2026

LLM·TNW

Critical Energy raises $22 million for geothermal turbines to power AI

Arcade.dev raises $60M to tackle the problem of unchecked AI agent permissions

LangSmith learns to compare fine-tuned open-source LLMs with automatic scoring

Google restricts Meta's access to Gemini due to a shortage of computing capacity

Google was unable to provide Meta with sufficient computing capacity and has imposed restrictions on its access to the Gemini AI models, acc

01Moral crumple zone: how developers became the liability buffer for AI systemsJun 29, 2026·2 min 02AMD and Intel finalize ACE instructions — AI accelerators to enter future x86 processorsJun 29, 2026·2 min 03Z.ai releases GLM-5.2 on Huawei chips — the model takes first place in the global rankingJun 29, 2026·2 min 04Anthropic and the White House negotiate AI rules at G7 — while the company's models remain bannedJun 29, 2026·2 min 05AWS shows how data mesh becomes the foundation for agentic AI in productionJun 29, 2026·2 min

Top themes

LLM·AI News

Computer vision on the shelves: Coresight Research calculates retail losses from shelving errors

A joint study by Coresight Research, Simbe and RELEX Solutions puts a figure, for the first time, on how much retail loses to shelving error

01Google DeepMind publishes roadmap for guarding against its own AI agentsJun 29, 2026·2 min 02Prosus launches ToqanClaw, a no-code business app builder for five million merchantsJun 29, 2026·2 min 03e2e-assure launches Cumulo — the UK's first sovereign AI-powered SOC platform for IT and OTJun 29, 2026·2 min

LLM·TNW

Tsuga raises $35 million for AI agent observability without per-byte billing

A Paris startup founded by Datadog alumni raised a $35 million Series A and wants to break the per-byte billing model in an era of explosive

LLM·AI News

SAP and Google Cloud roll out a multi-agent platform for retail automation

SAP and Google Cloud have joined forces to build an architecture with AI agents capable of autonomously managing marketing and retail operat

LLM·Hugging Face Blog

Hugging Face compares all LoRA alternatives: who wins at fine-tuning LLMs

The Hugging Face team tested five PEFT methods for fine-tuning LLMs — from DoRA to GaLore — and found out when LoRA can be beaten, and exactly what it costs.

LLM·3DNews AI

Claude Code found a bug in AMDGPU that has been freezing Linux laptop displays since 2017

LLM·Hugging Face Blog

ServiceNow: AI agent leaks corporate secrets through a chain of search queries

LLM·Hugging Face Blog

PaddleOCR releases PP-OCRv6: text recognition in 50 languages from 1.5 to 34.5 million parameters

Baseten raises $1.5 billion at a $13 billion valuation — a bet on cheap AI models

Abu Dhabi's MGX fund raises about $50 billion to accelerate AI investment

AI tokens are driving up corporate cloud bills — ZDNet flags the trend

Alphabet launches incubator for AI startups founded by former Google employees

Russian government approves AI bill with amendments from business

Oracle laid off 21,000 employees in a year — AI replaced some positions

Microsoft and Chevron Sign 20-Year Power Supply Deal for a Data Center in Texas

A Roadmap for Evaluating AI Agents: Metrics, Benchmarks, and Practical Methods

LLM·NVIDIA Developer Blog

NVIDIA Introduces Halos — a Functional Safety System for AI-Powered Robots

NVIDIA Halos is a full functional-safety stack for physical AI: robots that work alongside humans in factories, warehouses, and hospitals.

01California Gas Stations Accused of Using AI to Inflate Fuel PricesJun 29, 2026·2 min 02NVIDIA DAQIRI: Real-Time AI Inference for High-Speed Data AcquisitionJun 29, 2026·2 min 03TimeCopilot: How to Build a Forecasting Pipeline with Foundation Models and Anomaly DetectionJun 29, 2026·2 min 04GitLab 19.1: AI workflows now trigger automatically on events, with no human involvementJun 29, 2026·2 min 05Cisco open-sources FAPO: automatic LLM pipeline optimization with step-by-step diagnosticsJun 29, 2026·3 min

Top themes

SoftBank's Masayoshi Son: Musk's orbital data centers won't win the AI race

SoftBank founder Masayoshi Son has said Elon Musk's idea of putting data centers in orbit makes no sense: the AI race will be won by ground-

01Baseten raises $1.5 billion at a $13 billion valuation — Blackbird makes its biggest bet everJun 29, 2026·2 min 02MoonMath AI releases open-source HIP kernel for AMD MI300X that outperforms official AITER v3Jun 29, 2026·2 min 03xAI launches /goal in Grok Build: an autonomous agent plans and verifies multi-step tasksJun 29, 2026·2 min

LLM·3DNews AI

OpenAI launches Patch the Planet to protect open-source software with HackerOne and Trail of Bits

OpenAI, together with Trail of Bits, HackerOne and Calif, has launched Patch the Planet — a program of personalized support for open-source

LLM·MIT News

MIT creates a chip that lets tiny robots build 3D maps on the fly

MIT's specialized chip lets tiny robots build 3D maps of their surroundings directly onboard, with no need for external servers and minimal

A neural network for forecasting GDP in C++ without TensorFlow: runs on a 400-ruble Arduino

A developer wrote a neural network for forecasting GDP dynamics in C++20 with no frameworks — the model is trained using CUDA and runs even on an Arduino with 32 KB of memory.

The illusion of control: why prompts don't protect AI agents from capability chaining

Google restricts Meta's access to Gemini amid a compute capacity shortage

Stripe explains how AI agents help scale financial compliance

How Cara Creates Industry AI for Insurance Brokers in Partnership with AWS

LLM·TNW

BIS warns: an AI investment crash could hit credit markets like 2008

7 Best Coding Models for Local Deployment in 2026: Qwen, DeepSeek, and More

Qwen, Gemma, Phi-4: five open-source omni-models for every data type

Teamly AI Assistant: Chaotic Knowledge Base to Fast Onboarding

Teamly has updated its knowledge management platform: the AI assistant now runs on up-to-date internal data and helps new employees get up to speed on processes in days, not weeks.

VisionLabs found a way to train CV models with 50 images and no coding