Your Next Domain Comes With $1 Off

A small gift for your next big idea.

If you’re launching a business, newsletter, side hustle, or new project, don’t waste money on overpriced domain renewals.

Get transparent pricing, free WHOIS privacy, free SSL, and real human support 24/7 with Porkbun. Your next domain is $1 cheaper today.

Today In Ai

1  OpenAI launched ChatGPT Work, a computer-use agent that can hold focus on a single project for hours.

ChatGPT Work is OpenAI's new desktop agent, built on GPT-5.6, that can take actions across your computer or phone, pull context from your existing files to match your writing style, and sustain focus on a single project for hours without supervision. It ships inside a new desktop app that bundles ChatGPT, Work, and Codex under one roof, free on all plans. Try ChatGPT Work or see how teams are using it.

2  Meta launched its first paid AI model, ending a years-long commitment to keeping everything free and open.

Meta's Muse Spark 1.1 is a multimodal reasoning model built for agentic tasks, coding, and computer use, available through Meta's new Model API. Meta's AI chief Alexandr Wang described the pricing as aggressively competitive against other frontier models. The pivot from a fully open-weight strategy to a paid offering reflects growing pressure from Wall Street to show returns on the $145B infrastructure spend Meta committed to this year.

3  PromptQL relaunched as a multiplayer AI coworker with $136M raised and 2M views in its first week.

PromptQL is repositioning itself as the first AI version of Slack, giving teams a shared AI agent that operates across a common thread, can run on any frontier model, and turns every correction made by the team into a reusable skill the agent carries forward. The memory-from-mistakes feature is the most distinctive part: the agent gets better at the team's specific workflows the longer it is used. The relaunch video pulled 2M+ views in its first few days.

From The Frontier

OpenAI on voice and models. GPT-Live, OpenAI's new real-time voice family that listens and speaks simultaneously, is now rolling out inside ChatGPT, removing the turn-taking delay that has made AI voice feel unnatural. The GPT-5.6 model family went public in the same week after the US Commerce Department lifted a weeks-long export restriction. Two significant releases in one window.

Anthropic on brains, cowork, and time. New research revealed that Claude has a hidden internal workspace where it reasons through problems without exposing those thoughts in its visible chain-of-thought. Anthropic compared it to the way a person can hold a conversation while thinking about something else entirely. Watch the explanation here. Claude Cowork went cross-device in the same week, and Fable 5 access was extended through July 12 on paid plans.

Meta on models and money. Meta's Superintelligence Labs shipped three things in quick succession: Muse Image, which ranked third on Arena's text-to-image leaderboard on debut; a preview of Muse Video ahead of a wider rollout; and Muse Spark 1.1, the company's first paid model via its new Meta Model API. The shift to paid is the most consequential of the three.

China on chips and cost. Tencent shipped Hy3, a smaller open-source model that outperforms models significantly larger than itself, while Meituan fully open-sourced LongCat-2.0, described as the first trillion-parameter coding model trained entirely on Chinese chips rather than Nvidia GPUs. Both extend the same pattern: Chinese models matching or exceeding US model performance at lower cost, now increasingly without US silicon.

What people are actually watching and sharing

The AI flyer problem. A viral post with 3M views argues that AI-generated flyers are cluttering physical spaces with a recognizable aesthetic that is getting harder to ignore. This breakdown of the specific visual style is worth reading if you are producing anything designed to be printed and posted.

48 hours to irrelevance. Meta's first-ever image model debuted two days ago. It has already been outranked on Arena's leaderboard by a new model from a research lab. The current pace of the model leaderboard is not compatible with a release cycle measured in months.

Rotary phone, modern AI. Video director Karen Cheng connected a vintage 1920s rotary phone to an AI agent that responds through a mechanical message board. It is technically a Cursor ad, but the comment sections are praising the concept rather than calling it out for that.

Gear-shift model switching. Someone built a physical gear shifter that changes which AI model is active when you move it. The Reddit comments agree it is simultaneously useless and impressive, which is often how the most memorable projects land.

Claude now tracks your habits. A new reflect feature in Claude's settings generates a monthly recap of how you have been using the tool, along with options to set quiet hours or schedule break nudges. Here is what it looks like in practice.

Prompt Station

This ChatGPT prompt acts as a curriculum designer, elite teacher, and learning scientist in one. Give it a topic, your starting level, your target level, and a timeframe, and it returns a structured roadmap with daily, weekly, and monthly actions, practical projects, revision cycles, quizzes, common mistakes, prerequisite knowledge, and a difficulty score for each stage. It is designed to replace the hours most people spend figuring out what to learn and in what order.

Build a complete learning roadmap for any topic in one prompt

Act as an elite teacher, curriculum designer, and learning scientist. I want to master [TOPIC] from [CURRENT LEVEL: beginner/intermediate/advanced] to [TARGET LEVEL] in [TIMEFRAME]. Create a complete learning roadmap with foundations, intermediate concepts, advanced topics, practical projects, exercises, quizzes, milestones, common mistakes, prerequisite knowledge, recommended resources, real-world applications, revision cycles, and progress checkpoints. Structure everything into daily, weekly, and monthly actions. Also explain why each learning stage exists and estimate mastery difficulty on a scale of 1 to 10.

Replace [TOPIC] with what you want to learn: "Python for data analysis", "financial modelling", or "Spanish to conversational level". Replace [CURRENT LEVEL] and [TARGET LEVEL] with your honest starting point and where you want to land. Replace [TIMEFRAME] with a realistic window: "90 days", "6 months", or "one year with 30 minutes per day". The difficulty scores on each stage are the most underrated output. They give you an honest read on where you are likely to stall before you get there, so you can plan for those moments rather than be surprised by them.

Porkbun is the domain registrar trusted by folks who want low prices without the nonsense. Get most domains at cost, free features like WHOIS privacy & SSL certificates, plus real 24/7 human support. Get $1 Off Your Domain

Keep Reading