Live topic stream

Models & Research

Name: LDS Mentor
Availability: InStock

Model releases, research results, benchmarks, evaluations, and methods. Reader coverage is updated continuously; Google Search eligibility is decided separately.

Stream archive

Latest coverage

Browse all topics

7.2
ai-labor-market
Aug 25, 20262 sources
Stanford Finds No Broad AI Job Displacement
Stanford Digital Economy Lab researchers found no widespread economy-wide job displacement associated with AI through mid-2026, based on ADP payroll data. Their revised analysis found that employment among 22-to-25-year-olds in highly AI-exposed occupations was 19% below a comparable low-exposure trend, with the difference appearing primarily tied to reduced hiring rather than separations. The researchers caution that the descriptive data do not establish causation.
Read story
5.0
ai-research
Aug 25, 20262 sources
Geneformer-Guided Study Identifies Pbx1 in Aging Blood Stem Cells
Researchers at Tohoku University reported that a Geneformer-guided analysis of mouse blood stem cells identified Pbx1 as a regulatory hub in age-related changes to blood production. The Science Advances study, published Aug. 22, combined single-cell data, multi-omics work, and animal experiments; it does not establish a treatment for people.
Read story
6.4
linkedin
Aug 24, 20262 sources
Workers Retroactively Add AI Language to LinkedIn Profiles
A National Bureau of Economic Research working paper finds that 19.7% of established U.S. LinkedIn users retroactively edited a past job title or description between 2020 and 2026. Using monthly Revelio Labs data covering 29.4 million profiles, the researchers documented a sharp post-2022 rise in AI-related language, alongside declines in remote-work and DEI terminology.
Read story
6.8
rust
Aug 24, 20265 sources
Canonical Funds AI Research for C-to-Rust Translation
On August 20, 2026, Canonical and the University of Bristol began funding a three-year PhD project to investigate translating C repositories with hundreds of thousands of lines into safe, behaviorally correct, maintainable Rust. Reporting by InfoWorld identifies Alex Wood as the PhD student, supervised by Professor Luc Moreau with Canonical engineering VP Jon Seager and Professor Peter Flach. The research will use AppArmor and snap-confine as production-oriented case studies.
Read story+2
7.1
anthropic
Aug 23, 20262 sources
Ramp data shows limited early business uptake for Anthropic’s Fable 5
Ramp’s August AI Index found that Fable 5 accounted for 6% of Anthropic token purchases and 11.4% of dollar spend in the month it measured. The company says the sample tracks its token-spend management users and skews more tech-oriented, so the figures are an early usage signal rather than a measure of all enterprise demand.
Read story
6.4
gemini-3-5-pro
Aug 23, 20263 sources
Gemini 3.5 Pro Launch Remains Unconfirmed
NokiaPowerUser reported on August 23 that Gemini 3.5 Pro could launch that day, citing leaked specifications including a possible 2 million-token context window. Google has not announced a broad release: its July 21 blog post stated that Gemini 3.5 Pro was testing with partners and would become broadly available when ready.
Read story
5.8
drug-discovery
Aug 23, 20262 sources
SFU Researchers Present CGFlow for Drug Design
Simon Fraser University researchers presented CGFlow in July and August 2025, introducing an AI framework that designs drug molecules alongside feasible chemical synthesis routes. SFU coverage describes the method as combining stepwise molecular construction with 3D modeling to address a recurring limitation in computational drug design: candidates that appear promising but cannot be made in a laboratory.
Read story
5.9
ai research
Aug 23, 20262 sources
2018 Study Tested an AI System for Cooperation With People
In 2018, researchers reported that an algorithm called S# could sustain cooperation with people and other machines in repeated two-player games. The Nature Communications study tested communication-based mechanisms alongside machine learning, framing the result as research on social interaction rather than a deployed assistant or general-purpose system.
Read story
6.4
ai research
Aug 23, 20262 sources
Stanford's Decagon Model Predicted Side Effects of Drug Pairs
Stanford researchers reported in July 2018 that Decagon, a graph-based machine-learning system, could predict potential side effects associated with pairs of drugs. Built from more than 4 million known drug-side-effect associations and a network of roughly 19,000 proteins, the research was designed to prioritize interactions for follow-up, not replace clinical testing.
Read story
6.4
ox-alpha
Aug 22, 20262 sources
Ox Alpha Debuts Anonymously on OpenRouter
An anonymous model called Ox Alpha appeared on OpenRouter on August 20 with a 1,048,576-token context window and text, image, and video inputs, according to OpenRouter listing details reported by GLM5.app. Community testing has linked its tokenizer, video behavior, and backend errors to Zhipu AI's GLM family, but neither Zhipu AI nor OpenRouter has publicly claimed the model. Early coding benchmark claims remain unverified.
