What comes next.
New releases, interesting ideas, and the developments that matter to builders.
The latest
100 articlesAnthropic Puts Your Bank Balances Inside Claude With Personal Finance Beta
Anthropic is rolling out Personal Finance in Claude, a US beta that links bank and card accounts through Plaid so Claude can answer money questions with read-only access.
Read the news →AI Selloff Wipes $3 Trillion Off Wall Street as Nasdaq Enters Correction
The Nasdaq closed 5.5% lower in its worst day since April 2025 as chip and data center stocks slid on fears that Democratic midterm wins will slow the AI buildout.
Read the news →OpenAI Launches GPT-6 Cyber After Model Patches 948 Linux Kernel Bugs
OpenAI released GPT-6 Cyber for security research, saying the model powered a Linux kernel run that validated and patched 948 privilege-escalation bugs for about USD 1.2 million in API cost.
Read the news →Moonshot Opens Kimi K3.1 Pro, a 2.8T Open Model That Matches Claude on Coding
Moonshot released the weights of Kimi K3.1 Pro, a 2.8T parameter open model with a 1M token context that trades blows with Claude Opus 5.5 and GPT-6 Astra on coding benchmarks.
Read the news →OpenAI to Acquire Pierre Computer to Take Source Control Into Codex
OpenAI says it will buy Pierre Computer, the startup behind Code Storage and the Trees and Diffs libraries already embedded in the Codex app, as it pushes to make Codex a shared home for team software work.
Read the news →Revolut Moves 2,000 Engineers Off Claude Code and Codex to Self-Hosted GLM
Revolut runs about 2,000 engineers on self-hosted GLM across 56 Nvidia B200s, at roughly a tenth of its previous Anthropic and OpenAI coding bill.
Read the news →Anthropic Posts USD 17B Q3 Revenue and Warns It May Be Nearing Its Own RSI Threshold
Anthropic's amended IPO filing shows Q3 revenue of about USD 17 billion, a swing back to a USD 567 million operating loss, roughly USD 895 billion in compute commitments, and new language warning it may imminently cross its own RSI safety threshold.
Read the news →MIT CSAIL: GPT-6 Astra Fixes Qwen's Math by Editing 0.016% of Its Weights
MIT CSAIL researchers report that GPT-6 Astra lifted Qwen3.5-0.8B's MATH-500 score from 7.9% to 48.1% by directly editing 0.016% of its weights, with no gradients or training.
Read the news →Mistral Opens the Weights of Le Chonk, Its 1T Parameter Flagship
Mistral has published the open weights for Mistral Large 4, a 1-trillion-parameter multimodal model it says leads open-weight coding and cyberdefense benchmarks.
Read the news →Anthropic Blocks Users Whose Work Touches Model Training, Researcher Says
A researcher says Anthropic flagged his paid account for training a competitor and now blocks anything touching RFT, SFT or larger-model training. Anthropic has not responded.
Read the news →YC Puts 50 AI Agents in Minecraft: Ration Ledgers, Permits, and a Colony That Never Thrived
YC Labs put 50 GPT-6.1 Sol agents in survival Minecraft for three hours. The colony built ration ledgers and permits, never thrived, and the agent listed first in the directory ended up running the show.
Read the news →A Single Shared ChatGPT Link Could Hijack Any Signed-In Account
Parameter researchers turned an ordinary public ChatGPT share into full session hijacking. OpenAI rated the bug P1 and fixed it on October 26.
Read the news →