Subscribe for Free
Anthropic's Claude Opus 4.7 dropped this week—and it arrives with some of the most significant performance jumps the model has seen in a single release. Visual reasoning accuracy leapt from 69% to 82%, coding on SWE-bench Pro climbed over 10 points, and the model can now verify its own outputs before responding.
What does that mean for engineers who live inside Claude Code all day—and does this finally put Anthropic ahead of OpenAI on benchmarks that actually matter in the real world?
Today in AI Brief:
Anthropic’s Claude Opus 4.7 arrives with vision and coding upgrades
OpenAI turns Codex into a full AI superapp
Uber bets $10 billion on a driverless future
Claude Opus 4.7 Arrives With Vision and Coding Upgrades
In Brief: Anthropic released Claude Opus 4.7, a meaningful upgrade to its flagship model with a 13-point visual reasoning improvement, stronger coding performance, and a new ability to verify its own outputs before responding — available now across Claude, its API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
Claude Opus 4.7, a meaningful upgrade to its flagship model with a 13-point visual reasoning improvement, stronger coding performance, and a new ability to verify its own outputs before responding — available now across Claude, its API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.
The Details:
Visual reasoning accuracy jumped from 69.1% to 82.1%, with image processing now supporting up to 2,576 pixels on the long edge — 3x the prior limit — making it far more capable with complex documents, slides, and UI screenshots.
On SWE-bench Pro, Opus 4.7 scores 64.3% (up from 53.4%), and ranks #1 on Vals AI’s Vibe Code Benchmark at 71%, directly challenging OpenAI’s models for developer tooling dominance.
A new xhigh effort level in Claude Code and a /ultrareview command for bug detection give developers finer control — though the new tokenizer uses up to 35% more tokens for identical text, which could affect costs.
Take Away: Opus 4.7 strengthens Anthropic’s position at the top of the model leaderboard, with gains in coding and vision — the two categories that historically drive the highest engagement among developers. Pricing holds at $5/$25 per million tokens, making these upgrades effectively free for current users.
OpenAI Expands Codex Into a Full AI Superapp
In Brief: OpenAI transformed Codex from a coding-focused agent into a full-featured superapp, adding Mac computer use, an in-app browser, inline image generation, persistent memory, and 90+ integrations — all available free with a ChatGPT account.
The Details:
Codex now has 3 million weekly users, growing 5x in three months with 70% month-over-month gains — one of the fastest growth curves for any developer tool in history, and a direct competitive response to Anthropic’s Claude Code.
With background computer use, Codex can now operate any Mac application independently — clicking, typing, and running parallel agents without interrupting your own work in other windows.
The platform adds 90+ new plugins covering Atlassian Rovo, CircleCI, and the Microsoft Suite — positioning it as an enterprise workflow hub rather than just a coding helper.
Take Away: OpenAI is no longer just building models — it’s building the productivity layer that sits on top of them. Codex at 3 million weekly users with this feature set is the clearest signal yet that the real AI battleground is the daily developer workflow.
Uber Commits $10 Billion to a Driverless Future
In Brief: Uber revealed plans to commit over $10 billion to autonomous vehicles — $2.5 billion in equity stakes across robotaxi partners and $7.5 billion to grow its AV fleet — targeting deployment in 28 cities by 2028.
The Details:
Uber is partnering with Baidu, Rivian, and Lucid rather than building its own self-driving technology, betting on platform scale and its existing rider network over in-house autonomy — a repeat of the strategy it used when it sold its AV unit to Aurora Innovation.
The company targets 15 cities by end of 2026, scaling to 28 by 2028, with $2.5 billion going toward equity stakes in autonomous vehicle companies and $7.5 billion toward growing its robotaxi fleet.
This bet puts Uber directly against Tesla Robotaxi and Waymo — but the company’s advantage is its existing global ride-hailing network, which autonomous operators need to reach millions of riders.
Take Away: Uber’s $10 billion move is less about technology and more about platform dominance — locking in fleet capacity before Tesla Robotaxi and Waymo can route riders around its network entirely. If successful, Uber transitions from gig-economy operator to autonomous mobility infrastructure.
The Shortlist
Sabi emerged from stealth with a brain-reading beanie packed with 70,000–100,000 EEG sensors that converts imagined words into text at 30 words per minute — backed by Khosla Ventures and Accel, with consumer units planned for late 2026.
Perplexity launched Personal Computer for Mac, a 24/7 AI agent for Max subscribers that securely accesses local files, reads iMessage and Apple Mail, and runs ongoing automations — recommended on a Mac mini for always-on operation.
Factory raised $150 million at a $1.5 billion valuation from Khosla Ventures and Sequoia to scale autonomous AI “Droids” that handle enterprise coding, testing, and deployment end to end — with customers including Nvidia, Adobe, and MongoDB.