Read story
6.8
generative-ai
Aug 22, 20262 sources
Pew Study Finds Signs of AI Authorship on New Webpages
Pew Research Center found signs of AI writing or substantial AI editing in about 10% of a random sample of 10,000 English-language webpages collected in July 2026. The share rose to 35% among dated pages published after ChatGPT's 2022 release, but the study cautioned that detector results are not proof about any individual page.
Read story
6.8
ai-research
Aug 22, 20262 sources
Preprint Estimates Widespread AI Assistance in Biomedical Papers
An August 11 preprint analyzing 1.19 million open-access biomedical papers in PubMed Central estimated that 89% of papers published in December 2025 showed an excess of vocabulary associated with large-language-model assistance. The authors describe an estimate of assisted-writing prevalence, not proof that AI generated an entire paper, with implications for publication-policy debates.
Read story
6.8
gpt-4
Aug 22, 20263 sources
HUJI Study Uses GPT-4 to Build Personality Surveys
Hebrew University researchers reported in iScience on August 6 that GPT-4 generated personality questionnaires from DSM-5 and astrology texts, then forecast population-level response patterns before 600 participants completed them. The DSM-based questionnaire showed stronger internal consistency than the astrology-based version, a reminder that generated survey items still require human validation.
Read story
6.4
machine-learning
Aug 22, 20262 sources
UC San Diego team uses machine learning to map a gene-activation sequence
UC San Diego researchers reported on August 21 that they used machine learning and high-throughput sequencing to model the DNA initiator, a core-promoter element involved in starting gene transcription. The team analyzed about 500,000 sequence variants and found the initiator in roughly 60% of focused human promoters, a result that could help researchers study mutations that affect gene regulation.
Read story
7.8
nvidia
Aug 21, 20262 sources
NVIDIA AVO Achieves Perfect Public ARC-AGI-3 Score
NVIDIA's AVO agent achieved a 100.00 RHAE score on the public ARC-AGI-3 benchmark on August 21, completing all 183 levels across 25 environments. NVIDIA reports that AVO uses Claude Opus 5 as its primary model and that the full agent system outperformed the model's reported 30% baseline. The result applies to the public set, not a hidden evaluation set.
Read story
6.2
protein-design
Aug 21, 20262 sources
RFdiffusion2 Designs Functional Zinc Enzymes in Lab Tests
Researchers from the University of Washington, the Institute for Protein Design and ETH Zurich reported in Nature on December 3, 2025 that RFdiffusion2 generated zinc-dependent metallohydrolases from quantum-chemistry-defined active sites. In laboratory tests, the strongest first-round design recorded a catalytic efficiency of 16,000 M−1 s−1, while later designs reached 53,000, showing that the protein-generation pipeline can produce active enzymes without directed-evolution optimization.
Read story
6.9
computational-genomics
Aug 21, 20262 sources
DANDELION Identifies Overlooked Asthma Disease Genes
On August 11, 2026, researchers at Columbia University and the University of Chicago reported DANDELION, a computational framework that identified 21 asthma-related genes, most missed by other methods. The University of Chicago reports that CRISPR screens and mouse models validated asthma phenotypes, while two identified genes converged on a previously unstudied asthma pathway involving fatty-acid metabolism and protein palmitoylation.
Read story
6.4
qwen
Aug 20, 20262 sources
Qwen3.8 27B reaches 52 on Artificial Analysis Intelligence Index
Artificial Analysis lists Qwen3.8 27B at 52 on its Intelligence Index, placing the 27-billion-parameter open-weights model among the stronger small models in its August comparison. The benchmark page also lists a 256,000-token context window. The score is a comparative benchmark result, not a guarantee of equivalent performance on every production workload.
Read story
6.7
embodied-ai
Aug 20, 20263 sources
Unitree CEO Forecasts Embodied AI Breakthrough Window
Unitree founder and CEO Wang Xingxing said on August 20 that humanoid robotics may be two to three years from a "ChatGPT moment" for embodied intelligence. Speaking at Beijing's World Robot Conference, Wang described a future threshold in which robots could complete roughly 80% of tasks in unfamiliar homes from voice or text instructions. The estimate is a founder forecast, not a disclosed model release or independently validated benchmark.
Read story
3.0
ai-safety
Aug 20, 20262 sources
LeakGauge Research Proposes Early Detection for LLM Context-Leakage Attacks
An Aug. 19 research preprint introduced LeakGauge, a method that uses prefill token probabilities to estimate whether an LLM prompt may trigger context leakage before generation begins. Across 11 tested models, the authors report AUROC scores of 0.944 to 0.996 and released code for the approach; the findings have not yet been independently replicated.
