Skip to content
AI News HubLIVE

This edition’s highlights

Models

DevRev delivers autonomous self-service with GPT-6.1 Sol

  • DevRev's Computer and Operator leverage OpenAI models for reasoning, planning, and tool selection.
  • GPT-6.1 Sol performs well on DevRev's Enterprise-Bench benchmark, balancing reasoning, speed, and cost.
OpenAI NewsIn-site articleDevRev delivers autonomous self-service with GPT-6.1 Sol

Circuit Oracle: Automating Attribution Graph Analysis via Natural-Language Queries

  • Circuit Oracle combines a multi-agent harness with authored task skills to query attribution graphs via tool calls and answer natural-language questions about a target model.
  • It achieves 78.2±2.4% accuracy on spurious-feature detection, outperforming single-layer feature-ranking baselines.
arXiv Computational LinguisticsIn-site articleCircuit Oracle: Automating Attribution Graph Analysis via Natural-Language Queries

datasette-llm 0.1a10

  • datasette-llm 0.1a10 is an LLM integration plugin for other plugins to depend on.
  • It fixes a bug affecting the agent.datasette.io demo site where available models were not correctly limited to the configured set.
Simon Willison's WeblogIn-site articledatasette-llm 0.1a10
Agents

Together Compiler: Continuous Compilation for AI inference

  • TKCC raises experiment throughput by 252x and cuts engine startup time from 467 seconds to 9 seconds
  • Sets new low-latency records on SemiAnalysis InferenceX: 1111 TPS on MiniMax M3.1-Flash and 1205 TPS on DeepSeek v4.1-Flash
Together AI BlogIn-site articleTogether Compiler: Continuous Compilation for AI inference

Introducing India-based inference for Kimi K3 on Amazon Bedrock

  • Kimi K3 can now be invoked through an India geographic inference profile that keeps processing in Mumbai and Hyderabad Regions.
  • The global profile is the lowest-cost, most broadly available option for most workloads.
AWS Machine Learning BlogIn-site articleIntroducing India-based inference for Kimi K3 on Amazon Bedrock

Claude Platform now supports native OpenTelemetry observability in Databricks MLflow

  • Native OTel export for Claude Platform is in public beta, covering Anthropic SDKs, Messages API, and Managed Agents
  • Traces are ingested via Zerobus and stored in Unity Catalog tables, providing low-latency ingestion and governed access
Databricks BlogIn-site articleClaude Platform now supports native OpenTelemetry observability in Databricks MLflow
Research

Humanized but Still True? Auditing Facts and Detection in AI-Text Humanizers

  • All 11 commercial humanizers changed or dropped at least one checkable unit in 35-65% of documents containing such details.
  • An open 12B humanizer kept at least 97% of numbers and names in fact-dense text but reworded 67-78% of quoted spans.
arXiv Computational LinguisticsIn-site articleHumanized but Still True? Auditing Facts and Detection in AI-Text Humanizers

datasette-apps 0.3a0

  • Typing # in Datasette Agent chat autocompletes and tags apps the user can view
  • New app_add_stored_queries() and app_remove_stored_queries() tools let the agent manage an app's allowed stored queries
Simon Willison's WeblogIn-site articledatasette-apps 0.3a0
Policy

OpenAI to acquire Pierre Computer Company, Inc.

  • OpenAI is acquiring developer tools company Pierre to enhance Codex's code review and Git workflow capabilities.
  • Pierre's Trees and Diffs libraries are already integrated into the Codex app for file navigation and diff viewing.
OpenAI NewsIn-site articleOpenAI to acquire Pierre Computer Company, Inc.
Chips

You have reached the end of this edition.

Your reading list
Other updates (28)
Agents

See the agents, set the terms: new visibility and control for personal agents

Cloudflare introduces new analytics and granular access policies that let site owners identify personal AI agent traffic and decide how to allow, limit, or block it based on business needs.

Cloudflare AI BlogIn-site articleSee the agents, set the terms: new visibility and control for personal agents

Fidelity Collapses with Table Size, Not Turn Count: Delivery Architecture in a Deployed Tool-Using Agent

A controlled study on a production award-travel agent finds that feeding structured tool records into the model context and re-emitting them each turn causes fidelity to collapse as table size grows, while keeping records in session state and delivering rows over a typed side channel maintains high accuracy at lower cost.

arXiv Computational LinguisticsIn-site articleFidelity Collapses with Table Size, Not Turn Count: Delivery Architecture in a Deployed Tool-Using Agent

MEMENTO: Learning Context Management for Long Context Processing Agents

Researchers present MEMENTO, a long-context agent built on Qwen3-8B and trained to manage its context through compaction. It first bootstraps note-taking by fine-tuning on synthetic trajectories annotated with hindsight-generated notes, then applies RL with a reward combining answer correctness, note-quality heuristics, and confidence calibration. On book-level QA benchmarks, MEMENTO lifts Qwen3-8B's accuracy from 36.9% to 67.8% while never holding more than 20K tokens in context, outperforming GPT-5.5 agents and nearing a far larger GLM-5.3 agent at a fraction of the cost.

arXiv Computational LinguisticsIn-site articleMEMENTO: Learning Context Management for Long Context Processing Agents

Introducing Images 2.6

OpenAI launched Images 2.6 across ChatGPT, Codex, and the API, delivering 50% faster image generation, GIF animation, image markup and SVG export in the API, and a new complexity parameter for finer editing control.

