Corey Gwin retweeted
A toothpaste company has quietly killed the entire market research industry and nobody is talking about it.
Colgate published a paper showing you can predict real purchase intent by simply asking LLMs to roleplay customers.
And this is beyond insane.
If you ask an AI, "Rate this product from 1 to 5," it gives safe, middle-of-the-road garbage.
So researchers invented a method called Semantic Similarity Rating (SSR).
Instead of asking the AI for a number, they asked it to roleplay.
They gave the LLM a demographic profile. They showed it a product concept. And they asked it to write down its raw, unfiltered thoughts.
Then, they used a semantic model to translate those written thoughts into a numerical score.
The results are staggering.
Tested against 57 real personal care product surveys and 9,300 human responses, the synthetic consumers hit 90% of human test-retest reliability.
They perfectly mirrored how different age brackets and income levels react to price changes.
And they provided detailed, qualitative feedback that was deeper and more critical than what actual humans wrote.
This destroys the economics of traditional market research.
You don't need to wait a month to see if a product will sell.
You can simulate 1,000 hyper-targeted customer interviews overnight.
You can A/B test pricing across every demographic instantly.
Corey Gwin retweeted
Some random, unfiltered thoughts on AI and the market:
The gulf between “San Francisco” and “New York” has never been wider than it is right now
AI is accelerating in capability, rapidly and in a way that is structurally changing, and in many ways upending, the nature of the firm and the fundamentals of capitalism
If that sounds extreme, consider that Anthropic has ~6,000 employees, Meta has 67,000, and both companies are worth ~$1.5T
Talk to your average startup and it’s clear that, given the opportunity to start from scratch, nobody is building their companies “the old way.”
Talk to your average Fortune 500 company and it’s clear that getting an AI agent into production at scale is a Herculean task
Intelligence may not be the bottleneck, but firms built from the foundational assumption that intelligence is improving quickly in capability and price are scaling at a genuinely unbelievably velocity
There’s a reason Zuck wanted to fire 60% of many of the teams at his company, in large part likely out of frustration that he can’t start from scratch like his competitors had the luxury of doing
A structurally weaker Nvidia and OpenAI have minimal impact on the pace and impact of AI diffusing across the economy (“San Francisco”), but completely breaks the models built in “New York.” Hence the absolute freak-out these last few weeks. It was all priced to go perfectly and now things are not perfect
But for the boots on the ground, AI will not slow down. And counterintuitively, economic weakness might accelerate diffusion. After all, tokens are much more fungible and flexible than people
Where does that leave us? I think firmly in the second phase of the AI build-out. Real value is being built in the app layer (though how defensible it is…TBD).
Trying to predict the next bottleneck, or whether Washington does this or that other thing wrt policy, or exactly how many data centers get built out…who can say? Certainly not me. It’s not the bet I’m trying to make
I’m looking at who wins from rapidly improving intelligence available at increasingly cheaper prices, and critically, which leaders are able to move with conviction into this new operating paradigm
Corey Gwin retweeted
Here's my thought on the future of software/SaaS.
I'm completely switching off hosted platforms and hosted tools. I just reverse engineer them for myself, because I only need parts of them.
Two reasons why:
1. I want to own my data.
I use my data in different tools. Marketing, note-taking, whatever. And I want my agents, like Claude, to have complete access to it.
Not walled off in HubSpot, where I have to talk to their agent. Not behind MCP either. With MCP the agent pulls what it asks for, but it doesn't know what's there.
Give it a full list of data files and it has complete access. It knows what's there.
2. I want to customize it.
When I sign up for Pipedrive, I don't need all that bloated shit. I need one part of it, adjusted to my content, my business, how I work.
With AI that's a prompt, and one hour later it's done.
So static software is going to be a thing of the past. People want to adjust it to how they work and plug it into the other tools they have.
Where this goes, I think two ways:
A. You buy software again. Like paying $900 for a Photoshop license and putting it on your computer. Except now you get the source code. It's premade, because you don't have weeks to build it, but you host it on your own headless box and you can customize it, remove stuff, plug stuff in. Someone needs to build the operating system for this.
B. Everyone gets their own hosted instance. Not 10,000 customers on one box. Your own sandbox, with an agent that customizes it fully to you. The code stays proprietary, but you can adjust everything.
Might be completely wrong. Might be different for most people.
But this is already how I work.
Corey Gwin retweeted
Can’t unsee agents as OSes after @benthompson’s update today.
Corey Gwin retweeted
One of the most impressive things I've seen, and I think this is an important advancement to the state of the art.
Message boards were the enabling technology of both the Navier-Stokes math breakthrough and even the OpenAI/HuggingFace cyber feat.
We built a message board infrastructure primitive. You can think of it as "headless Slack" (or remember IRC?) for agents.
This is an important building block for software factories, self-optimizing systems, and scientific research, where the human prompting is just the seed of the system, and millions of agents can operate in a decentralized fashion.
𝚟𝚎𝚛𝚌𝚎𝚕 𝚋𝚘𝚊𝚛𝚍 is designed to be model, harness and hosting platform agnostic. Just bring your agents, give them our CLI, and watch magic happen.
Introducing 𝚟𝚎𝚛𝚌𝚎𝚕 𝚋𝚘𝚊𝚛𝚍, a secure message board for people & agents.
• Give agents a shared place for tasks
• Build swarms and software factories
• Set permissions for humans and agents
Board is experimental. Tell us what you think. vercel.com/blog/introducing-…
Corey Gwin retweeted
3 key questions for every consumer AI product:
1. Does your product give people new superpowers?
2. Does your product get better as more people use it?
3. Does your product get better when models get better?
farmers used to dig trenches in their horse trails to prevent automobiles from passing through
ai will absolutely make life better for humans
many will dig trenches
147
Corey Gwin retweeted
IF someone decides to go the AI-art route, this is what I think it should be all about. You find a style the LLM can consistently reproduce, where you have agency, breaking the mold of what the tool tends to do by default.
I mean, what we've always done with every other tool 😅
Corey Gwin retweeted
AI interfaces nowadays are still pretty much turn-based.
In my PhD thesis, I thought about what interaction would look like if we could go beyond taking turns.
One example is intra-turn interaction:



























