OpenAI released GPT-6 Astra this week and called it AGI. The model outperformed every competitor across software engineering, science, and professional reasoning. Every high-skill task you currently outsource just got a credible AI alternative.

The same week, OpenAI also launched ChatGPT Sites, a tool that lets you describe a website in plain language and have it built for you. No code, no templates, no drag-and-drop. A non-coder built multiple working sites in a single week of beta testing.

Meanwhile, data from Agentio shows that YouTube's new view-counting method inflates public viewcounts by 40%. YouTube changed how it registers a view last month. The audience did not grow. The metric changed. Sponsors running their own analytics will catch it before you explain it.

One more thing I am watching: Nvidia confirmed a $12.9 billion acquisition of Hugging Face this week. One company now controls the hardware that runs AI and the platform where most open-source models are shared. That story is coming next week.

With ❤️
Humpty Calderon

OpenAI released GPT-6 Astra on September 3 and declared it the arrival of artificial general intelligence, with Greg Brockman calling it a historic inflection point. The model scored 99.9% on ARC-AGI-3, a benchmark specifically designed to prevent models from gaming the test. It outperformed all frontier competitors across software engineering, cybersecurity, science, and professional reasoning tasks.

The creator implication is direct: every high-skill task you currently pay someone to do now has a credible AI alternative. Legal review, technical writing, code, research, design briefs. The argument for hiring specialists to fill gaps in your solo operation just got significantly harder to make.

This is the most capable model available to creators right now.

See which creators get early access →

—

OpenAI launched ChatGPT Sites in public beta — a tool that generates functional websites from plain-language descriptions, no code required. It is available to ChatGPT Plus, Pro, Business, Enterprise, and Edu subscribers. A non-coder writing for Fast Company built multiple working sites in a single week of beta testing, including a companion guide and a feedback page with a form.

For creators who have been waiting until they could afford a developer, or until they had time to learn a no-code tool, that excuse is gone. A landing page for a course, a portfolio, a newsletter sign-up hub, a simple storefront: describe it in a sentence and the site exists. The sites load, accept form submissions, and work.

AI-generated sites can look generic without deliberate design input. But the floor just dropped out of the technical barrier to having a professional web presence. Describe what you want, then shape the result with your own design sense.

See what it can actually build →

—

Data from Agentio shows that YouTube's new view-counting method over-reports actual engagement by 40%. Last month, YouTube changed how it registers a view on long-form content — shifting away from a minimum watch-duration threshold to a system that logs views much earlier in a session. The change was not announced prominently. Most creators noticed their numbers going up and assumed it reflected better performance.

It does not. The audience did not grow. The metric changed. For creators who use viewcount as a benchmark in brand deal negotiations, this is a problem: sponsors who run their own analytics will catch the discrepancy before you explain it. Facebook ran a similar inflation in 2016 and it triggered significant advertiser budget reallocation away from creators.

Get ahead of this now. Know what your actual watch time and retention numbers say, and lead with those in sponsor conversations rather than the inflated top-line view count.

See exactly how YouTube changed the count →

Consumers can tell the difference. 79% say AI voices should come from real, attributed actors, and 61% say a voice is most memorable when it's exclusively tied to one brand. 

Voices is the only platform that designs, licenses, and captures your brand's AI voice from real, consenting professional talent—never from scraped data. From strategic voice casting and brand calibration to consent-based rights management and studio-grade capture, we handle the entire process end to end. 

BMW, SuperBloom, and Cresta already trust Voices to build their signature AI voice. Get a demo and see how a licensed, brand-owned voice becomes a real asset—not a legal liability.

Get a Demo

Meta launched Muse Voice Transcribe, a real-time speech-to-text API with speaker diarization for up to 20 speakers, priced at $0.18 per hour. Last week we covered Google's Gemini 3.5 Transcribe as a dedicated transcription model worth testing. Meta just entered the same race at a lower price point. For creators producing podcasts, running live interviews, or recording panel discussions, real-time diarization at this price makes automated transcription a no-brainer addition to your post-production workflow. The price war in transcription is good news for anyone who has been paying more for less.

See how it compares to what you're paying now →