The week in AI

Nvidia's SoL-Pi system cuts coding agent token usage nearly in half by optimizing the harness

SoL-Pi cuts coding agents' token usage by up to 49 percent with little change in performance by optimizing the control layer between the model and its environment. A research agent tested 152 approaches across more than 3,000 runs to develop the system, though the gains were smaller on other benchmarks. The article Nvidia's SoL-Pi system cuts coding agent token usage nearly in half by optimizing the harness appeared first on The Decoder.

OpenAI's GPT-6 Astra can now tell you exactly where you screwed up your IKEA shelf

OpenAI's GPT-6 Astra can look at a photo and tell whether an IKEA furniture piece was assembled incorrectly, hitting an 80 percent accuracy rate. Back in November 2025, the best model managed just 28 percent. According to Epoch AI, the speed isn't quite fast enough yet for real-time assembly guidance, but the gap is closing quickly. The article OpenAI's GPT-6 Astra can now tell you exactly where you screwed up your IKEA shelf appeared first on The Decoder.

OpenAI pauses its "most capable models" after agents exploit loopholes and leak data

OpenAI has shared new details from its ongoing AI safety investigation. One research model exploited a DNS loophole to reach the internet from a locked-down environment, while another deliberately leaked a GitHub token and twice ignored a researcher's direct instructions. OpenAI has paused tool-based training, evaluation, and inference for its most capable models. With government and university sites among those affected, the question of who's liable when AI agents hack is getting harder to igno

Proaction boosts sales 60% and saves 75+ hours with Codex

With Codex, GPT-Live-1, and GPT-6 Astra, Proaction builds, operates, and sells modern fleet management faster.

Pentagon was right to slap Anthropic with a security supply chain risk label, federal court says

A federal appeals court has upheld the Pentagon's decision to bar Anthropic from military contracts. Defense Secretary Hegseth argues the company's safety restrictions could jeopardize military operations. Anthropic says the designation has already cost it billions. The article Pentagon was right to slap Anthropic with a security supply chain risk label, federal court says appeared first on The Decoder.

Another Google Deepmind researcher quits, says building superintelligent AI soon is "inherently irresponsible"

Google Deepmind researcher Robert O'Callahan has quit, saying AI's "current rate of change is far too high." He worked on chip design tools that helped make AI cheaper and faster, a contribution he can no longer justify. Many colleagues share his concerns but rarely speak out, he says. The article Another Google Deepmind researcher quits, says building superintelligent AI soon is "inherently irresponsible" appeared first on The Decoder.

Microsoft gives Copilot another makeover, adding an Autopilot agent and usage-based billing

Microsoft is splitting its Copilot app into three sections: Home, Code, and a new agent called "Autopilot." Built on OpenClaw, the agent runs continuously in the cloud, where it can monitor Teams channels and complete tasks on its own, according to Microsoft. For Autopilot and Code, the company is also switching to usage-based billing instead of flat-rate pricing, moving further away from its AI subsidy model. The article Microsoft gives Copilot another makeover, adding an Autopilot agent and us

Batching by Length Instead of Looping Item by Item for SLM Optimization

We finish off our short series on SLM optimization with the third entry, focused on batching by length instead of looping item by item.

Meta's Muse agent gives every user a full cloud computer running Ubuntu Linux

Meta gives every Muse user a free cloud computer running Ubuntu Linux where they can install software, write code, and browse the web. A "Sentinel" process monitors sensitive actions outside the user's workspace, while users can inspect every file in the system. With over 500,000 users in its first week, Meta is betting on product reach over model power. The article Meta's Muse agent gives every user a full cloud computer running Ubuntu Linux appeared first on The Decoder.

Intelligence doesn't come cheap as AI drives up costs for the NSA, hospitals, and insurers

The NSA is already spending billions of dollars to test advanced AI models, mostly on computing power, according to The Washington Sun. Lawmakers expect full-scale AI oversight to cost tens of billions of dollars a year, while earlier estimates from the Congressional Budget Office put the figure at just $20 million. In US health care, AI-assisted billing codes drove up costs by nearly $1 billion over two years. The article Intelligence doesn't come cheap as AI drives up costs for the NSA, hospit

7 Advanced Python Tricks to Level Up Your Coding Skills

Leveling up rarely means new syntax. It means learning what the language already promised you.

Google's "Call for Me" lets Gemini phone businesses for you

Google is testing "Call for Me," a feature that lets Gemini call businesses on a user's behalf. The article Google's "Call for Me" lets Gemini phone businesses for you appeared first on The Decoder.

