OCTOBER 2, 2025

• PromptCraft: 5 Prompts to 100x Quality of Outputs; ChatGPT's Scraping Talents; 5 Household Management Prompts; AI 'humanizer' tools
• NewsDive: Sora 2 Is a Pain in the Butt, But Also a Lot of Fun
• THE HEADLINES: Yale study finds AI has yet to seriously disrupt the job market; Disney sends cease-and-desist to Character.AI; Agency refuses to 'sign' AI actress; Behind the millions of new materials AI is dreaming up
• Tool Spotlight: Imagine with Claude Builds Software in Real Time (Test Only)

5 Prompts to 100x Quality of Outputs; ChatGPT's Scraping Talents; 5 Household Management Prompts; AI 'humanizer' tools

• ​Five prompts that 100× results (SPLIT, BIONIC…): A compact toolkit: debate your own answer (SPLIT), ask the model to up-level its reply (BIONIC), summon expert personas (CONJURE), future-retrospective (TIME TRAVEL), and style mimicry (MIMIC). Useful for pressure-testing reasoning and quickly improving first drafts.
• ​Scaling a 5M-visit jobs platform with ChatGPT: Postmortem from Jobright.ai: LLMs weren’t reliable as “one-shot parsers.” Wins came from using ChatGPT to write/maintain scrapers, parsing fields separately to cut hallucinations, and validating against an internal company map—cleaner ingestion and fewer junk listings.
• ​Five household-management prompts that actually help: Ready-to-copy prompts for weekly meal planning + grocery checklist, seasonal deep-clean schedule, family budget template, maintenance log/calendar, and rotating chore system. Each includes inputs and packaging so outputs are immediately usable with minimal edits.
• ​Testing “AI humanizers” on a rulebook: Head-to-head on humanizer tools editing a dense game manual. Takeaway: some over-formalize or over-hype; Rephrasy balanced tone best. Practical combo: draft with ChatGPT, then selectively humanize to smooth stiffness while preserving author voice.
• ​Ontological Hierarchy of Knowledge prompt method: A philosophy-forward prompt to derive a multi-level ontology from “The Absolute,” enforcing emergent naming and recursive expansion. Handy for research maps, world-building, or structuring an unfamiliar domain before deeper exploration and iterative prompts.
• ​OpenAI Academy: Prompt Pack for Students (Edu): Classroom-ready prompts for ChatGPT Edu: study planning, research synthesis with citations, problem-set tutoring, and presentation drafting. A useful catalog for students and instructors to standardize academic workflows—best paired with course policies on AI use.

Sora 2 Is a Pain in the Butt, But Also a Lot of Fun

My hopes for Sora 2 were not particularly high when I embarked on the convoluted journey to get my account up and running on Wednesday afternoon.

I've gotten spoiled to one-click signups through my Google account, but OpenAI, perhaps having learned a lesson from previous traffic logjams, decided to make Sora 2 a chore to access.

First, try to access Sora through the link provided on the sidebar of your eligible ChatGPT account. I pay the $200 a month for Pro (for now), so this part wasn't difficult for me. Not sure if it extends to Plus users just yet.

After doing this, it gets tricky. You'll get a screen that looks like this.

Click Enter invite code, but don't get your hopes up too much just yet. After all, that code has to come from somewhere, and that somewhere is Discord.

Go to OpenAI's official Discord server. Make sure you're logged in and accept the invite. (See below.)

Hold up, though. You've still got a bit to go. Once here, check the sidebar. You'll find this option under the OpenAI heading.

Choose #openai-verification, then select Verify with OpenAI. From there, you'll log in to your ChatGPT account.

This will open you up to the #sora-2 subchannel.

Once you've clicked this, it'll take you to a giant chat scroll where you can request an invite code. Even so, you'll get yours stolen a time or two, or in my case six, before you succeed at stealing someone else's and entering it into the code verification screen in time (that's the place OpenAI took you when you clicked Enter invite code earlier in this article.

After the code is accepted, you can begin creating videos. That's as simple as entering a prompt and waiting 2-3 minutes for the result. You can also upload images for a starting point, though it says it can't do videos of photorealistic people at this time.

Even so, there's a way around it by creating a "Cameo" of yourself. The "Cameo" feature can be overwritten at any time, so you have some room for trial and error.

Just center your face in the screen, say three numbers that pop up, turn your head to the left, right, and look up. That's all Sora 2 needs to create a pretty convincing likeness of you that can then be dropped into any video. Here's me lifting a car.

​CHECK OUT THE REST ON LINKEDIN​

HIGHLIGHTS: Yale study finds AI has yet to seriously disrupt the job market; Disney sends cease-and-desist to Character.AI; Agency refuses to 'sign' AI actress; Behind the millions of new materials AI is dreaming up

1. ​OpenAI launches new AI video app spun from copyrighted content​
2. ​OpenAI Is Preparing to Launch a Social App for AI-Generated Videos​
3. ​Meta plans to buy chip startup Rivos to boost semiconductor efforts​
4. ​CoreWeave signs $14 billion AI infrastructure deal with Meta​
5. ​Citigroup forecasts Big Tech’s AI spending to cross $2.8 trillion by 2029​
6. ​Former OpenAI and DeepMind researchers raise whopping $300M seed to automate science​
7. ​Disney sends cease-and-desist letter to Character.AI, Axios reports​
8. ​Gersh Will Not Sign AI Actress Tilly Norwood: ‘We’re Not That Agency’​
9. ​Cerebras Systems Raises $1.1 Billion Series G at $8.1 Billion Valuation​
10. ​Peloton debuts AI coaching and new products in sweeping overhaul​
11. ​Spec-driven development: Using Markdown as a programming language when building with AI​
12. ​US jobs market yet to be seriously disrupted by AI, finds Yale study​
13. ​U.S. Energy Department Seeks Proposals for AI Data Centers​
14. ​AI is dreaming up millions of new materials. Are they any good?​
15. ​Gemini for Home: Expanding the Platform for a New Era of Smart Home AI​

Imagine with Claude Builds Software in Real Time (Test Only)

​

​

Imagine with Claude is a live, hands-on demo where Claude Sonnet 4.5 builds mini-apps as you chat. No prewritten code, no fixed UI.

You describe what you want, and Claude generates and refines the software in real time, so you can prototype ideas, test flows, and iterate with plain language instead of spinning up a dev environment.

Standouts: instant code + interface generation in one place, natural-language edits (“add a filter,” “make it dark mode,” “hook up a chart”), and a sandboxed experience that adapts to your requests on the fly. Currently offered as a short-term research preview for Max subscribers (through Oct 4, 2025).

So, yes, you'll have to pay to try it out. For now, you may just want to enjoy the demo video above. But I'd look for something like this to become a permanent part of Claude in the not-too-distant future.

Learn more: https://claude.ai/imagine​

​Read All mAIn Street Back Issues Here​

Your input fuels the content. Let me know how I did today!