Skip to content
AI News HubLIVE

This edition’s highlights

Agents

Route coding agent requests to the right model on Amazon Bedrock with NVIDIA NeMo Switchyard

  • Coding agents work in tasks, so routing should be judged by cost per solved task rather than price per token
  • NeMo Switchyard supports Stage routing between a capable and an efficient model, and Random routing for weighted A/B tests
AWS Machine Learning BlogIn-site articleRoute coding agent requests to the right model on Amazon Bedrock with NVIDIA NeMo Switchyard

Cost attribution and controls for AI developer and productivity tools on Amazon Bedrock

  • Bedrock captures the caller's full IAM ARN at the endpoint layer, enabling per-user attribution without any API changes.
  • IAM Identity Center with session tags (team, cost center) flows organizational dimensions into CUR 2.0 and Cost Explorer.
AWS Machine Learning BlogIn-site articleCost attribution and controls for AI developer and productivity tools on Amazon Bedrock

An update on our recent security incident · Cursor

  • The attack began with an access key Cursor issued to Fireworks that Fireworks exposed, leading to exposure of API credentials for less than 0.01% of customers.
  • The attacker used stolen credentials to place malicious Claude Code extension builds in the marketplace cache; at most 5,000 users were potentially affected over roughly 11 hours.
Cursor BlogIn-site articleAn update on our recent security incident · Cursor

Kitesurf is now open source

  • Kitesurf is a stateless browser for AI agents running entirely on Cloudflare Workers
  • Open sourced under Apache 2.0 license on GitHub, with local build and deployment support
Cloudflare AI BlogIn-site articleKitesurf is now open source
Policy

Apache Spark vs Apache Flink: which should you use?

  • In a Databricks benchmark, Spark Real-Time Mode matched or beat Apache Flink on p99 latency for stateless, aggregation and join workloads.
  • Spark provides a unified engine for batch, SQL, ML and streaming, avoiding the overhead of running a second engine.
Databricks BlogIn-site articleApache Spark vs Apache Flink: which should you use?

Open-source AI without Jensen

  • Jensen Huang's departure may weaken Nvidia's support for open models, affecting funding and advocacy for the open ecosystem.
  • The long-term benefit of open models as commoditized intelligence is far off due to compute scarcity, limiting their economic impact.
Interconnects (Nathan Lambert)In-site articleOpen-source AI without Jensen
Models

GenPFN: An In-Context Tabular Foundation Model Approach for Augmenting Small Datasets in Engineering

  • GenPFN keeps TabPFN weights fixed and generates candidates purely through in-context learning from labeled evaluated designs.
  • Across fourteen problems it achieves higher class satisfaction and lower realism distance than adversarial, VAE, and diffusion generators, though VAE offers better coverage and diversity.
arXiv AIIn-site articleGenPFN: An In-Context Tabular Foundation Model Approach for Augmenting Small Datasets in Engineering
Research

SynthCorp: A Platform for Creating Benchmarks for Data Agents

  • SynthCorp simulates business events on a virtual calendar to generate complex data estates comprising relational databases, document stores, and data catalogs.
  • The platform automatically generates analytical tasks and derives verified gold programs from a simulation log, eliminating manual annotation.
arXiv AIIn-site articleSynthCorp: A Platform for Creating Benchmarks for Data Agents

Advancing science & engineering with Antigravity’s /teamwork

  • Antigravity /teamwork uses multi-agent workflows with parallel exploration and dedicated falsifiers to scale long-horizon reasoning
  • Systems helped prove the 2013 Courtade–Kumar conjecture in information theory, formalizing it in 50 million lines of Lean 4 code
Google Research BlogIn-site articleAdvancing science & engineering with Antigravity’s /teamwork
Startups

LangChain Raises $300M Series C at $3B Valuation

  • LangChain raised $300M Series C at a $3B valuation, led by existing investor CapitalG.
  • New investors GIC and DFJ Growth joined, with existing investors IVP and Sapphire Ventures participating.
LangChain BlogIn-site articleLangChain Raises $300M Series C at $3B Valuation

You have reached the end of this edition.

Your reading list
Other updates (12)
Models

When Is Proof-Step Verification Worth Its Cost? Operating Requirements of Certify-or-Decline Verification for Open-Weight Mathematical Reasoning

A NeurIPS 2026 MATH-AI workshop paper evaluates certify-or-decline proof-step verification on open-weight models. Certification boosts answer-key precision by 31–46 points on GPQA and AIME but only answers 35% and 21% of attempts; on near-ceiling MATH-500, gains are at most 5 points at a 13–27 point coverage loss. The authors also identify invalid proof construction as a major execution failure and caution that current audits lack human or formal-kernel ground truth.

arXiv AIIn-site articleWhen Is Proof-Step Verification Worth Its Cost? Operating Requirements of Certify-or-Decline Verification for Open-Weight Mathematical Reasoning

