AllLLMRoboticsStartupsResearchEthicsBusinessHardwareAutonomousMedicineCreative

All news

10600
Thomson Reuters engineer runs an offline coding agent on Qwen3.5 9B in the Himalayas
LLM·Latent Space (swyx)

Thomson Reuters engineer runs an offline coding agent on Qwen3.5 9B in the Himalayas

Oct 28, 2026·3 min
NVIDIA Kumo Relational Beats Supervised ML on 41 Tasks Without Any Training
Business

NVIDIA Kumo Relational Beats Supervised ML on 41 Tasks Without Any Training

Oct 28, 2026·3 min
Moonshot releases Kimi K3.1 Pro: 2.8 trillion parameters, weights due November 4
Business

Moonshot releases Kimi K3.1 Pro: 2.8 trillion parameters, weights due November 4

Oct 28, 2026·3 min
Switzerland's Apertus 1.5 Gains Vision, Hearing and a 262K-Token Context Window
LLM

Switzerland's Apertus 1.5 Gains Vision, Hearing and a 262K-Token Context Window

Oct 28, 2026·3 min
NVIDIA Releases 120B Nemotron 3.5 Super VL and Bolt Models for Local Agents
Business·NVIDIA AI Blog

NVIDIA Releases 120B Nemotron 3.5 Super VL and Bolt Models for Local Agents

NVIDIA opened Nemotron 3.5 Super VL, a 120B-parameter model with a 5x speed gain, plus two Bolt models. Bolt 74B-A9B fits in 24 GB thanks to 2-bit quantization.

Oct 28, 2026·3 min
Goodfire Researcher Claims Neural Networks Are Far Simpler Than Interpretability Experts Thought
Research·Latent Space (swyx)

Goodfire Researcher Claims Neural Networks Are Far Simpler Than Interpretability Experts Thought

Nick Cammarata of Goodfire argues that most neurons in trillion-parameter models are actually readable by humans. The famous examples of "tangled" neurons, he says…

Oct 28, 2026·3 min
aiedit hides individual lines of code from an AI agent: 5 levels and 857 tests
LLM·Habr AI

aiedit hides individual lines of code from an AI agent: 5 levels and 857 tests

A Python beginner built the aiedit editor with Claude: the model sees the project but not the locked lines. Code, not a prompt, enforces it: 11K lines, 857 tests…

Oct 28, 2026·3 min
vps-forge: Agents With Root on a VPS Get Auto-Rollback, Self-Bans Drop From 7 in 60 Runs to 0
LLM·Habr AI

vps-forge: Agents With Root on a VPS Get Auto-Rollback, Self-Bans Drop From 7 in 60 Runs to 0

The open-source Agent Skill vps-forge gives agents a 5-minute dead man's switch. Across 120 benchmark runs, Opus 5.5 scores rose from 79–92 to 99–100, but each run…

Oct 28, 2026·4 min
Top themes
01BigSleep LLM Agent Finds 45 V8 Vulnerabilities, Fuzzilli 37, With 25% OverlapOct 28, 2026·3 min02rwkv-quant: keeping one matrix in bf16 removes 39% of RWKV-7 2.9B's quantization lossOct 28, 2026·3 min03OpenAI Releases GPT-6.1 Terra: Half the Price, Twice the Accuracy in CodeOct 28, 2026·3 min04Qwen3-8B on the Intel Core Ultra 5 226V NPU: 10.7 tok/s at 15 W and 24 dB of NoiseOct 28, 2026·3 min05AWS Launches Multi-Turn RL in SageMaker: gpt-oss-20b Climbs From 15% to 69% for $339Oct 28, 2026·4 min
Microsoft and Hugging Face trained Qwen3.8-27B on database-state rewards: from 38 to 107 tasks
Business·Hugging Face Blog

Microsoft and Hugging Face trained Qwen3.8-27B on database-state rewards: from 38 to 107 tasks

Oct 28, 2026·3 min
Gemma 26B fine-tuned on a 24 GB MacBook: offline assistant gets 703 of 766 right
Hardware·Habr AI

Gemma 26B fine-tuned on a 24 GB MacBook: offline assistant gets 703 of 766 right

Oct 28, 2026·4 min
Cognition releases FrontierCode 2.0: the best score is just 45.5%
Business·Cognition Blog

Cognition releases FrontierCode 2.0: the best score is just 45.5%

Oct 28, 2026·3 min
Dharma-AI trimmed Qwen3-4B's vocabulary and cut inference cost by 77%
Business·Hugging Face Blog

Dharma-AI trimmed Qwen3-4B's vocabulary and cut inference cost by 77%

Oct 28, 2026·3 min
Codex on GPT-6.1 Sol trims its window to 828 400 but still accepted an 888-thousand-token input
LLM·Habr AI

Codex on GPT-6.1 Sol trims its window to 828 400 but still accepted an 888-thousand-token input

The config says model_context_window = 1 000 000, yet Codex reports 828 400. A 890 547-token test shows it is a history threshold, not a limit on a single input.