Read story
5.7
ai-research
Aug 20, 20262 sources
University of Essex Study Uses AI to Design Intrabodies for Cell Research
A University of Essex-led study reported a way to redesign antibody fragments so they can remain functional inside human cells. Published in Nature Communications in March, the work generated more than 600 candidate intrabody sequences aimed at intracellular targets, including proteins linked to neurodegenerative disease. It is a research advance and early therapeutic route, not a treatment proven in people.
Read story
7.2
plasma-proteomics
Aug 20, 20262 sources
Blood Proteomics Study Links Cell-Specific Aging to Disease Risk
A June 15 Nature Medicine study used more than 7,000 plasma proteins from 60,542 participants to estimate biological age across more than 40 cell types. The models' cell-specific aging signatures were associated with disease and mortality outcomes over as much as 15 years of follow-up, a research result rather than a clinical diagnostic.
Read story
4.4
minecraft
Aug 19, 20262 sources
Minecraft Creator Demonstrates a Generative Language Model Built With Command Blocks
A Minecraft creator has demonstrated a small generative language model built entirely with vanilla command blocks, according to an original community post and a Tom's Hardware report published August 19. The optimized version uses 445,782 command blocks, down from more than 1.9 million, and runs without mods, plugins, or datapacks. It is a technical proof of concept rather than a practical chat service.
Read story
5.8
artificial-intelligence
Aug 18, 20262 sources
Pew: 55% of Young Adults Are More Concerned Than Excited About AI
A Pew Research Center survey found 55% of U.S. adults under 30 are more concerned than excited about AI, up 24 points from 2021. In the same analysis, 73% said AI would reduce U.S. jobs over the next two decades, showing how employment expectations shape a more skeptical view among younger adults.
Read story
7.6
diffusion-models
Aug 18, 20262 sources
MIT Study Finds Diffusion Outputs Often Defy Individual Training-Data Attribution
A peer-reviewed MIT CSAIL study published August 18 found that diffusion-model outputs become harder to attribute to any single training image, person, or artist as training sets grow. Across 24 model ensembles trained on up to 162,770 images, the researchers observed attribution decay, while stressing that the result does not rule out memorization or apply automatically to large language models.
Read story
7.4
ai agents
Aug 18, 20262 sources
Databricks Reports 63.3% Win at Grounded Reasoning Cup
Databricks published results from its June 17 Grounded Reasoning Cup, where Stanford's agent won with 63.3% accuracy on a held-out benchmark built from about 120,000 pages of U.S. Treasury documents. The average team scored about 41%, showing that retrieval, document parsing, verification and latency engineering still materially shape agent performance.
Read story
6.8
rippling
Aug 18, 20262 sources
Rippling Shares 2,100-Run AI Agent Benchmark
Rippling President and CPO Matt MacInnis said on July 30 that the company ran roughly 2,100 scored attempts per model across 15 AI models on payroll, personnel, and finance tasks. SaaStr's August 18 analysis found the leading models clustered within a few accuracy points while differing sharply in cost and latency, underscoring that model selection depends on the workload rather than one overall winner.
Read story
7.6
qwen
Aug 17, 20262 sources
Alibaba Releases Qwen3.8-27B and a 2.4T Open-Weight Model
Alibaba released Qwen3.8-27B, a 27-billion-parameter multimodal model whose official repository uses Apache 2.0, and published weights for the 2.4-trillion-parameter Qwen3.8-2.4T-A95B model. CNBC reported the rollout on August 17 and Alibaba's claim that the 27B model can match a model ten times larger. The two releases have different license terms, so weight availability does not imply identical commercial rights.
Read story
5.7
large-language-models
Aug 17, 20263 sources
Google Research Examines Entity Order in LLM Recall
Google Research published findings on August 12 showing that factual errors in frontier LLMs can stem from recall failures rather than missing encoded knowledge. Search Engine Journal reports that the evaluated models encoded 95-98% of tested facts but could not directly recall 26-34%, with reverse subject-object question phrasing contributing to retrieval difficulty.
Read story
7.6
glm-5-3
Aug 17, 20264 sources
Z.ai Unveils GLM-5.3 for Coding and Cybersecurity
Z.ai unveiled GLM-5.3 on August 14, reporting large post-training gains for long-horizon coding and cybersecurity tasks without changing GLM-5.2's 743 billion-parameter base model. The company reports an 84.5% CyberGym score but a lower 54.4% ExploitBench result, and says open weights will follow a two-week safety review and hardening period.
Read story+1

