Historical news archive. Items here are older than the 24-hour Latest News window. This page is for reference; the live AI news feed is at /news/.
Showing the 60 most recent of 3678 archived items. Browse the full history by month.
Anthropic's Python SDK v1.13.0 adds typed Chat and Cowork analytics metrics, plus workflows, multi-agent configuration, and thread-status filtering for Managed Agents.
Developers building on Managed Agents should upgrade to the typed analytics and workflow calls, because hand-rolled untyped requests risk silent breakage.
Impactful scheduling for GPU clusters
Teams affected by Impactful scheduling for GPU clusters need to decide whether its documented change alters their current workflow. For builders, this may affect how you choose, deploy, or govern AI tools this week.
3.27.0 (2026-10-09) Features api: add prewarmed hosted environments ( #4043 ) ( c511a77 )
What's actually new here is openai/openai-python - see if it moves anything you maintain. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Hermes Agent canary 20261009T070410Z
Operators using related systems should check whether NousResearch/hermes-agent changes compatibility, cost, or access requirements. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Discover how Sophos uses OpenAI’s Daybreak to cut cyber-threat investigation time by 96% and automate 52% of MDR cases while preserving human oversight.
Teams affected by Sophos cuts threat investigation time by 96% need to decide whether its documented change alters their current workflow. For builders, this may affect how you choose, deploy, or govern AI tools this week.
Using GPT-6 Astra in Codex, Asana made its browser agent 76x cheaper and 5x faster in tests to offer customers more capable models.
At its core this is about Asana cuts model costs 76x in browser tests, worth a look if that's in your workflow. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2607.16204v2 Announce Type: replace Abstract: Recent growth in reinforcement learning (RL) has surfaced a need for diverse, specialized training environments. Hand-curated environments with fixed task and reward difficulties become ineffective signals as model performance improves, and sparse rewards over long ho
The thing to notice is Masked Diffusion Language Models are Strong and Steerable; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2608.10434v2 Announce Type: replace Abstract: Machine learning-based Intrusion Detection Systems (IDS) have demonstrated superior performance in securing Unmanned Aerial Vehicle (UAV) networks. However, the 'black-box' nature of these models, combined with the high dimensionality of multimodal cyber-physical data
The thing to notice is Conversational versus Dashboard Explainable AI for UAV Intrusion; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.00874v2 Announce Type: replace Abstract: As agentic systems become compound systems, increasingly important decisions move above task execution itself: when should a higher-level controller preserve the strategy guiding another process, and when should it revise it? We study this meta-level control problem i
What's actually new here is A Score Is Not a Policy - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30294v2 Announce Type: replace-cross Abstract: Scientific presentations are more than summaries of research papers. They need to present the work in a coherent sequence, explain the main ideas clearly, and help the audience follow the presentation.
This lands on SlideLab - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.39607v2 Announce Type: replace-cross Abstract: Skills extend an agent's capabilities by injecting instructions and information into the context, and are widely used by agents such as OpenClaw and Claude Code. Prior work shows third-party marketplaces host malicious skills that give attackers direct influence
At its core this is about Pretext, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.10541v1 Announce Type: new Abstract: Knowledge Graph (KG) quality depends not only on downstream graph validation, but also on the quality of tabular metadata used before integration. In metadata-only Semantic Table Interpretation (STI), where cell values are unavailable, noisy, or unsuitable, column headers
This lands on An Explainable Header-Centric Framework for Large-Scale Semantic Table - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.10549v1 Announce Type: new Abstract: Tool-calling agents have become central to enterprise AI, yet training and evaluating them at scale remains severely constrained due to business and legal restrictions on enterprise systems, data, and database schemas. Tabular data synthesis offers a natural alternative
Operators using related systems should check whether Synthesis Through Simulation changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11123v1 Announce Type: new Abstract: Most human-agent interaction today remains text-based. Natural language can impose cognitive overload, ambiguity, information chaos, and slow input for complex tasks; ephemeral generative UIs can present structured information and guide users toward task completion.
This lands on When Interfaces Speak - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11317v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) architectures have become essential for scaling large language models, with recent work demonstrating the benefits of fine-grained expert designs. Training such models from scratch is expensive, and sparse upcycling from pre-trained dense models i
This lands on DivMoE - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11344v1 Announce Type: new Abstract: Physics-based models connect scientific explanation with quantitative prediction. Constructing them requires selecting physical processes, defining states and governing equations, specifying couplings, and identifying parameters from experiments.
