← Back to Payloads
ai2026-05-19

Gemini Extended Thinking , ChatGPT finance , Claude Code at

Google rolls out a Standard/Extended thinking-level toggle for Gemini 3 Flash and 3.1 Pro. OpenAI plugs Plaid into ChatGPT, giving 12,000+ institutions read-only access for personal-finance guidance. Anthropic reframes Claude Code as a self-improving engineering organization at the Code with Claude 2026 keynote.
Quick Access
Install command
$ mrt install ai
Browse related skills
Gemini Extended Thinking , ChatGPT finance , Claude Code at

Gemini Extended Thinking , ChatGPT finance , Claude Code at

Three different labs shipped three different "model + product surface" plays within 48 hours of each other. Google added a new thinking-level control inside the Gemini app. OpenAI plugged Plaid into ChatGPT. And Anthropic used its annual Code with Claude developer conference to make the case that Claude Code is now a self-improving engineering organization, not an assistant.

What You Need to Know: Google is rolling out a "Thinking level" option (Standard / Extended) inside the Gemini app for Fast (Gemini 3 Flash) and Gemini 3.1 Pro selections, ahead of I/O 2026. OpenAI partnered with Plaid to let ChatGPT Pro users in the US connect bank accounts from 12,000+ institutions for personalized financial guidance. And Anthropic used Code with Claude 2026 (May 17 keynote) to reframe Claude Code as an "engineering organization that teams can orchestrate and trust at scale."

Why It Matters

  • The thinking-level toggle is the new product surface for reasoning models. Anthropic, OpenAI, and Google now all expose a user-facing control that trades latency for depth. Builders wiring any of these into an app need to expose the same control to their end users or watch usage metrics split.
  • ChatGPT is now a personal-finance dashboard that reads your accounts. With Plaid integration, ChatGPT Pro users can ask the model about spending, subscriptions, and upcoming payments. This is OpenAI's first major move into sensitive-vertical data custody (with Intuit "coming soon"), and it sets a precedent for what a chatbot's permissions look like when money is on the line.
  • Claude Code is being repositioned as the runtime, not the assistant. The keynote framing — "a reliable, self-improving engineering organization that teams can orchestrate and trust at scale" — is the strategic signal that Anthropic is going after the orchestration layer, not the model.

What Actually Happened

Gemini app rolls out "Extended" thinking level

Ahead of Google I/O 2026, Google began rolling out a "Thinking level" option inside the Gemini app's model picker, per 9to5Google. The new menu only appears when the user selects either Fast (Gemini 3 Flash) or Gemini 3.1 Pro — the "Thinking" model selection is excluded, since it already runs at maximum depth. The two new options are "Standard" and "Extended," and the rollout is currently limited. This mirrors what Google AI Studio already offers (Low / Medium / High). Alongside this, the Gemini app is preparing integrations with @Canva, @Instacart, and @OpenTable, joining @GitHub, @OpenStax, @Spotify, and @WhatsApp in the third-party app connector roster. Canva will let Gemini create designs and manage folders; Instacart will let it add ingredients to carts; OpenTable will let it book, modify, and cancel restaurant reservations through Reserve with Google.

OpenAI partners with Plaid to put ChatGPT inside your bank account

On Friday, May 15, 2026 OpenAI started rolling out a personal finance experience inside ChatGPT in preview for U.S. Pro subscribers, per OpenAI and TechCrunch coverage. The product, built on Plaid, lets users link more than 12,000 financial institutions (Schwab, Fidelity, Chase, Robinhood, Capital One, American Express among them) and ask ChatGPT portfolio, spending, subscription, and upcoming-payments questions. The feature is reachable via an @Finances mention in chat or a new sidebar entry, and is fed by GPT-5.5's contextual reasoning. Intuit support is "coming soon." OpenAI told TechCrunch: "More than 200 million users already ask financial questions to ChatGPT every month." The launch follows OpenAI's April acquihire of personal-finance startup Hiro and lands the same week Perplexity rolled out its own Computer-based finance research tool.

Claude Code at scale: Code with Claude 2026

Anthropic held Code with Claude 2026 on or around May 17, 2026, with a public opening keynote that has been viewed widely. The framing: Claude Code is no longer a coding assistant — it is "a reliable, self-improving engineering organization that teams can orchestrate and trust at scale." The keynote featured Anthropic engineers walking through how the Claude Code team at Anthropic itself has doubled in the last year while shipping more product, and how the engineering workflows have been re-architected around CLAUDE.md layered context files (root to project to subdirectory), agent teams running concurrent sessions, and the new Skills system (SKILL.md files) for repeatable procedural knowledge. The 90% autonomous coding target — the number Anthropic engineering leadership has cited publicly — was reiterated as the near-term goal, not a multi-year aspiration.

The Take

Three labs, three different bets on where the leverage is. Google is extending Gemini into the OS layer via Android and thinking-level controls that work in the consumer app. OpenAI is going vertical with data custody (Plaid now, Intuit next) and is willing to ask users to plug primary asset classes into the model. Anthropic is going after the engineering-organization layer, betting that runtime ownership is a bigger moat than model ownership. For builders: if you are wiring any of these into a product, expose the thinking-level control in your UI, treat the Plaid integration as a data-custody commitment not a feature, and assume the agent runtime you pick is the one you will be running for the next 3-5 years.

Quick Summary

Google is rolling out a "Thinking level" Standard/Extended option for Gemini 3 Flash and 3.1 Pro. OpenAI's ChatGPT now reads your bank accounts through Plaid, with 12,000+ institutions supported. And Anthropic is reframing Claude Code as a self-improving engineering organization that teams can orchestrate at scale.


Sources:


Source: TLDR | mr.technology — The Master Skill Index

Related Dispatches