OpenAI NewsIn-site articleIntroducing Images 2.6

Polymorphic Inference: self-tuning inference for open models

Together AI introduces Polymorphic Inference to adapt inference systems to rapidly changing models, architectures, traffic patterns, and hardware, and launches Together Compiler, an AI-agent-driven continuous compilation system that keeps production inference tuned.

Together AI BlogIn-site articlePolymorphic Inference: self-tuning inference for open models

Together Compiler: Continuous Compilation for AI inference

Together AI introduces Together Compiler (TKCC), which uses agents and persistent engines to optimize open models in under 24 hours, delivering 1.3 to 3x higher throughput as traffic changes.

Together AI BlogIn-site articleTogether Compiler: Continuous Compilation for AI inference

datasette-agent 0.5a0

datasette-agent 0.5a0 is a significant release introducing the "panel" feature, which lets the agent chat open and sit next to Datasette pages. Combined with datasette-apps, users can prompt database-backed apps into existence and modify them from the adjacent chat.

Simon Willison's WeblogIn-site articledatasette-agent 0.5a0

Why human insight is a strategic advantage for building better agents

LangChain's blog outlines three principles for effective human-in-the-loop workflows to evaluate and improve agents: expert judgment defines quality, evaluation in full context, and turning feedback into reusable evaluation data.

LangChain BlogIn-site articleWhy human insight is a strategic advantage for building better agents

LangSmith BYOC is now available on AWS, Azure, and Google Cloud

LangSmith BYOC now runs on AWS (GA), Azure and Google Cloud (Beta). Your traces, evals, and agent deployments stay in your cloud account while LangChain runs it.

LangChain BlogIn-site articleLangSmith BYOC is now available on AWS, Azure, and Google Cloud

Bring structured and unstructured data together with Amazon Quick

Amazon Quick unifies structured and unstructured data by connecting S3-backed knowledge bases with enriched datasets, enabling a custom chat agent to answer business questions across both data types through a single conversational interface.

AWS Machine Learning BlogIn-site articleBring structured and unstructured data together with Amazon Quick

ServiceNow on-behalf-of token exchange with Bedrock AgentCore Gateway

Learn how to configure Amazon Bedrock AgentCore Gateway for an on-behalf-of (OBO) token exchange with Microsoft Entra ID, so an AI agent can call the ServiceNow REST API while preserving the calling user's identity. AgentCore Gateway performs the exchange natively, with no interceptor Lambda or custom code.

AWS Machine Learning BlogIn-site articleServiceNow on-behalf-of token exchange with Bedrock AgentCore Gateway

How Gen Digital built Norton Revamp’s AI-curated News Engine with Amazon Bedrock AgentCore Web Search

Gen Digital built Norton Revamp, a generative AI-powered professional reputation management platform, with an AI-curated News Engine grounded by Web Search on Amazon Bedrock AgentCore. The platform delivers real-time, cited content suggestions, goal setting, and news feeds with zero data egress and Norton-grade privacy.

AWS Machine Learning BlogIn-site articleHow Gen Digital built Norton Revamp’s AI-curated News Engine with Amazon Bedrock AgentCore Web Search

Build a semantic layer for agentic AI with EDB Postgres AI and Amazon Bedrock AgentCore

Learn how to build a governed semantic layer for agentic AI inside EDB Postgres AI and query it from a Strands Agents agent on Amazon Bedrock AgentCore. Business meaning, join paths, and access control all live in Postgres, with GRANT and row-level security enforcing what each agent can read.

AWS Machine Learning BlogIn-site articleBuild a semantic layer for agentic AI with EDB Postgres AI and Amazon Bedrock AgentCore
Robotics

A thousand fewer paths to root

OpenAI built an autonomous pipeline that found, validated, and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel, at an API cost of $1.2 million.

OpenAI NewsIn-site articleA thousand fewer paths to root
Models

Daybreak Red is now available in AWS GovCloud (US-East)

OpenAI makes GPT-5.6 Cyber available through Daybreak Red in AWS GovCloud (US-East) via Amazon Bedrock, giving approved government security teams access for authorized cybersecurity work.

