Everything May Habib said on any show that made the record, most notable first. Each card names its show and opens the statement there.
Habib predicts AI will be able to run companies within five years
“This is AI that in three to five years is going to be able to run companies.”
Enterprise AI drives 500% capacity shifts, not just 10% productivity gains
“None of this is about productivity, right? You being able to go through Three decades of consumer research on Listerine to be able to develop a new flavor, right? You being able to develop all of the go to market around that new product in literally a third of…”
Habib: Writer's largest models match GPT-4 quality at lower hosting cost
“Now we've got 18 models. The biggest ones are GPT-IV level quality, but at a fraction of the cost of hosting. Super, super fast for the enterprise use cases that That our customers run them on.”
Habib: AI capabilities will outpace most organizations' capacity to absorb them
“And I do think the capabilities that are possible are gonna outstrip most organizations' capacities to absorb them, for most of their employees to absorb them without a lot of effort.”
Habib: Writer qualifies out enterprise prospects with weak champions, even CIOs
“And so our team is really good at politely declining to engage, right? Not just because the use case doesn't fit or it's not ICP, but we qualify the champion out, even if it's a CIO.”
Habib: Writer has the potential to become a 400 percent NRR company
“I think this should be a 400% NRR company.”
Habib: Current transformer architectures fail at enterprise agentic workflows
“You know, our kind of tests on inference on quality for agent use at work. It's just transformers. Just, it was current capabilities, really not able to do what we wanted them to do. And so new architectures that we're working on are super, super promising.”
Habib: Vector databases fail at RAG retrieval beyond 50,000 pages
“Like, it's fine when I am, I've indexed 50 pages, but when it goes to 50,000, you know, I can't get this needle in a haystack out of my kind of VectorDB approach.”
Vector RAG struggles with nested tables and overlapping context, Habib explains
“When you are doing a vector-based approach, you're going through, and you are, you've chunked Right, that policy, and, you know, against the prompt, you are trying to find the most relevant chunks, right, in that policy, which means it does really badly, right…”
Vector RAG requires full re-indexing for single-file updates, Habib claims
“And when that policy, even one policy gets updated, you're throwing away the entire embedding store and starting over.”
Writer trained its Palmyra X 004 model using just $700k in compute
“Just on GPUs, but yeah.”
Habib: Unequal AI access will cause 100x more inequity than language barriers
“Being able to use AI or not, Is likely to create a hundred X more inequity in society than all of us speak in different languages.”
Habib: Enterprise champions target $100M in cost reductions using Writer
“Showing people that we've got champions who are signed up with procurement to take a hundred million dollars of costs out of their organizations completely built on writer.”
Habib: Enterprise innovation and POC budgets are easily cut distractions for startups
“I think, you know, the deals around POCs and innovation and, you know, sort of, I call it funny money. You don't really want that stuff, actually. I think that is a distraction, because it's easy to cut. Your team will have spent a lot of time, and it'll be fo…”
Habib: Scaling GenAI in SMB and mid-market is hard due to switching
“And I think it is just so much easier in mid-market and SMB from a functionality perspective to just switch, right, to something else, especially in generative AI where it'll be confusing for a while. It's gonna be hard to scale a big business if your buyers a…”
Habib: Most enterprise buyers reject data sharing and Azure AI vendor lock-in
“And for us, the vast majority of enterprises we talk to, that stuff can't leave their environment, and they're not excited about the kind of lock-in that an Azure AI requires, right?
And the way that that scales the amount of engineering effort that's required…”
Habib: Writer runs without regular one-on-ones to preserve asynchronous culture
“And so we're able to have a really low meeting culture as a result because, you know, we are able to do things as a group. That meeting is 90 minutes. Sometimes it goes two hours, a little bit longer. But, you know, as a result, I don't really have one on ones…”
A customer rebuilt Habib's previous startup Qordoba over a single weekend
“We had a customer basically in a weekend rebuild Cordoba using Writer.”
Writer relied on GPT-3 for only a brief three-month period
“So we, we've always had our own technology down to the LLM. There was a very brief period, literally like three months, where we, for some of our apps, used GPT-III because it was just so much better.”
Writer stores its knowledge graph as flat JSON in Postgres
“I mean, it is a flat JSON file. This is stored in a Postgres database. It's not even, A graph database.”
Writer provides 1B randomized tokens for customers to audit copyright compliance
“We give folks, this is another reason we build our own models, a billion randomized tokens that they can query themselves to validate that we use copyright free information and that, you know, our bias and toxicity distribution is what we say it is”
A Fortune 50 CIO personally demoed an eight-person AI startup
“I mean, literally, I was talking to a CIO yesterday. Fortune 50 had just taken personally a demo from an eight-person startup, because people are so curious.”
Hyperscalers spend more on single enterprises than AI startups raise total
“The consultancies and the strategic advisory firms and every hyperscaler can literally throw more innovation dollars at the enterprise than you will raise in three years from venture capital, right?”
Habib: Writing is the last unstructured enterprise business process
“Writing is super strategic, but is the last unstructured business process that a company does.”