Stop telling ChatGPT “brainstorm.” Use these prompts to find a $10k/month idea 💰

Welcome to another edition of Horizon AI,

You don’t need 50 ideas. You need the one that can actually make money. Today, we’re sharing how to use ChatGPT to be brutally honest, kill weak ideas, and focus on the ones with a realistic path to $5k–$10k per month.

Let’s jump into it!

Read Time: 4.5 min

Here's what's new today in the Horizon AI

• Zoom Launches AI Companion 3.0 with Browser Access and Agentic Workflows

• AI Reasoning Models Now Pass All Three CFA Exam Levels

• AI Tutorial: Use These 9 ChatGPT Prompts To Find a $5k–$10k/Month Business Idea

• AI Tools to check out

• AI Findings/Resources

• The Latest in AI and Tech 💡

AI News

ZOOM

Zoom Launches AI Companion 3.0 with Browser Access and Agentic Workflows

Zoom has rolled out AI Companion 3.0, expanding its workplace assistant beyond meeting summaries to include browser-based access and agentic automation, “transforming conversations into action.”

Details:

• AI Companion 3.0 is now accessible through the browser at ai.zoom.us, allowing users to interact with the assistant across both desktop and web workflows instead of being limited to the Zoom client.

• The update adds personal workflows (beta), a low-code, drag-and-drop system that lets users automate follow-ups, summaries, daily reports, and task execution using AI agents.

• Zoom uses a federated AI approach that combines its own small language models with larger frontier models from providers like OpenAI and Google, escalating complexity only when needed to improve quality while controlling costs.

• New agentic features include cross-meeting retrieval, post-meeting follow-up generation, daily reflection reports, AI-assisted writing and editing, and easy export to formats like PDF, Word, Markdown, and Zoom Docs.

• The assistant can retrieve information across past meetings, Zoom Workplace notes, and connected tools such as Google Drive and OneDrive, with Gmail and Outlook integrations coming soon.

Zoom says its multi-model orchestration strategy showed measurable gains on advanced reasoning benchmarks, reinforcing the value of coordinating multiple models rather than relying on a single system.

Source

TOGETHER WITH ROKU

Find customers on Roku this holiday season

Now through the end of the year is prime streaming time on Roku, with viewers spending 3.5 hours each day streaming content and shopping online. Roku Ads Manager simplifies campaign setup, lets you segment audiences, and provides real-time reporting. And, you can test creative variants and run shoppable ads to drive purchases directly on-screen.

Bonus: we’re gifting you $5K in ad credits when you spend your first $5K on Roku Ads Manager. Just sign up and use code GET5K. Terms apply.

Use code GET5K now

AI RESEARCH

AI Reasoning Models Now Pass All Three CFA Exam Levels

A new academic study shows that today’s reasoning-focused AI models can successfully pass all three levels of the CFA exam, one of the most demanding certifications in finance.

Details:

• Researchers tested six leading reasoning models across 980 CFA-style questions spanning all three exam levels, covering multiple-choice, case-based analysis, and open-response formats.

• Models from Google, OpenAI, Anthropic, xAI, and DeepSeek all met established passing thresholds, with several achieving near-perfect results on specific sections.

• Gemini 3.0 Pro set a new record on Level I with a 97.6% score, followed closely by GPT-5 and Gemini 2.5 Pro, while even the lowest-performing model still exceeded 90%.

• GPT-5 led Level II, which focuses on applied analysis and case studies, while Gemini 3.0 Pro dominated Level III constructed responses, showing major gains in complex synthesis and reasoning.

• The study highlights ethics as a consistent weakness, with higher error rates across models, suggesting limits in contextual judgment despite strong technical performance.

• Researchers caution that passing exams does not equal real-world job readiness, noting issues such as benchmark limitations, potential data contamination, and gaps in handling ambiguity and human decision-making.

In 2023, the leading LLMs could answer some CFA exam questions, but overall performance was mixed, with most failing to pass. Now, just two years later, AI models have gone from struggling to scoring near the top, bringing us one step closer to a point at which financial institutions could begin thinking about integrating AI into actual workflows.

Source

AI Tutorial

Use These 9 ChatGPT Prompts To Find a $5k–$10k/Month Business Idea

Most people brainstorm dozens of business ideas and end up stuck, overwhelmed, or building something that never makes real money.

The problem isn’t a lack of ideas. It’s a lack of brutal filtering.

Here is a set of ChatGPT prompts you can copy and adapt to cut through the noise, kill weak ideas, and focus only on the ones with a realistic path to $5k–$10k per month.

1. Turn chaos into a clean idea list

Act as my brutally honest co-founder.

I’m overwhelmed with too many business ideas. Your first job is NOT to be nice, it’s to organize them so we can kill most of them later.

First, ask me to paste ALL my ideas in one message, even if they’re messy.

Then:
1. Turn my brain dump into a clean, numbered list of ideas.
2. For each idea, write:
   - A 1-sentence description in plain English
   - Who would pay for it (specific type of customer)
   - HOW it might make money (1 sentence: subscription, service, product, etc.)

Do NOT evaluate them yet. Just organize the chaos so we can judge them ruthlessly later.

2. Money-first scoring matrix

Act as a skeptical angel investor.