OpenAI NewsIn-site articleDaybreak Red is now available in AWS GovCloud (US-East)

GAVAGAI: Referential Alignment for Infant-Scale Vision-Language Models via Balanced Optimal Transport

GAVAGAI introduces a referential alignment loss using balanced optimal transport, allowing words to map to nothing and enforcing capacity constraints on regions. It improves word-to-picture accuracy on simulated and SAYCam infant video data, but the authors note it does not set a new state of the art.

arXiv Computational LinguisticsIn-site articleGAVAGAI: Referential Alignment for Infant-Scale Vision-Language Models via Balanced Optimal Transport

Cross-Lingual Persona Effects in LLM Negotiation

A study finds that persona instructions given to LLM negotiators (e.g., "headstrong") produce different negotiation behaviors depending on the language of interaction. Korean negotiations reached agreement more often and featured larger seller concessions than English ones, but the gap appeared only for neutral and headstrong buyers, not cooperative ones.

arXiv Computational LinguisticsIn-site articleCross-Lingual Persona Effects in LLM Negotiation

From Fragmentation to Fusion: A Survey and Roadmap for Omnimodal Embeddings

This position paper by Ziyin Zhang and 17 co-authors argues that embedding research across modalities is converging and proposes a roadmap built on three progressive capabilities: coverage, comparability, and compositionality.

arXiv Computational LinguisticsIn-site articleFrom Fragmentation to Fusion: A Survey and Roadmap for Omnimodal Embeddings

Embedding at the Speed of Light

Static embedding models replace transformer networks with a lookup table and mean pooling, embedding 32.9 million passages in 35 minutes on a laptop CPU versus about 53 days for a transformer-based model, at the cost of retrieval quality that can be offset with hybrid search.

Qdrant BlogIn-site articleEmbedding at the Speed of Light

llm 0.37.1

LLM, the command-line tool for accessing large language models, shipped version 0.37.1, a bugfix release with contributions from Steven Huang, Tao An, Arjan Mossel, and XQ.

Simon Willison's WeblogIn-site articlellm 0.37.1

Qwen 3.8 27B on my laptop can interpret model weights, too

Simon Willison reproduces a key experiment from the Weight-Space Programming paper using Qwen 3.8 27B running locally on his laptop. Given only raw policy-network weights and no tools, the model correctly identified Breakout, Pong, and Pac-Man, but failed on the more obscure Freeway. He also pushes back on broader claims that source code is becoming obsolete.

Simon Willison's WeblogIn-site articleQwen 3.8 27B on my laptop can interpret model weights, too
Research

From Lookups to Lessons: Turning Retrieved Knowledge into Parametric Memory

The paper introduces ResMem, a method that trains parametric memory offline using retrieved text to reduce storage and inference costs, improving performance across multiple QA datasets.

arXiv Computational LinguisticsIn-site articleFrom Lookups to Lessons: Turning Retrieved Knowledge into Parametric Memory

Cross-Prompt Reuse of Entity-Replacement Responses

The paper measures how entity-replacement responses transfer across English factual questions with fixed model weights, finding hidden reuse reduces no-change error but effects depend on model scale, format, and cohort, with advantages not replacement-specific.

arXiv Computational LinguisticsIn-site articleCross-Prompt Reuse of Entity-Replacement Responses

Queering the Audits: Community-Based Auditing of AI Harms to Queer Communities

A paper for the Identity Aware AI workshop at LREC 2026 reports on a participatory auditing workshop at EurIPS 2025, where 16 queer community members used the 4Cs harm taxonomy to audit four case studies across the AI lifecycle, tracing harms to problem definition and data collection and flagging extractive contractual structures.

arXiv Computational LinguisticsIn-site articleQueering the Audits: Community-Based Auditing of AI Harms to Queer Communities

Community detection in complex networks via node similarity, graph representation learning, and hierarchical clustering

The paper proposes three general hierarchical frameworks for community detection that combine linkage-based clustering, vertex proximity matrices, and graph representation learning models. Comparing 364 module combinations on Stochastic Block Model graphs and real datasets, the best pipelines (Wasserman-Faust and mutual-information-based PPMI proximity, plus deep learning-based DNGR representations) perform competitively with Leiden and Louvain while remaining hierarchical, but time complexity limits them to networks of about 10,000 vertices on personal machines.

arXiv Machine LearningIn-site articleCommunity detection in complex networks via node similarity, graph representation learning, and hierarchical clustering

Declare and optimize with UNIQUE constraints on Databricks

Databricks has made UNIQUE constraints available in Public Preview, letting tables declare multiple nullable or composite uniqueness guarantees beyond a single primary key and allowing foreign keys to reference them. Once validated and marked RELY, the optimizer can use them to eliminate redundant joins and DISTINCT aggregations, cutting over 99.99% of shuffle and delivering up to 12x faster queries in 10 TB TPC-DS benchmarks.

Databricks BlogIn-site articleDeclare and optimize with UNIQUE constraints on Databricks

Fireworks unveils new products at Forge to help companies build specialized intelligence

At Forge, Fireworks launched a suite of new products including ultra-fast inference, background mode, the Nexus model-routing platform, training and evaluation tools, and enterprise controls, all aimed at helping companies build specialized AI around their own expertise, data, and tasks.

Fireworks AI BlogIn-site articleFireworks unveils new products at Forge to help companies build specialized intelligence
Chips

Building ultra-low-latency voice AI applications on AWS

Cartesia uses purpose-built State Space Models (SSMs) and AWS infrastructure to deliver sub-100ms text-to-speech and speech-to-text at global scale. Learn how to deploy its Sonic model on Amazon SageMaker AI or self-managed Amazon EC2 GPU instances.

AWS Machine Learning BlogIn-site articleBuilding ultra-low-latency voice AI applications on AWS