Facebook once bought a VPN app for $120M and turned it into a surveillance tool that spied on 33M+ users' entire phones for years. This app helped Zuck buy WhatsApp for a whopping $19B and break Snapchat's encryption. Thread

We booked 460 meetings in the last 90 days at ColdIQ. And we'd love to help you do the same. Visit us at https://t.co/7UfsTtrciX to scale your outbound without adding more people to your team.

This cold email helped us build a $6M ARR agency: "Hi [Name], Saw you're [specific trigger]. Usually, that means [pain point]. We helped [Company] go from [before] to [after] using [method]. They saw [specific result] in [timeframe]. Mind if I share the 3-step process we

During the Polish Heritage Night at the Mayor’s residence – Gracie Mansion – @NYCMayor New York City Mayor Eric Adams gave an exclusive interview to Radio RAMPA and @RampaTV The conversation was conducted by Monika Adamski, editor-in-chief of Radio RAMPA. It was the second event

Want to launch or scale your own AI sales agency? Go to:

"Target people who hate your competitors." We saw 15-25% reply rates with this strategy for one of our clients. But the question is: How do you do it? ↳ Visit G2, Capterra, and Trustpilot. These platforms are full of people actively complaining about tools and looking for a

There is something seriously wrong with this study: 1. Dataset flaw The authors filtered their dataset down to just ASCII text and used about 1.22 million tokens per group (junk vs. control). Real AI models are trained on trillions of tokens, 1000X more data. But their "brain

Loved how deeply Karpathy explained why training AI on AI output destroys intelligence. > "You may look at each example of the LLM's reflective thoughts, and say this looks great, let's train on it. But you should actually expect the model to get much worse after training." Why

The full breakdown of my system:

> you are > running outbound campaigns in 2025 > still manually writing cold emails > watching others book meetings while no one replies to you > so you take a course sold by an "online guru" > but still nothing happens. zero meetings > but then you decide to go full AI mode >

“Real-world impact has come much faster with AI than it did with mobile.” -Mike Krieger AI is already changing budgets, workflows, and P&Ls. We sat with @mikeyk, CPO at Anthropic. Here’s what he told me: • How AI is driving revenue today • Sonnet 4.5: Opus-level brains at

RT the first tweet if you found this thread valuable. Follow me @itsalexvacca for more threads on outbound and GTM strategy, AI-powered sales systems, and how to build profitable businesses that don't depend on you. I share what worked (and what didn't) in real time.

Thanks for making it to the end! I'm Alex, co-founder at ColdIQ. Built a $6M ARR business in under 2 years. We're a remote team across 10 countries, helping 400+ businesses. Here's how I make $450k+ every month with AI: https://t.co/hyV1JWbitQ

Lastly, why does SSR work? When humans answer surveys, they first feel something about the product, then translate that feeling into a number. Traditional AI prompting skips the feeling & forces the number. But SSR mirrors the human process: feeling → interpretation → rating

With traditional research, you test 3-5 concepts max (too expensive). With SSR, you can test 50+ concepts for the same price.

Today we’re launching Scholar Agent, your AI copilot for scientific exploration. Scholar Agent breaks the shackles of the Google Scholar search experience. It plans and runs multiple searches, applies filters, and synthesizes results into a clear, citation-backed report⤵️

Now repeat this for hundreds of synthetic consumers with different demographics: Ages 18-75, various income levels, different regions, gender distribution matching your target market. Each will respond based on their persona. You just aggregate the results like a real survey.

The AI's response might be: 15% similar to "Very unlikely I'd buy" (Rating 1) 25% similar to "Not sure, maybe" (Rating 3) 60% similar to "Might try" (Rating 4) We turn these similarity scores into probabilities. Higher similarity = higher chance that's what the AI meant.

This is where SSR magic happens. You see, every sentence has a "meaning fingerprint" in the LLM's brain. Its response "I'm somewhat interested..." gets converted into a mathematical fingerprint. Then we compare how similar that fingerprint is to our 5 reference statements.

Ask: "How likely would you be to purchase this product?" The AI will respond naturally: "I'm somewhat interested. If it works well & isn't too expensive, I might give it a try after looking at the reviews." Notice that there are no numbers, just reasoning like humans.

Big news!!! FOUR Organics is officially at PopUp Grocer in NYC's West Village! Our mission has always been simple: create the cleanest lip balm possible with just FOUR organic ingredients. No compromises, no unnecessary additives—just pure, effective hydration. We're thrilled

Then comes concept attributes. You need to show the LLMs your product concept (as an image or text description): The paper tested with actual product concept images: full descriptions with benefits, pricing tiers, and brand positioning. Images work slightly better than texts.

Then you need to prompt the AI to be a synthetic consumer with specific demographics: "You are a 32-year-old female consumer from the Midwest, household income $60k-80k. You're being shown a new product concept in a market research survey." The demographics matter. A lot.

Let's say you're testing a new body wash concept. First, you need to create 5 reference statements that anchor each rating: Rating 1: "It's very unlikely I'd buy this" Rating 3: "I'm not sure, maybe if the price is right" Rating 5: "It's very likely I'd buy this"

SSR = Semantic Similarity Rating Instead of forcing AI to pick a number, you: • Ask it to explain its purchase intent • Compare that text to 5 reference statements • Use embedding similarity to create a probability distribution Let's now learn to use it:

We're looking to hire a full-time research assistant in my laboratory starting in January 2026. This is a great opportunity for someone excited about molecular and systems neuroscience, and considering graduate school. Our overall goal is to understand how we learn from our