Oct 28, 2026·3 min
12 Pentesting Tools vs. Juice Shop: The Best AI Agent Found 5 Holes Out of 116
Business·Habr AI

12 Pentesting Tools vs. Juice Shop: The Best AI Agent Found 5 Holes Out of 116

The author ran 6 AI agents, 4 DAST scanners and a Claude + Burp MCP combo against OWASP Juice Shop's 116 vulnerabilities. The hybrid won with 9/10; headline…

Oct 28, 2026·3 min
Google Releases Gemini 4 Argon: $2 per Million Tokens and First Place on the Vals Index
Business·DeepMind Blog

Google Releases Gemini 4 Argon: $2 per Million Tokens and First Place on the Vals Index

Gemini 4 Argon is live in the API and Gemini app: $2/$10 per million tokens, 95% off cached input, #1 on the Vals Index. Chrome, Gemma and quantum kernels got faster with it.

Oct 28, 2026·3 min
Reflection Releases Beam: An Open 501B-Parameter MoE Model Under Apache 2.0
LLM·Hugging Face Blog

Reflection Releases Beam: An Open 501B-Parameter MoE Model Under Apache 2.0

Oct 28, 2026·4 min
Meta open-sources Muse Spark Open: an 816B-parameter MoE with a 1M-token context
LLM·Hugging Face Blog

Meta open-sources Muse Spark Open: an 816B-parameter MoE with a 1M-token context

Oct 28, 2026·3 min
Modal Raises $750M at a $15.75B Valuation and Reaches $1B in ARR
Business·Latent Space (swyx)

Modal Raises $750M at a $15.75B Valuation and Reaches $1B in ARR

Oct 28, 2026·2 min
Property Finder Found a Flaw in Its Labels and Cut VLM Costs by 86%
Business·AWS Machine Learning Blog

Property Finder Found a Flaw in Its Labels and Cut VLM Costs by 86%

Oct 28, 2026·4 min
DeepSeek Open-Sources DeepInfer as the US and China Prepare Controls on Open Models
Business·Interconnects (Nathan Lambert)

DeepSeek Open-Sources DeepInfer as the US and China Prepare Controls on Open Models

Oct 28, 2026·3 min
Qwen 3.8 Routed Requests to an Alibaba Cloud Proxy, With Up to 4% Failures in Tests
Business

Qwen 3.8 Routed Requests to an Alibaba Cloud Proxy, With Up to 4% Failures in Tests

Oct 28, 2026·3 min
Can You Teach a Language Model While It's Reading? Researchers Tested Every Architecture
Research

Can You Teach a Language Model While It's Reading? Researchers Tested Every Architecture

Oct 28, 2026·3 min
OpenAI releases GPT-6.1 Terra: half the price and nearly on par with its senior models
Business

OpenAI releases GPT-6.1 Terra: half the price and nearly on par with its senior models

Oct 28, 2026·3 min
OpenAI Releases GPT-6.1 Terra: $1 per Million Input Tokens and Three Times Faster Than Astra
Business·TG: @neyr0graph (NeuroGraph)

OpenAI Releases GPT-6.1 Terra: $1 per Million Input Tokens and Three Times Faster Than Astra

OpenAI unveils GPT-6.1 Terra, a mid-tier model for agents that can see the screen: $1 per million input tokens, $5 per million output, and three times faster than…

Oct 28, 2026·3 min
2026 Open-Source Models Claim 1M Context but Read It Selectively
LLM·Habr: Сбер

2026 Open-Source Models Claim 1M Context but Read It Selectively

GLM-5.3 picks 2048 tokens from history, DeepSeek-V4-Pro holds 1024 compressed blocks, and Qwen3.8 and Kimi K3 use full attention in only every fourth layer. What…

Oct 28, 2026·3 min