In this issue:
OpenAI is hitting the brakes on its "side quests." The company is shelving its web browser, hardware experiments, and consumer distractions to refocus everything on enterprise and coding tools. The reason? Anthropic is winning business customers, and OpenAI knows it.
This isn't just an internal reshuffling. Alongside the strategic pivot, OpenAI dropped GPT-5.4 mini and nano, two smaller models built for high-volume business workflows. When a company stops chasing shiny objects to compete on practical tooling, you know the enterprise AI race just got real.
Topics of the day:
OpenAI sounds 'code red' to catch Anthropic
Nvidia charts a future of 'Agents-as-a-Service'
Anthropic's new feature completes tasks on your computer
Gamma revamps its AI design app as it nears 100M users
The Shortlist: Raycast Glaze, Cloudflare's /crawl API, Proof editor, PicoClaw, Google Workspace CLI
OpenAI sounds 'code red' to catch Anthropic
What's happening: OpenAI is cutting back on "side quests" like its browser and hardware projects to refocus its strategy on enterprise and coding tools, a direct response to Anthropic's growing dominance with business customers.
In practice:
This pivot means you can expect OpenAI to prioritize features that solve tangible business problems, think better coding assistants and workflow automations, not just flashy consumer apps.
The company just launched new smaller models, GPT-5.4 mini and nano, designed to make high-volume business tasks faster and more cost-effective.
For operators, this competition is a win, as it forces AI labs to build more practical and focused tools for professionals instead of chasing general-purpose hype.
Bottom line: The AI race is shifting from broad consumer chatbots to specialized tools for work. This focus will give you more powerful and efficient options to automate tasks and find growth levers in your business.
Nvidia charts a future of 'Agents-as-a-Service'
What's happening: At its annual GTC keynote, Nvidia's CEO declared the era of Software-as-a-Service is evolving into "Agents-as-a-Service." The company is backing this vision with new chips and software designed to help businesses build and deploy autonomous AI assistants.
In practice:
Think of this as hiring specialized AI agents for tasks like market research or software testing, paying for the outcome instead of just a software license.
Nvidia's new NemoClaw software stack gives your team a toolkit to start building these agents securely, with privacy controls for handling sensitive company data.
The new Vera Rubin platform is the underlying engine, providing the massive computing power needed to run these complex agent workflows at scale.
Bottom line: The infrastructure for a digital workforce is being built right now. It's time to start identifying the repetitive, high-value tasks in your business that an AI agent could eventually own.
Gamma revamps its AI design app for 100M users
What's happening: AI presentation builder Gamma launched a major overhaul of its platform, adding a new tool called 'Gamma Imagine' to create custom, on-brand visuals from a simple text prompt.
In practice:
Use the AI-powered builder to generate first drafts of pitch decks, proposals, or social posts in minutes, turning hours of design work into a simple prompt.
Create consistent graphics for your marketing materials without a designer by using the new Imagine tool to generate logos, charts, and posters that fit your brand guide.
Build out a full presentation directly from your research in ChatGPT or Claude, creating visuals and slides without switching apps.
Bottom line: Tools like Gamma are making professional-grade design accessible to anyone who can write a sentence. This puts pressure on legacy software to keep up with AI-native workflows that turn ideas into assets instantly.
The Shortlist
Raycast launched Glaze, a new tool that lets you create any desktop app in minutes by chatting with AI, with beautiful, production-ready output.
Cloudflare introduced a /crawl endpoint that crawls an entire website with a single API call. No scripts, no browser automation, just one request.
Every launched an agent-first document editor designed from the ground up for AI-assisted writing, where the agent is a first-class collaborator, not a sidebar feature.
PicoClaw shipped an ultra-lightweight AI assistant built in Go, prioritizing raw performance over feature bloat for developers who want speed.
Google released the Workspace CLI, an open-source command-line tool built for both humans and AI agents to interact with Gmail, Drive, Calendar, and Docs programmatically.
Anthropic's new feature completes tasks on your computer
What's happening: Anthropic released a new feature called Cowork Dispatch, letting you assign tasks to Claude from your phone that it completes remotely on your desktop. Think of it as sending your AI a to-do list while you're out, and returning to finished work.
In practice:
Automate research or data pulls by sending a quick text to kickstart complex work while you're commuting or in a meeting.
It uses your desktop's access to local files and apps, making it possible to delegate multi-step workflows like drafting a presentation from a messy folder of assets.
Bridge the gap between mobile ideas and desktop execution by having your AI prepare documents you can download and refine once you're back at your desk.
Bottom line: This is a major step from conversational AI to practical, agentic AI that you can delegate real work to. It's less about asking questions and more about assigning outcomes.
What I read/use lately
Tools, articles, and people worth your attention.
MiniMax M2.7: Early Echoes of Self-Evolution - A new model with self-evolution capabilities, pushing the boundaries of what smaller models can do autonomously.
Introducing Agentic Vision in Gemini 3 Flash - Google gives Gemini 3 Flash the ability to see and act on visual inputs, bringing agentic AI into the real world.
OpenAI Symphony: Open-Source Multi-Agent Framework - OpenAI open-sources a framework that turns project work into isolated, autonomous implementation runs across multiple agents.
How We Hacked McKinsey's AI Platform - A security team walks through vulnerabilities found in a major consulting firm's AI infrastructure. A cautionary read for anyone deploying enterprise AI.
Perplexity Announces Personal Computer - An always-on, local-first AI that merges Perplexity's search with your desktop environment for persistent, context-aware assistance.
This newsletter is where I (Kwadwo) share products, articles, and links that I find useful and interesting, mostly around AI. I focus on tools and solutions that bring real value to people in everyday jobs, not just tech insiders.