Using the numbered list of my ideas, score each one purely on its MONEY potential.

Create a table where each idea gets a 0–10 score for:
- Market demand
- Willingness to pay
- Pricing power (how high I can realistically charge)
- Speed to first $1,000
- Realistic path to $5k–$10k/month within 12–18 months

Then:
1. Add a “Total Score” column and rank ideas from highest to lowest.
2. Briefly justify the score for each idea in 1–2 sentences.
3. Bold the top 3 ideas by total score.

Do not sugarcoat anything. If something looks weak, say it clearly.

3. Kill the weak ideas (brutal filter)

Act as a ruthless idea executioner.

Using the scored list we already created, your job is to aggressively CUT ideas that are unlikely to make real money.

For each idea:
1. Give a quick verdict:
   - “STRONG CONTENDER” (keep)
   - “MAYBE LATER”
   - “KILL IT”
2. For every “KILL IT”, give a blunt 1–2 sentence reason focused on MONEY:
   - tiny market, low pricing power, hard to reach customers, etc.

Then:
- Show me ONLY the ideas labeled “STRONG CONTENDER” in a short list.
- Make sure at least 50–70% of ideas are killed or moved to “maybe later”.

Be direct. I care more about making $5k–$10k/month than protecting my feelings.

4. Reality-check my time, skills, and runway

Act as my brutally honest advisor.

Here are my constraints:
- Time I can work on this per week: [HOURS PER WEEK]
- Money I’m willing to invest upfront: [BUDGET]
- My main skills: [LIST SKILLS]
- Things I refuse to do (or hate doing): [LIST THINGS]

Using ONLY the “STRONG CONTENDER” ideas:
1. For each idea, estimate:
   - Time to launch a basic MVP (in weeks)
   - Time to first realistic sale
   - Startup cost (LOW / MEDIUM / HIGH, with a rough range)
2. Note how well my current skills match each idea (GOOD / OK / BAD).
3. Label each idea:
   - “REALISTIC PATH TO $5k–$10k/MONTH FOR YOU”
   - “POSSIBLE BUT HEAVY LIFT”
   - “NOT A GOOD FIT WITH YOUR CURRENT REALITY”

Sort the list so the best fit ideas for ME are at the top. Don’t be optimistic, be realistic.

AI Tools to check out

📽 ImagineX: A professional AI video and image generator built for modern creators to help produce engaging content faster than ever.

✍ LazyTyper: Advanced voice typing powered by 12 AI speech models. Write 3× faster with superior accuracy.

👨‍💻 AppWizzy: Professional vibe‑coding platform to build scalable web apps and websites by simply chatting with AI.

🦾 FAQ Ally: Train AI agents with your company data for easy knowledge access.

✨ TruGen AI: Delivers hyper-realistic Video Agents that see, hear, remember, and act in real time, transforming conversations into natural, human-like interactions.

TOGETHER WITH LEVANTA

Why AI Isn’t Replacing Affiliate Marketing After All

“AI will make affiliate marketing irrelevant.”

Our research shows the opposite.

Shoppers use AI to explore options, but they trust creators, communities, and reviews before buying. With less than 10 percent clicking AI links, affiliate content now shapes both conversions and AI recommendations.

Download the full report

AI Findings/Resources

📝 Gemini 3 prompting: Best practices for general usage

👀 Explore real world Claude use cases across research, writing, coding, analysis, and everyday tasks for solo or team work.

🚀 10 prompts that can transform your day to day life

💡 Can AI make us more creative? Study reveals surprising benefits of human-AI collaboration

⭐ The great AI hype correction of 2025

The latest in AI and Tech

Merriam-Webster's 2025 word of the year is 'slop'

Merriam-Webster named “slop” its 2025 word of the year, defining it as “digital content of low quality that is produced usually in quantity by means of AI.” The dictionary said the term captures public skepticism about the new technology replacing human creativity.

Google is gradually rolling out a NotebookLM integration for Gemini

This will allow users to use selected notebooks as context for their queries, expanding the chatbot’s knowledge base beyond its training data and web results.

A Perplexity study shows AI agents are mostly used for productivity and learning

Perplexity and Harvard researchers have published a new study based on data from the startup’s AI browser and digital assistant, Comet.

Researchers analyzed hundreds of millions of queries and found that 36% focused on workflow and productivity, while 21% were related to learning and research.

US government launches ‘Tech Force’ to recruit AI and tech talent

The US Office of Personnel Management is launching the “US Tech Force” to attract early career AI, software, and tech talent to government roles.

The two-year program aims to hire 1,000 employees across agencies, offering salaries of $130K–$195K and mentorship from tech companies like Microsoft, Amazon, Meta, Adobe, and xAI.

Disney's OpenAI deal is exclusive for just one year

Disney’s new three-year licensing partnership with OpenAI grants exclusivity for just the first year, CEO Bob Iger said.

The deal lets OpenAI’s Sora video generator use over 200 Disney, Marvel, Pixar, and Star Wars characters, and once that exclusive year is up, Disney can later sign similar agreements with other AI companies.

That’s a wrap!

Thanks for sticking with us to the end! Let’s stay connected on LinkedIn and Twitter.

Not subscribed yet? Sign up here and send it to a colleague or friend!

See you in our next edition!

Gina 👩🏻‍💻