Models & Research

Latest coverage

Stanford Finds No Broad AI Job Displacement

Geneformer-Guided Study Identifies Pbx1 in Aging Blood Stem Cells

Workers Retroactively Add AI Language to LinkedIn Profiles

Canonical Funds AI Research for C-to-Rust Translation

Ramp data shows limited early business uptake for Anthropic’s Fable 5

Gemini 3.5 Pro Launch Remains Unconfirmed

SFU Researchers Present CGFlow for Drug Design

2018 Study Tested an AI System for Cooperation With People

Stanford's Decagon Model Predicted Side Effects of Drug Pairs

Ox Alpha Debuts Anonymously on OpenRouter

Pew Study Finds Signs of AI Authorship on New Webpages

Preprint Estimates Widespread AI Assistance in Biomedical Papers

HUJI Study Uses GPT-4 to Build Personality Surveys

UC San Diego team uses machine learning to map a gene-activation sequence

NVIDIA AVO Achieves Perfect Public ARC-AGI-3 Score

RFdiffusion2 Designs Functional Zinc Enzymes in Lab Tests

DANDELION Identifies Overlooked Asthma Disease Genes

Qwen3.8 27B reaches 52 on Artificial Analysis Intelligence Index

Unitree CEO Forecasts Embodied AI Breakthrough Window

LeakGauge Research Proposes Early Detection for LLM Context-Leakage Attacks

University of Essex Study Uses AI to Design Intrabodies for Cell Research

Blood Proteomics Study Links Cell-Specific Aging to Disease Risk

Minecraft Creator Demonstrates a Generative Language Model Built With Command Blocks

Pew: 55% of Young Adults Are More Concerned Than Excited About AI

MIT Study Finds Diffusion Outputs Often Defy Individual Training-Data Attribution

Databricks Reports 63.3% Win at Grounded Reasoning Cup

Rippling Shares 2,100-Run AI Agent Benchmark

Alibaba Releases Qwen3.8-27B and a 2.4T Open-Weight Model

Google Research Examines Entity Order in LLM Recall

Z.ai Unveils GLM-5.3 for Coding and Cybersecurity

Models & Research

Latest coverage

Stanford Finds No Broad AI Job Displacement

Geneformer-Guided Study Identifies Pbx1 in Aging Blood Stem Cells

Workers Retroactively Add AI Language to LinkedIn Profiles

Canonical Funds AI Research for C-to-Rust Translation

Ramp data shows limited early business uptake for Anthropic’s Fable 5

Gemini 3.5 Pro Launch Remains Unconfirmed

SFU Researchers Present CGFlow for Drug Design

2018 Study Tested an AI System for Cooperation With People

Stanford's Decagon Model Predicted Side Effects of Drug Pairs

Ox Alpha Debuts Anonymously on OpenRouter

Pew Study Finds Signs of AI Authorship on New Webpages

Preprint Estimates Widespread AI Assistance in Biomedical Papers

HUJI Study Uses GPT-4 to Build Personality Surveys

UC San Diego team uses machine learning to map a gene-activation sequence

NVIDIA AVO Achieves Perfect Public ARC-AGI-3 Score

RFdiffusion2 Designs Functional Zinc Enzymes in Lab Tests

DANDELION Identifies Overlooked Asthma Disease Genes

Qwen3.8 27B reaches 52 on Artificial Analysis Intelligence Index

Unitree CEO Forecasts Embodied AI Breakthrough Window

LeakGauge Research Proposes Early Detection for LLM Context-Leakage Attacks

University of Essex Study Uses AI to Design Intrabodies for Cell Research

Blood Proteomics Study Links Cell-Specific Aging to Disease Risk

Minecraft Creator Demonstrates a Generative Language Model Built With Command Blocks

Pew: 55% of Young Adults Are More Concerned Than Excited About AI

MIT Study Finds Diffusion Outputs Often Defy Individual Training-Data Attribution

Databricks Reports 63.3% Win at Grounded Reasoning Cup

Rippling Shares 2,100-Run AI Agent Benchmark

Alibaba Releases Qwen3.8-27B and a 2.4T Open-Weight Model

Google Research Examines Entity Order in LLM Recall

Z.ai Unveils GLM-5.3 for Coding and Cybersecurity