Anthropic signs $11.6 billion cloud deal with Akamai, pushing its compute spending past $500 billion in under a year

Anthropic has reportedly signed a seven-year, $11.6 billion cloud deal with Akamai Technologies and will receive a warrant for up to 5 percent of Akamai's shares. Its compute deals have reportedly totaled $517 billion in 11 months. CEO Dario Amodei has warned that Anthropic could go bankrupt if its revenue forecasts are even slightly off. The article Anthropic signs $11.6 billion cloud deal with Akamai, pushing its compute spending past $500 billion in under a year appeared first on The Decoder.

Ruby on Rails creator DHH says he's done writing code by hand

David Heinemeier Hansson, creator of Ruby on Rails, has quit writing code by hand after 25 years. He says he hasn't typed a single line since March 2026. "English is a better programming language than Ruby," DHH told the audience at Rails World 2026. Just a year ago, he was pushing back against AI coding. Now he calls it "the biggest thing that has happened in the history of computing." The article Ruby on Rails creator DHH says he's done writing code by hand appeared first on The Decoder.

The Pentagon wants $30 million to build an AI-powered lie detector

The US government wants to spend $30.3 million over the next five years on an improved form of lie detector, according to a Department of Defense budget request. The program, called “Polygraph+” or “Polygraph Next”, will focus on scoring algorithms that use artificial intelligence and machine learning and on a technique called “standoff sensing”, which…

White House tells OpenAI and Anthropic to let U.S. review new models before sharing them with British testers

The White House wants OpenAI and Anthropic to hold back new AI models from the U.K.'s AI Safety Institute until U.S. agencies get to review them first. The article White House tells OpenAI and Anthropic to let U.S. review new models before sharing them with British testers appeared first on The Decoder.

Top AI experts badly underestimated how fast the field is moving, study finds

Leading AI experts have consistently underestimated how fast AI is advancing, according to the Forecasting Research Institute. AI reached gold-medal level at the International Mathematical Olympiad five years ahead of the median expert forecast, and Anthropic's annualized revenue is about five times what experts predicted. But forecasts for real-world uses like self-driving cars paint a more mixed picture. The article Top AI experts badly underestimated how fast the field is moving, study finds

Sakana AI hires Jürgen Schmidhuber, inventor of deep learning, world models, and your next ChatGPT update

Tokyo-based Sakana AI has hired Jürgen Schmidhuber as Chief Scientific Advisor. Sakana calls him the "father of modern AI." He'll help lead the company's new RSI Lab, which works on recursive self-improvement, meaning AI that keeps developing itself. His ideas from the 1990s have already shaped Sakana projects like the Darwin Gödel Machine. The article Sakana AI hires Jürgen Schmidhuber, inventor of deep learning, world models, and your next ChatGPT update appeared first on The Decoder.

Google's Suncatcher project aims to put AI data centers in orbit powered by solar energy

Google's "Suncatcher" project aims to run AI infrastructure in orbit on solar power. A fridge-sized experimental satellite is set to launch on a SpaceX Falcon 9 on October 1. But the challenges are steep: you'd need around 10,000 satellites to match a single 1-gigawatt data center on Earth, and Jeff Bezos thinks it could take 20 years before orbital data centers beat ground-based ones on cost. The article Google's Suncatcher project aims to put AI data centers in orbit powered by solar energy ap

Black Forest Labs launches FLUX 3 Action, an open robotics AI model

Black Forest Labs is entering robotics with FLUX 3 Action. The open-world-action model uses camera feeds to predict what action a robot should take next. With just seven billion parameters, it sets a record on the RoboLab-120 benchmark while running up to 3.95 times faster than the previous top model. The article Black Forest Labs launches FLUX 3 Action, an open robotics AI model appeared first on The Decoder.

Introducing Gemini 3.8 Live with Live Avatar

Anthropic says Claude discovered a new enzyme system, but CRISPR researchers call it routine genome mining

Anthropic's AI model Claude found a previously unknown enzyme system in DNA databases, doing most of the analysis on its own. The article Anthropic says Claude discovered a new enzyme system, but CRISPR researchers call it routine genome mining appeared first on The Decoder.

AI performance costs are falling faster than those of any previous technology

AI is hitting a fixed benchmark performance level at a rapidly falling cost. Epoch AI measures a price decline of about 13x per year. After stripping out hardware gains and competition, MIT puts annual algorithmic progress at about 3x. That doesn't mean today's best models are cheaper, though. Reasoning models, for example, can cost more because they use far more compute per task. When picking a model for real-world use, quality, speed, and error rate matter just as much as price. The article AI

Accelerating vision-language models with LFM2.5-VL-DSpark

MCP Explained in 5 Minutes

A visual guide to MCP that explains how it works, how to use it with Claude Code, Tavily, GitHub, and Playwright, and what is new through simple diagrams that make the whole concept easy for anyone to understand.

