and agentic benchmark gains versus GLM-5.2.
An EPFL-led preprint reports a three-agent MCP framework that translates natural-language microscopy requests into checked commands, assesses AFM images, tunes bounded parameters and diagnoses artifacts across live nanoscale-characterization experiments.
Skild AI says S1 can execute previously unseen manipulation tasks lasting up to about 10 minutes from a single human video demonstration, without task-specific fine-tuning or post-training. In a controlled scaling comparison at 100K hours of pre-training, the company reports 66% performance for in-context demonstration prompting versus 9% for language prompting; the result remains first-party and lacks independent reproduction.
Figure unveiled Index, a global physical-data collection system for Helix that has already accumulated more than 16 million videos from 264,000 app downloads across 108 countries, processing roughly 30 minutes of new video per second. Figure says it will spend more than $1B on data and compute over the next 12 months. The signal is data-scaling infrastructure for general-purpose humanoid learning; no new independently validated capability result was released with the announcement.
OpenAI reports first measured results for its Jalapeño custom inference system on SemiAnalysis InferenceX. Across GPT-OSS 120B, DeepSeek R1 670B and Kimi K2.5 1T, Jalapeño delivered 1.5–1.9x higher peak AI work per watt and 1.7–3.6x lower end-to-end latency than compared Blackwell systems, with deployment inside OpenAI planned by year-end.
Accelerated Understanding launched a neural-operator-based physical AI system that directly predicts 4D physical evolution, reports inference beyond 5 trillion context elements, training up to 1 trillion parameters and scaling tests to 35 trillion parameters.
NVIDIA says Groq 3 LPX is entering production deployment; Artificial Analysis measured 3,431 output tokens/s on Gemma 4 31B at 100K input context, about 4x the fastest public endpoint cited by NVIDIA, while Nebius is the first AI cloud adopter.
NVIDIA reports first on-silicon Vera Rubin NVL72 measurements on SemiAnalysis AgentX production-style coding-agent sessions, reaching up to 30x higher throughput per megawatt and up to 35x lower token cost than GB300 NVL72 on selected workloads. Results are NVIDIA-measured and pending SemiAnalysis review.
Levent Alpoge publicly claims a construction of a complex structure on the six-sphere, a classical long-standing open problem, and attributes substantial work to Claude/Opus 5. The claim is not yet independently verified and S^6 has a history of failed claimed solutions.
Ava Howell submitted curve #302 to the NSF ICARM Elliptic Curve Rank Leaderboard with 31 explicitly exhibited independent rational points, proving rank >=31 unconditionally. The leaderboard commentary attributes the discovery to Claude, Levent Alpöge and Ava Howell. Exact rank 31 is additionally certified conditional on BSD+GRH. This advances the rank frontier but does not resolve whether elliptic-curve ranks over Q are unbounded.
Zeta Lab reports five original mathematical theorems produced during an AI-directed research pursuit, all accepted by the Lean proof kernel with no unfinished proof steps. The lab separately distinguishes its stronger zeta-zero bound as a candidate rather than a theorem.
NVIDIA reportedly agreed to pay about $6 billion to license Poolside AI technology, extend offers to more than 100 Poolside employees, and separately invest about $1 billion in the startup as it builds a stronger U.S. open-weight model stack around Nemotron.
NVIDIA reports that its Agentic Variation Operators (AVO) long-horizon agent architecture, paired with Claude Opus 5, achieved a 100.00 RHAE score across all 25 ARC-AGI-3 public environments and all 183 levels. The result is system-level, not a controlled model-only uplift: ARC Prize separately reports about 30.2% for Claude Opus 5 at High reasoning effort, while NVIDIA used a different agent system, reasoning setting and text-grid observation setup. No semi-private or fully private ARC-AGI-3 result is reported.
Keefer Rowan constructs the first smooth random fast dynamo on T^3. The paper states that the central proof idea was generated autonomously by ChatGPT 5.6 Sol Ultra, while the author wrote and verified the manuscript. The result is a random-flow analogue and does not fully solve the deterministic smooth fast-dynamo problem.
A newly submitted curve on the iCARM Elliptic Curve Rank Leaderboard has rank at least 30, satisfying a named AIM problem that asked for an elliptic curve over Q of rank 30. The leaderboard uses exact 2-descent certification to establish independence of the exhibited rational points. Public discussion links the discovery to Levent Alpöge, Ava Howell and Claude, but the discovery method and AI contribution have not yet been documented in a primary paper, so the AI role remains unclear.
Moderna and Merck reported that individualized mRNA neoantigen therapy intismeran autogene plus pembrolizumab met the primary recurrence-free-survival endpoint and the key distant-metastasis-free-survival endpoint in the 1,137-patient Phase 3 INTerpath-001 study after resection of high-risk stage IIB-IV melanoma. Full effect sizes and detailed data have not yet been released.
Jihao Liu presents a smooth polarized projective fivefold that is K-polystable but admits no constant-scalar-curvature Kähler metric, thereby disproving the original cscK Yau–Tian–Donaldson conjecture. The paper documents that GPT-5.6 Sol, Fable 5 and Danus produced the counterexample and proof; an improved Danus run starting only from the original problem reportedly reproduced a complete solution in 5h29m without the key human intervention used in the first run.
Generalist AI reports that GEN-1.5 can execute previously unseen short-horizon manipulation tasks after a single 3–12 second physical demonstration with no gradient update, averaging 59% success across 10 internal tasks; few-step adaptation reaches 83% after about five minutes of task data. The company also reports compositional prompting, sim-to-real prompting and some human-to-robot transfer. Results are first-party and success remains modest on short atomic tasks.
Deep Origin reran its computational drug-discovery challenge with 2026 frontier models. The best autonomous run recovered 922/1000 hidden top structures, above the prior unrestricted human reference of 77.8%; five runs exceeded that reference. Performance is strongly harness-dependent and the public benchmark may have been exposed in training, limiting causal inference.
Ouyang, Spector and Stockdale prove a dimension-free weak-type (1,1) bound with constant 2 for the vector Riesz transform, settling Stein’s 1986 ICM problem. The paper states that the proof strategy was developed through LLM dialogues involving GPT-5.6 Sol, OpenAI reasoning agents and Claude Opus 5.0; the authors then independently checked, rewrote and validated the argument.
Avik Chakravarty, Daebeom Choi and Shengjing Xu construct counterexamples disproving Sato's weak F-equivalence conjecture for nonsingular projective toric weak Fano varieties in every dimension at least three, then formulate a Gorenstein refinement. The paper states that the results were developed with assistance from GPT-5.6 Sol.
Anthropic reports that Claude autonomously orchestrated open-source protein-design and folding tools from a human-written research prompt and produced binders against 14 of 15 targets. Adaptyv Bio and Twist Bioscience independently synthesized and tested the designs; reported hit rates were 22–35%, above the 10–15% field baseline cited by Anthropic.
The Find-Attempt-Recommend pipeline scans thousands of combinatorics papers, identifies thousands of apparently open conjectures, surfaces hundreds of potential resolutions, and narrows them to a small expert-review set. This is a material AI-for-Science signal because it automates problem selection and triage, two human bottlenecks in research-level mathematics.