Robust Task-Aware State Estimation from Pre-trained Perception

This paper introduces a method to construct a θ-confidence Minimal Belief Set (MBSθ) from the probability outputs of pre-trained perception models like VLMs, enabling robust task planning under uncertainty without determinizing predictions or maintaining full distributions.

arXiv AIIn-site articleRobust Task-Aware State Estimation from Pre-trained Perception

Rational Use of Time for Social Coordination in Humans and AI

A new study uses a cooperative card game, The Mind, which forbids explicit communication, to compare how 150 humans and two classes of AI models (computer-use agents and standard text-prompted LLMs) coordinate using only temporal signals. Human behavior is best captured by a Bayesian strategy that infers a partner's belief from timing, and frontier models show similar patterns. However, unlike humans, none of the tested models efficiently adapt their behavior to partners over the course of the game. The findings suggest humans perform implicit probabilistic reasoning about continuous-time signals in dynamic social settings, and such mechanisms may emerge without real-time experience.

arXiv AIIn-site articleRational Use of Time for Social Coordination in Humans and AI

When Should Semantic Priors Enter Partially-Identified Causal Discovery?

A new arXiv paper systematically studies three decisions for integrating LLM-provided semantic information into observational causal discovery: where to allocate semantic queries, which orientation proposals to admit, and which causal graphs to retain. Benchmarks show semantic information improves recovery over random orientations, but targeting unstable edges offers no established advantage over matched-random allocation; the tested admission certificate cannot support nontrivial admission with available calibration tasks; and learned references contain the true DAG in only 3 of 80 planned tasks.

arXiv AIIn-site articleWhen Should Semantic Priors Enter Partially-Identified Causal Discovery?

Together Compiler: Continual Compilation for AI Inference

Together AI introduces Together Compiler (TKCC), the first truly continual agent loop for AI inference optimization. It addresses the throughput-fidelity tradeoff with a Tiered Persistent Engine and a cost model, outperforming hand-written Megakernels by up to 1.7x on Llama 1B and achieving up to 2.6x higher TPS than the best open source engines on frontier models, with state-of-the-art batch-size-1 performance of 1111 TPS on M3.1 Flash and 1205 TPS on DeepSeek v4.1 Flash.

arXiv AIIn-site articleTogether Compiler: Continual Compilation for AI Inference

Build a model improvement flywheel with Amazon SageMaker HyperPod Data Capture

Amazon SageMaker HyperPod Data Capture records live inference inputs and outputs to Amazon S3. This post walks through one end-to-end example: capturing a support router's production traffic, surfacing a new-product blind spot with Amazon Athena, fine-tuning on the captured data with the HyperPod training operator, and measuring the before-and-after lift.

AWS Machine Learning BlogIn-site articleBuild a model improvement flywheel with Amazon SageMaker HyperPod Data Capture
Research
Agents

Cleaning an Agent's History Hurts When the World Still Holds the Error

A new arXiv paper shows through controlled experiments on AgentCE-Bench and AppWorld that editing an agent's history only helps when the real-world state changes along with it; making a history look clean while the error persists can leave a receiver agent worse off than no edit at all.

arXiv AIIn-site articleCleaning an Agent's History Hurts When the World Still Holds the Error

Zugwise: Domain-Aware Game Reconstruction for Digitizing Handwritten Chess Scoresheets

Zugwise reframes handwritten chess scoresheet digitization as game reconstruction, using chess engine machinery to detect silent legal-but-wrong errors and merging both players' sheets to reduce mistakes dramatically.

arXiv AIIn-site articleZugwise: Domain-Aware Game Reconstruction for Digitizing Handwritten Chess Scoresheets

Production MLOps at GS Caltex with Amazon SageMaker Canvas, part 1

GS Caltex connects Amazon SageMaker Canvas to Amazon SageMaker Pipelines so process engineers' no-code models run in production. This post shows how ML engineers reuse validated preprocessing flows and models, cutting deployment from about a week to a few hours and running batch inference every 30 minutes.

AWS Machine Learning BlogIn-site articleProduction MLOps at GS Caltex with Amazon SageMaker Canvas, part 1
Chips

OpenFold3 on Amazon SageMaker JumpStart: Predict biomolecular structures

OpenFold3-preview is now deployable from Amazon SageMaker JumpStart, letting users predict 3D structures of proteins, DNA, RNA, and ligand complexes inside their own AWS account. The post covers asynchronous inference, multi-GPU parallelism, bringing your own MSAs for accuracy and privacy, and scaling to zero to reduce cost.

AWS Machine Learning BlogIn-site articleOpenFold3 on Amazon SageMaker JumpStart: Predict biomolecular structures

Cost-optimizing AssemblyAI speech-to-text inference on AWS with Amazon EC2 Spot

AssemblyAI runs cost-optimized speech-to-text inference at scale on AWS using Amazon ECS, EC2 Spot GPU instances, and an SQS-based retry architecture, achieving 70% cost savings while maintaining SLA commitments across European Regions.

AWS Machine Learning BlogIn-site articleCost-optimizing AssemblyAI speech-to-text inference on AWS with Amazon EC2 Spot