At its core this is about EvoSim, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11483v1 Announce Type: new Abstract: Diffusion-based driving planners capture diverse behaviors but can generate unsafe trajectories under distribution shift. We propose BridgeGuard, a safety-constrained diffusion planning method that progressively strengthens a constraint term during denoising to drive inte
What's actually new here is BridgeGuard - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11527v1 Announce Type: new Abstract: Atomistic simulation is a fundamental tool for studying long-term materials evolution, from diffusion and defect dynamics to interfacial reactions and fracture. Yet conventional simulators typically advance at microscopic resolution, spending substantial computation on lo
Builders evaluating AtomWorld-Mirror should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11723v1 Announce Type: new Abstract: Multiplayer world models must ensure that independently controlled views remain consistent with one shared and persistent world. We introduce MultiWorldBench, a diagnostic Minecraft benchmark containing 495 case configurations across seven task suites and ten capabilities
Operators using related systems should check whether MultiWorldBench changes compatibility, cost, or access requirements. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2610.11738v1 Announce Type: new Abstract: Reliable uncertainty quantification (UQ) is essential for deploying neural networks in scientific and high-stakes applications, but full Bayesian inference over the network parameters is computationally infeasible. We propose a low-rank generalized Laplace approximation f
The thing to notice is Scalable AI Uncertainty Quantification via Generalized Laplace Active; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11754v1 Announce Type: new Abstract: Some defects in an AI-generated study can be identified from its artifacts; others require knowledge of what was approved before execution. We propose study contracts that bind declared experimental choices, run obligations and claim scope to recorded execution evidence
Builders evaluating What Output-Only Review Cannot Verify should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11775v1 Announce Type: new Abstract: Sparse Mixture of Experts (MoE) models scale more efficiently than dense models by routing tokens to modular expert networks that are only active for processing a fraction of tokens. A leading hypothesis for the performance of MoE models is that each expert specialises in
At its core this is about RouterInterp, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11833v1 Announce Type: new Abstract: Neural networks trained on modular addition tasks often develop Fourier-structured representations that support exact generalization. While prior work has identified these Fourier circuits, the mechanism by which gradient-based training selects them from the data distribu
Teams affected by Probability-Signature Dynamics need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11989v1 Announce Type: new Abstract: On-policy distillation (OPD) trains a student on its own generated responses using token-level teacher supervision. However, uniform weighting overlooks differences in token learning value, while existing weighting methods rely on predefined mappings from prediction signa
This lands on MetaOPD - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.12129v1 Announce Type: new Abstract: A large body of research measures model coherence based on output variance without adequately considering competing causes. We identify two such causes, ambiguity and indifference, and we introduce a set of 175 questions where contradicting answers cannot easily be explai
The thing to notice is An Investigation of Model Coherence; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.10845v1 Announce Type: cross Abstract: A large language model can only use the text that fits in its context window, and it recomputes its internal key-value (KV) state for a prompt every time the prompt is sent. We test a memory layer, the public package galahad-kv, that saves the KV state of each block of
This lands on Real Long-Term Memory for AI - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.10924v1 Announce Type: cross Abstract: The Tensor-Memory Equilibrium (TME) model of "FP8 is All You Need (Part 1)" calculates the execution time of Ozaki Scheme II emulation of fp64 as the maximum of a tensor-core term and a High-Bandwidth Memory (HBM) traffic term, plus a per-output reconstruction term. How
The thing to notice is The Missing Fourth Term for the Emulation Tensor; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.10961v1 Announce Type: cross Abstract: Coding agents increasingly share a workstation while drawing on separate providers and subscription allowances. A second agent can inspect a completed answer, but the call spends another pool and may provide no substantive finding.
The thing to notice is Cross-Provider Review as a Runtime Contract for Coding; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11008v1 Announce Type: cross Abstract: Language-model agents favor their own group because they have watched their members favor each other. The group label alone does little once the decision has a cost; what drives favoritism is observed behavior, and an individual's own record can override it.
Teams affected by Why LLM Agents Favor Their Group need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11252v1 Announce Type: cross Abstract: This study presents the Neuro-Memory Fuzzy Inference System (NeMeFIS), a hierarchical machine learning architecture that asymmetrically models acceleration and deceleration in car following behavior by integrating five human memory types procedural, working, episodic, s
The thing to notice is Neuro-Memory Fuzzy Inference System for Mimicking Human-like Car; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11316v1 Announce Type: cross Abstract: System One models output constrained decisions and probability distributions rather than free-form text generation. While prevailing paradigms rely on structured schema objects to encode state, intent, and candidate choices, we revisit a fully natural language-based Sys
Builders evaluating MetaEncoder should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11690v1 Announce Type: cross Abstract: In the graph model for conflict resolution (GMCR), a decision maker (DM) either moves the conflict to another state or does nothing. The basic definitions leave inaction implicit, so every action that leaves the state unchanged is treated as doing nothing.
