Nightly
No. 18Friday, 31 July 20262 min read

Fri 31 Jul

You’re reading an archived edition. Today’s is here →

Top picks

read these, skip the rest
Top pickContestCodeforces

Codeforces Round starts August 4

  • Aug 4
  • 2.2h

The contest is an ICPC contest and will start at 14:35 UTC.

Top pickResearcharXiv cs.CL

ORCA-bench evaluates language models for oncall root cause analysis.

  • 1,079 RCA tasks
  • 6 days of metrics

ORCA-bench provides a production-fidelity oncall setting with real telemetry interfaces and full source-code access.

arxiv.org
Top pickResearcharXiv cs.LG

CoT reasoning-enabled triage classifier reaches 82.6% test accuracy.

  • 82.6% test accuracy

The system combines automated prompt optimization, self-training, and reinforcement learning with verifiable rewards.

arxiv.org
Top pickResearcharXiv cs.AI

MANTA framework enables self-evolving multi-agent systems.

MANTA initializes a task-conditioned topology and applies bounded structural updates during deployment.

arxiv.org
Top pickAnalysisHacker News105 pts

Distilled GPT-OSS model does not transfer censorship from teacher model.

  • 83.61% FinanceReasoning score

The distilled model's behavior remained the same as its American base model.

ctgt.ai
10 stories
Top pickResearcharXiv cs.CL

ORCA-bench evaluates language models for oncall root cause analysis.

  • 1,079 RCA tasks
  • 6 days of metrics

ORCA-bench provides a production-fidelity oncall setting with real telemetry interfaces and full source-code access.

arxiv.org
Top pickResearcharXiv cs.LG

CoT reasoning-enabled triage classifier reaches 82.6% test accuracy.

  • 82.6% test accuracy

The system combines automated prompt optimization, self-training, and reinforcement learning with verifiable rewards.

arxiv.org
Top pickResearcharXiv cs.AI

MANTA framework enables self-evolving multi-agent systems.

MANTA initializes a task-conditioned topology and applies bounded structural updates during deployment.

arxiv.org
Top pickAnalysisHacker News105 pts

Distilled GPT-OSS model does not transfer censorship from teacher model.

  • 83.61% FinanceReasoning score

The distilled model's behavior remained the same as its American base model.

ctgt.ai
Newsopenai blog

Avatarin built a 24/7 retail agent with GPT-Realtime.

  • 30,000 users
  • 92% positive survey responses

The agent provided multilingual support to Yamada Denki shoppers.

huggingface blog

Idle GPUs are a significant issue in GPU management.

Idle GPUs can lead to wasted resources and decreased productivity.

Newsopenai blog

GPT-5.6 offers lower pricing for Luna and Terra.

The more efficient models help enterprises deploy AI workflows at scale.

Newsdeepmind blog

Gemini Robotics ER 2 enables robots to reason and collaborate.

The system represents a step change in video understanding and multi-robot collaboration.

NewsAnthropic

Frontier Red Team investigates real-world incidents.

  • Jul 30, 2026

The investigation is part of the company's cybersecurity evaluations.

RepoHF trending132 likes

Audio8 model is a text-to-speech model.

  • 225 downloads
  • 132 likes

The model is available on the HF trending platform.

6 stories
Releaselwn

Stable kernels released with single fix for use-after-free vulnerability.

  • CVE-2026-64560
  • 6.18.41

Users are advised to upgrade to the latest stable kernels.

lwn.net
Newslwn

Security updates issued for various packages.

The updates are available for AlmaLinux, Debian, Fedora, Oracle, and SUSE.

lwn.net
Newslwn

LWN.net Weekly Edition for July 30, 2026 released

  • July 30, 2026

The edition covers various topics including hazard pointers and BPF inlined functions.

lwn.net
Researchlwn

Reconsidering O_CREAT|O_DIRECTORY

Jori Koolstra is working on a solution to create and open a directory in a single, race-free call.

lwn.net
Researchhackaday

Indian scientists create brainstem map

The map is a critical step in understanding how the brainstem works.

hackaday

Custom beach robot handles gear

The robot is designed to haul beach gear, making it easier to enjoy a day at the beach.

1 story
Top pickContestCodeforces

Codeforces Round starts August 4

  • Aug 4
  • 2.2h

The contest is an ICPC contest and will start at 14:35 UTC.

6 stories
ResearcharXiv cs.DB

Fully Inductive Cardinality Estimation introduced

FICE is a graph neural network that generalizes to unseen graphs without retraining.

arxiv.org
ResearcharXiv cs.DC

FAIR-Compute allocation roadmap

The project studies allocation in a federated Digital Research Infrastructure.

arxiv.org
ResearcharXiv quant-ph

CPU+DCU framework for quantum circuit cutting

The framework reconstructs nonzero-probability states in the original output distribution.

arxiv.org
ResearcharXiv cs.DC

Cloud Continuum research infrastructure

The infrastructure combines heterogeneous Edge, Fog, Cloud, and high-performance computing resources.

arxiv.org
ResearcharXiv cs.DC

ESBT sequence CRDT for collaborative editing

ESBT provides a dense, deterministic, and compact identifier space.

arxiv.org
ResearcharXiv cs.DC

Safe quotes for retroactive liquidity pools

The quotes ensure the integrity of the pool and are computed based on virtual pools.

arxiv.org
6 stories
RepoGitHub trending · c++23,633 ★

Tencent/ncnn neural network framework

  • mobile platform

ncnn is a high-performance neural network inference framework.

RepoGitHub trending · rust27,102 ★

Pingora library for network services

  • Rust

It aims to provide a foundation for building fast and reliable network services, with a focus on evolvability.

RepoGitHub trending38,313 ★

ASP.NET Core framework

  • .NET
  • Cross-platform

It supports building web applications on Windows, Mac, or Linux, making it a versatile choice for developers.

RepoGitHub trending · c++11,599 ★

F´ flight software framework

  • C++

It is designed for building embedded systems, particularly for flight software applications, and provides a foundation for reliable and efficient systems.

RepoGitHub trending · rust66,743 ★

ripgrep search tool

  • Rust
  • Regex

It recursively searches directories while respecting gitignore, making it a useful tool for developers working with large codebases.

RepoGitHub trending48,149 ★

Chrome DevTools for coding agents

It provides a set of tools for coding agents, aiming to enhance the development experience with Chrome DevTools.

You’re all caught up.

29 stories from 223 scanned overnight.

archived edition · 223 fetched → 29 publishedj/k move · enter expand · o open · / search