What I’ve Learned About DeepSeek Harness

KDnuggets team member Shittu Olumide tested out DeepSeek Harness. Here's what he found.

How to Use NVIDIA Warp and MjWarp to Accelerate Robotics Simulation and Learning Workflows

Google Beam expands with new regions, partners, and customers

We’re expanding Google Beam to five new countries, and partnering with Industrious for an extended network.

ChatGPT Voice gets closer to "Her" with email, calendar, and Slack access

ChatGPT Voice now runs on OpenAI's new GPT-6 Astra, Sol, and Luna models and can tap into plugins like email, calendar, and Slack. Users can manage appointments, send emails, or build websites just by talking. The update moves OpenAI closer to the everyday AI assistant Sam Altman has long compared to the one in the sci-fi film "Her." The article ChatGPT Voice gets closer to "Her" with email, calendar, and Slack access appeared first on The Decoder.

Google's new Flash TTS models let you design AI voices from scratch using text descriptions

Google is introducing two new text-to-speech models, Gemini 3.8 Flash TTS and Flash-Lite TTS, which support more than 100 languages. Flash TTS can create new voices from text descriptions, and both models let users add stage directions to individual lines and generate two-voice dialogue from a single script. A voice cloning feature can build a voice profile from a 30-second sample. The article Google's new Flash TTS models let you design AI voices from scratch using text descriptions appeared fi

YouTube adds AI tools to Creator Studio with script coaching, smart thumbnails, and Gemini editing

YouTube is adding AI tools to its creator studio. A storytelling assistant analyzes scripts and rough cuts, Gemini becomes a chat-based editing assistant for Shorts, and a new live translation feature turns English streams into Spanish. The article YouTube adds AI tools to Creator Studio with script coaching, smart thumbnails, and Gemini editing appeared first on The Decoder.

Advancing Private AI Compute with secure, server-side memory

Introducing private, server-side memory to Private AI Compute for personal AI.

Two years of OpenAI Academy

Marking two years of OpenAI Academy and bringing AI skills to even more communities.

Gemini 3.8 text-to-speech says hello

Anthropic engineer explains why Claude's writing got worse although the model got smarter

Anthropic employee Jackson Kernion explains why newer Claude models write so oddly. Optimizing for math, code, and technical explanations aimed at other AI models has created a style that sounds like "overly-dense info dumps" to humans. Opus 5.5 tries to fix this, but Opus 4.6 remains unmatched as a pure writing model. The article Anthropic engineer explains why Claude's writing got worse although the model got smarter appeared first on The Decoder.

Everything Claude Opus 5.5 Actually Ships With

This article pulls together every verifiable number and detail from Anthropic's announcement, the platform documentation, the system card, and independent coverage, so you have one place to check the facts.

Nvidia-backed Nscale keeps its biggest customer, Bytedance, out of its IPO filing

Nscale, the Nvidia-backed AI cloud provider, leaves its most important customer, Bytedance, out of the main prospectus for its planned US IPO. The article Nvidia-backed Nscale keeps its biggest customer, Bytedance, out of its IPO filing appeared first on The Decoder.

Meta's AI agent Muse draws 500,000 users in a week along with claims it copied OpenClaw

Meta's AI agent Muse picked up more than 500,000 users in its first week and hit number one in Apple's App Store. But Meta admits the product is "heavily inspired" by the open-source project OpenClaw, and some of the file names and contents are nearly identical. OpenAI is already discussing a response of its own. The article Meta's AI agent Muse draws 500,000 users in a week along with claims it copied OpenClaw appeared first on The Decoder.

Why Most Data Science Notebooks Die After Day One: How to Build Ones That Survive

Six habits that keep a notebook runnable after you close the laptop.

Inside Basecamp Research, the AI startup turning evolution into training data

Basecamp Research has raised $140 million from investors including Nvidia and Anthropic's Anthology Fund. The London company trains AI models on genetic material from rainforests, oceans, and hot springs to design antibiotics and tools for cell therapies. In an interview with THE DECODER, CTO Philip Lorenz explains why biology is a far bigger problem for AI than language, and why good scores on paper don't guarantee good molecules. The article Inside Basecamp Research, the AI startup turning evo

OpenAI extends cyber access to Ukraine for civilian defense

OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure.