Builders evaluating Epistemic Disturbance in the Graph Model for Conflict should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11923v1 Announce Type: cross Abstract: The brain maintains stable cognition despite continuously changing neural activity. How to extract stable cognitive states from variable neural observations remains a central problem in neural decoding.
This lands on Neural Decoding as Cognitive Inference - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.11963v1 Announce Type: cross Abstract: A no-code fix resolves an invalid bug report by directing the user to change a setting, update to a version where the problem is already fixed, or adjust their workflow. Manually verifying whether a proposed no-code fix resolves the reported bug takes considerable devel
At its core this is about Can LLMs Fix It Without Code? Toward Automated, worth a look if that's in your workflow. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2610.12007v1 Announce Type: cross Abstract: Flow-based vision-language-action (VLA) models generate action chunks for temporally coherent robot motion, but chunked control creates a fundamental closed-loop trade-off: long chunks provide smooth execution, whereas frequent replanning improves reactivity at the cost
Teams affected by REACT need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.12033v1 Announce Type: cross Abstract: Robotic systems operating over extended tasks must maintain a world state assembled from observations arriving at different times, with varying confidence and potential revisions. Conventional representations emphasize latest estimates, hindering fact provenance, decisi
At its core this is about Traceable World State, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.12182v1 Announce Type: cross Abstract: Very-high-resolution Earth-observation satellites acquire more data than they can store and downlink, while in maritime surveillance the vessels cover a tiny fraction of each scene. We study onboard vessel detection as a way to select what is downlinked, which reduces t
The thing to notice is AI-Based On-Board Maritime Object Detection for Earth Observation; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2610.12449v1 Announce Type: cross Abstract: Bifurcations are ubiquitous in physical systems, from structural buckling to fluid and climate dynamics, yet they remain largely unexplored in deep learning. At a symmetry-breaking bifurcation, a single input admits multiple equally valid solutions, violating the one-to
What's actually new here is Bi-FORK - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
A new paper compares 2026 agent-security incidents at OpenAI, Anthropic, and Google agents that reached real systems beyond test scope, and proposes a Proactive Agent Security Assurance Cycle with a five-layer boundary model.
Operators deploying tool-calling agents should test boundary enforcement at runtime, because the paper shows assumed test boundaries failed and agents reached real systems, raising shutdown and incident-response risk.
arXiv:2501.07440v3 Announce Type: replace-cross Abstract: Paying attention improves performance, but attention is metabolically costly, so how should a resource-efficient agent allocate it? We study optimal allocation strategies using a normative model of a signal detection task in which attention comes at a cost.
At its core this is about Attention when you need, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2511.04000v2 Announce Type: replace-cross Abstract: Decision trees are widely used in high-stakes fields like finance and healthcare due to their interpretability. This work introduces an efficient, scalable method for generating synthetic pre-training data to enable meta-learning of decision trees.
This lands on Towards Scalable Meta-Learning of near-optimal Interpretable Models - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2601.18751v2 Announce Type: replace-cross Abstract: Preference-based reinforcement learning (PBRL) offers a promising alternative to explicit reward engineering by learning from pairwise trajectory comparisons. However, real-world preference data often comes from heterogeneous annotators with varying reliability
Teams affected by Trust, Don't Trust, or Flip need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2602.01683v2 Announce Type: replace-cross Abstract: Transitioning Multimodal Large Language Models (MLLMs) from offline to online streaming video understanding is essential for continuous perception. However, existing methods lack flexible adaptivity, leading to irreversible detail loss and context fragmentation.
