wiz @ feat @feat_studio

@usmanshabbir I can see AI agents as a new channel. Over time - you are right, it will become the dominant channel through which services get sold

Twitter Essay #1: LLMs have started the transition from selling intent to fulfilling it. When the internet emerged as a new social layer, selling intention became the dominant way platforms monetized their business. After all, users were already spending countless hours interacting with the internet, and the ability to capture intention and package it as a way to enable ad targeting seemed like an obvious thing to do. Little did we know that the motivation to drive revenues would not only turn behavior into the intention that gets sold, but change the behavior altogether to drive a particular intention pushed by adverts. @Google today has a cash pile of $119.93 billion, @Meta is $49.6 billion, & @TencentGlobal $58.3 billion. In hindsight though, it seems weird that selling intention became the dominant way to monetize the internet rather than fulfilling it most efficiently. Marketing spend creates inefficient marketplaces. There is only one cheapest car rental in economy class for next Friday, but good luck finding it. All search queries have "one" single final destination, and while one may imagine that the internet would be built to make that journey as efficient as possible, instead it's a myriad of distractions and shiny objects that get you to get anywhere but the destination. Engagement is the north star metric for all social platforms. Internet giants are built on exchanging intent for attention. The fact that the intent graph is owned by a few big players creates friction for entrants who want to make better use of user intent. Search intent is entirely owned by Google. Interests intent is owned by @Meta (Facebook, Instagram, Whatsapp) and business intent (B2B) is owned by @LinkedIn and @X is somewhere in the mix there. This is where I see LLM model completely upend these business models. If done right, they will not be monetized on selling intent, but fulfilling intent. This is also a warning call to all businesses out there - you can only compete on merits of providing the best service at the lowest cost. Traditional moats of distribution & brand power will give away as LLMs equalize the playing field. We are thinking about recruiting in the same way @useparallel. Instead of selling Job Ads & Tools - we think the power of AI should enable & fulfill the basic ask: high quality talent for your open roles fast & cheap. #usshabb

mll

@elonmusk Sufism.

@Alex_elnaugh it feels ambiguous - like what is this?

@PeterDiamandis three themes: 1. Their society: who lives in it, where they are in terms of privilege and access to opportunity. 2. Their money: how to earn it, manage it, grow it, and give back. 3. Their expression: what’s their craft, language to express their craft and medium to distribute it

This makes a broader point about LLM systems - they are super power agents that can think through answers. You are interacting with a computer interface not a ML type predictive rules engine of the past.

Jack Clark @jackclarkSF

3 years ago

One thing I regularly obsess about is how 'today's AI systems are more powerful than they appear' - here's a nice example for Claude via @AnthropicAI of how by prompting your system more intelligently you can eke out significant performance boosts

12 43 250 82K 139

@omarsar0 I think there are people are building LLM apps and those that are talking - like being against prompt engineering isn’t even a a position.

@wbetiago the picture of great - until you take your hands off to stretch and your laptop falls in water 😂

@glennwrites1 I think part of the problem is that course sellers sell snake oil leading to inflated expectations

@venturetwins regulated capture.

@masterchief164 keep grinding

@lexfridman For me - it was when I could buy @ChipotleTweets . I had to save $2.00 every day for 5 days I worked in college - Chipotle was $6.25 and I shared a @Zipcar with my friend which costed $4.75 - this is 2012.

Another large surface area opportunity for existing companies with ChatGPT is to automate onboarding flows. You can now spin up placeholder profiles to get users into your product - reducing your time-to-value.

@collibra is a good example here. They can completely append the way Data Quality works vis a vis ChatGPT - the work they need to do is trivial - fine tune a GPT model for data consistency rules. However, knowing Collibra - they can be super slow to implement it.

ChatGPT hardly represents a wedge to go after incumbents. Given the low barrier to entry - most of value will accrue to existing players, however there will be some super dumb companies that will not roll out these features - and they will die.

Super excited with my POC results today - essentially I have built logic that allows you to talk to APIs using GPT like interface. You can connect your bankaccount - and then just start asking questions. LLM on structured data that also provides structures data can unlock so much

@nico_jeannen Hey Nico - I tried talk notes. Its really good. What I love is just how easily I could get started - I rambled for a bit and then asked it to scratch what I said and simply create a to do list. I must say - results were pretty accurate.

@SophiaKianni @alexisohanian congrats!!

@SamMendelsohn4 @elonmusk I heard from @WalterIsaacson that he is turning @X into a hardcore company - let’s see how hardcore 😂