🤖 ChatGPT connects to 12,000+ banks via Plaid 🤖
Pro users access spending dashboards through 12,000+ connected financial institutions.

OpenAI has launched a preview of a personal finance experience within ChatGPT, initially available to Pro subscribers in the US. The feature lets users securely connect bank accounts, brokerages, and credit cards through Plaid, with support for Chase, Schwab, Robinhood, and over 12,000 other institutions. Once linked, users get interactive spending dashboards and can ask questions grounded in their actual financial data and goals. Intuit integration is planned as a next step. Critically, ChatGPT can analyze but cannot move money, make trades, pay bills, or file taxes — keeping a clear line between insight and action. This puts real competitive pressure on dedicated budgeting apps like Mint and YNAB.

💰 Thousands trash real Monet as AI slop 💰
AI label bias triggers replacement of genuine artistic critique online.

Conceptual artist SHL0MS sparked a viral experiment on X by presenting an authentic Claude Monet painting — a circa 1915 Water Lilies — and falsely claiming it was AI-generated, then inviting users to critique it. Thousands trash real Monet as AI slop: respondents confidently described the genuine masterpiece as "emotionless," criticizing its depth, reflections, and composition — completely unaware they were judging a real work. The experiment aligns with 2024 Norwegian research finding that people actually prefer AI art but exhibit strong negative bias when told it's machine-made. Perception labels, not actual quality, drive art judgment — a finding with real implications for how AI-generated imagery gets evaluated across creative industries and online discourse.

🧠 Gemini rolls out Extended thinking level 🧠
Google's rollout adds deeper reasoning to keep pace with rivals.

Google is gradually rolling out a new "Extended" thinking level within the Gemini app, giving users more granular control over the model's reasoning depth. The option has started appearing for select users when choosing either the Fast model or Gemini 3.1 Pro, enabling deeper chain-of-thought processing for complex queries. The move is a direct response to competitive pressure from OpenAI's o-series models and Anthropic's extended thinking in Claude. Broader availability is expected in the coming weeks. By adding this reasoning tier, Google signals that deep thinking capabilities are becoming a baseline expectation — not a premium differentiator — in the AI model landscape.

INTERESTING TO KNOW

💻 Anthropic releases Claude Code enterprise scaling guide 💻

Anthropic published an official guide detailing how Claude Code integration is working in production across large-scale codebases with thousands of developers. The resource addresses enterprise-specific challenges including monorepos with millions of lines of code, legacy systems built over decades, and microservice architectures spanning multiple repositories. The practical patterns shared signal that Claude Code is gaining real enterprise traction alongside GitHub Copilot in the AI developer tools space.

🖥️ Codex eyes remote macOS control via Computer Use 🖥️

OpenAI is developing remote access capabilities for Codex that would let it operate macOS applications via Computer Use even when a laptop is locked or asleep. Currently, Computer Use requires an active, unlocked session to see the screen, move the cursor, and type. Removing this restriction represents a capability jump toward truly autonomous desktop agents — users could direct Codex remotely without physical access to their machines. No release timeline has been confirmed.