The thing to notice is FreshMem; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2603.07980v2 Announce Type: replace-cross Abstract: As language models (LMs) evolve from chat assistants to long-horizon agents capable of multi-step reasoning and tool use, existing benchmarks remain largely confined to structured or exam-style tasks that fall short of real-world professional demands. To this en
What's actually new here is \$OneMillion-Bench - see if it moves anything you maintain. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2603.08993v2 Announce Type: replace-cross Abstract: System prompts for LLM-based coding agents are software artifacts that govern agent behavior, yet lack the testing infrastructure applied to conventional software. We present Arbiter, a framework combining formal evaluation rules with multi-model LLM scouring to
This lands on Arbiter - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2605.18162v2 Announce Type: replace-cross Abstract: Vision-Language Models (VLMs) have made striking progress, yet their spatial reasoning remains fragile. Models that answer an original input correctly can still fail under valid transformations with predictable answer mappings, revealing a gap between instance-l
This lands on Self-Evolving Spatial Reasoning in Vision Language Models - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2605.22342v2 Announce Type: replace-cross Abstract: While 4D Gaussian Splatting (4DGS) has revolutionized high-fidelity dynamic reconstruction, safeguarding the intellectual property of these assets remains an open challenge. Conventional steganographic techniques often neglect the underlying kinematic manifolds
The thing to notice is 4D-GSW; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2608.15108v2 Announce Type: replace-cross Abstract: Large language model agents are increasingly connected to high-value resources, including external APIs, GPUs and servers, and workflows such as deployment and approval. Existing agent security research mainly focuses on attacks against information and agent beh
The thing to notice is Beyond Direct Access; decide whether it changes your next build. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
{"attempt":5,"autopublish":false,"claimEpoch":1791518021,"commit":"17...
Teams affected by NousResearch/hermes-agent need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
{"attempt":4,"attemptRef":"rc.4-v0.21.7","schema":1,"version":"0.21.7"}
Builders evaluating NousResearch/hermes-agent should verify the source before changing a production default. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Ollama v0.40.2 migrates previously downloaded models to llama.cpp runners in the background on first run, keeps a temporary on-disk backup for safe downgrades, and fixes duplicate entries in ollama list.
Users running local models should upgrade promptly but budget extra temporary disk space, because first-run migration keeps a full backup copy until a future release removes it automatically.
GitHub Releases published a source-backed AI update around vllm-project/vllm: proto-v0.5.0. RepoRadar is keeping the source link direct for verification.
At its core this is about vllm-project/vllm, worth a look if that's in your workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
OpenAI's October 9, 2026 ChatGPT release notes list Composer predictions in Codex. Direct fetch was bot-walled (403); the dated entry is corroborated via the indexed notes snippet.
Prediction-style completion inside Codex changes the cost/latency profile of agentic coding sessions.
BabyTalk launched on Show HN October 9, 2026: an MIT-licensed library for offline speech-to-text and text-to-speech on ESP32-S3/P4 with 4-bit integer kernels.
Launch-day record for the BabyTalk card; links the two surfaces.
Veil launched on Show HN October 9, 2026, presenting a zero-knowledge messenger architecture with post-quantum encryption (item 50015160). Audit status is unverified.
Launch-day record for the Veil card; links the two surfaces.
Apogee launched on Show HN October 9, 2026 as a fully local, privacy-focused rebuild of Mozilla's discontinued Orbit summarizer, connecting to local Ollama with WebGPU/Transformers.js options (item 50017301).
Launch-day record; the Apogee card already exists and is linked this cycle.
Google and BIDMC researchers report in The Lancet that 98 urgent-care patients consulted the AMIE diagnostic chatbot before visits: no session needed interruption under the study's stop criteria, visit summaries aided prep in 75% of cases, and differentials matched final diagnoses 90%.
Operators running clinic intake pilots should test chatbot triage against supervised-visit outcomes before wider deployment, because the study links AI prep summaries to visit readiness while its authors note larger trials are still needed.
Changes since langchain==1.4.3 release(langchain): 1.4.4 ( #41163 ) fix(langchain): (SummarizationMiddleware) retry summary step on context overflow ( #41159 ) chore(deps): bump fsspec from 2026.6.0 to 2026.9.0 in /libs/langchain_v1 ( #41097 ) chore(deps): bump multidict from 6.7.0 to 6.9.1 in /libs/langchain_v1 ( #410
Teams affected by langchain-ai/langchain need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Changes since langchain-core==1.6.8 release(core): 1.6.9 ( #41160 ) feat(core): accept a callable in with_retry ( #41158 )
Teams affected by langchain-ai/langchain need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub now lets users with the triage role or higher archive and unarchive pull requests; archiving closes the PR, makes it fully read-only, and hides it from public view while staying visible to admins.
Teams running busy open-source repos should enable triage-role archiving to retire spam, duplicate, and abandoned pull requests, because moderation no longer waits on admins and archived threads stay read-only without extra workflow.