Alibaba launches Qwen Audio 3.1 with new models and slashes AI audio prices by up to 95 percent

Alibaba's AI team Qwen has released Qwen-Audio-3.1, a lineup of five models for speech recognition (ASR), text-to-speech (TTS), and real-time interaction. The ASR model improves multilingual and dialect recognition and automatically cleans up filler words and repetitions. ASR-Next adds multi-speaker identification with timestamps and detects emotions, ambient sounds, and machine noise. TTS handles multilingual synthesis […] The article Alibaba launches Qwen Audio 3.1 with new models and slashes

High-Performance Data Processing with Polars: A KDnuggets Cheat Sheet

Polars is a DataFrame library written in Rust on the Apache Arrow memory format, and the speed comes less from the language than from the model. The model? Describe your work as expressions, and the Polars query engine plans them out.

Sam Altman’s remarks at the United Nations Security Council

OpenAI CEO Sam Altman discusses AI safety, human control, and international cooperation in remarks to the United Nations Security Council.

Harvey turns legal context into stronger drafts with GPT-6 Astra

GPT-6 Astra produces more structured, context-aware legal documents, freeing lawyers to focus on strategy.

How invideo improves color grading 3x with GPT‑6 Astra

With GPT‑6 Astra, invideo plans edits with greater precision, improves color correction and grading threefold, and produces 50 custom effects in one day.

Ringg’s AI agents resolve up to 65% of customer calls with OpenAI

Using GPT-5.6, Ringg powers multilingual agents across voice, chat, WhatsApp, and web for 90% less cost vs. GPT-4.1.

OpenAI hires Patreon co-founder Sam Yam to lead a new Creator Product division

OpenAI has hired Patreon co-founder Sam Yam to lead a new "Creator Product" division. After more than 13 years at the creator platform, he's bringing two Patreon executives with him. The team will build new tools for creators, and a first look could come at OpenAI DevDay next week. The article OpenAI hires Patreon co-founder Sam Yam to lead a new Creator Product division appeared first on The Decoder.

Introducing MentalHealthBench

MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses across realistic mental health conversations.

The AI Hype Index: AI loves cheating

Brace yourself: It turns out AI is being optimized for cheating. OpenAI’s agents hacked into Hugging Face to get the answers to a cybersecurity test. Next, they solved a prestigious math problem (or just stole from two top mathematicians’ answer sheets). Anthropic’s models have also hacked into other companies’ systems four times already. And that’s…

ChatGPT Ads expands to Southeast Asia and Taiwan

ChatGPT Ads is expanding to Southeast Asia and Taiwan, giving eligible businesses new ways to reach people across more than 60 countries.

Airbnb widens access to GPT-6 Astra and OpenAI frontier models

Learn how Airbnb is expanding access to GPT-6 Astra and OpenAI frontier models to help engineering teams solve bugs, design systems, and ship faster.

Grab and OpenAI bring practical AI skills to Southeast Asia

OpenAI and Grab launch GO Forward with AI, a regional programme helping 30,000 partners build practical AI skills across Southeast Asia.

Better prompt caching for GPT-6

Learn how GPT-6 improves prompt caching with higher cache hit rates, new diagnostics, explicit breakpoints, and controls that reduce latency and costs.

OpenAI's GPT-6 Sol and Luna cut prices in half but barely move the needle on performance

With GPT-6 Sol and Luna, OpenAI adds two cheaper models that deliver their predecessors' performance at half the token price and take aim at Anthropic's pricier offerings. Independent analyses find little gain in actual intelligence, though, and OpenAI likely didn't see Anthropic's simultaneous launch of Opus 5.5 coming. The article OpenAI's GPT-6 Sol and Luna cut prices in half but barely move the needle on performance appeared first on The Decoder.

Introducing GPT-6 Sol and Luna

Meet GPT-6 Sol and Luna, two models that bring frontier intelligence to everyday work with different balances of capability and cost.

Bravely AI Browsing with Leo

Learn about private AI browsing for data professionals.

Roundtables: The Deadly Failures of The Virtual Border Wall

The US has spent billions building a “virtual wall” of surveillance towers along its southern border over the past 25 years, promising they will help detect and apprehend border crossers and save lives. But a groundbreaking investigation by MIT Technology Review has documented over a thousand people who moved through areas watched by these towers…

7 Open-Source Alternatives to ChatGPT You Can Run Locally

Explore seven open-source ChatGPT alternatives, from lightweight local chat interfaces and document assistants to agent platforms, multi-user setups, and complete self-hosted AI workspaces.

Parallel cut research time and cost in half with GPT‑6 Astra

GPT‑6 Astra allowed Parallel’s agents to research and synthesize labor-market data in half the time and at half the cost vs. prior models.