What comes next.
New releases, interesting ideas, and the developments that matter to builders.
The latest
100 articlesOpenAI Ships GPT-6 Sol and Luna With API Prices Cut in Half
OpenAI's GPT-6 Sol and Luna land at $2/$10 and $0.10/$0.50 per million tokens, roughly half of GPT-5.6, arriving 90 minutes after Anthropic's Opus 5.5.
Read the news →Anthropic Launches Claude Opus 5.5, 40% Cheaper Than Opus 5
Anthropic ships Claude Opus 5.5, the first model in its Claude 5.5 family, leading agentic coding benchmarks while costing 40% less to run than Opus 5.
Read the news →Mini-AGI Trains a Continual Learning Model From Scratch on an 8GB Laptop GPU
A byte level language model that grows its own architecture, pages weights from disk instead of VRAM, and keeps learning from a single stream of data without forgetting.
Read the news →xAI Ships Grok 4.7 With Frontier Coding Scores at Half the Price
xAI's new flagship matches rival coding benchmarks at 2 and 6 dollars per million tokens, with a bigger base model trained on multi-hour tasks.
Read the news →Google Open-Sources AX, a Distributed Runtime for Agent Workloads
Google has open-sourced AX (Agent Executor), a distributed runtime that sandboxes, schedules and recovers agent tasks at cluster scale.
Read the news →OpenAI Ad Pixel Quietly Links Your Browsing to Your ChatGPT Account
A researcher says OpenAI bzr.openai.com ad collector sets an __obi cookie tied to your ChatGPT account, then receives it back from advertiser sites you visit.
Read the news →Alibaba Open-Sources Damo Radar, a Medical AI That Reads Abdominal CTs
Tested on nearly 40,000 real-world scans, Alibaba's Damo Radar vision-language model matched or beat most radiologists across 146 clinical findings.
Read the news →GPT-6 Astra Cracks a Century-Old WWI German Radio Cipher
GPT-6 Astra has decoded a WWI German ADFGVX radio message that had gone unsolved since 1918, then verified the plaintext against HMS Canterbury's original ship logs.
Read the news →US Military Nearly Boarded a Chinese Ship Over an AI Hallucination
An AI chatbot misidentified a Chinese ship's cargo as nuclear weapons components, and the US military prepared to intercept it. The false report nearly started a war.
Read the news →Microsoft Exec Called AI Scraping 'the Largest Theft of Labor in Human History'
Newly unsealed court filings in the New York Times copyright suit reveal Microsoft and OpenAI privately described their AI training data practices as theft.
Read the news →GLM Used Its Own Model to Build the Inference Stack That Serves It
Z.ai says an Infra Agent powered by GLM-5.3 built a production inference stack on 100,000 Chinese-made accelerators, tripling throughput in under two weeks
Read the news →A 4B Model Beats Postgres at Planning Its Own Queries
A researcher taught a 4B Qwen model to produce Postgres query plans via agentic reinforcement learning, cutting latency 44.7% across 113 join-heavy queries.
Read the news →