the @aisdk is probably the best way to build an ai app right now

you can go from idea -> working ai app in 15 mins.

its made working with llms 10x more enjoyable ( and we ship faster) its literally one loc to swap to over any LLM provider

Google recently launched Vertex - their new AI offering

It includes various ML models, including text, image recognition and more all in one spot

And the best part: Anyone can leverage them for their business with a single developer

Here's my favourite ones + example use cases 1/ PaLM 2, their foundational text model - similar to chatGPT & powers bard, but can be tuned for a wide variety of use-cases.

Ex: You're a small bakery, you copy paste all your reviews into it, and find out people love cupcakes, but want gluten free options

So its been a few days since the vision pro launch, and I think i figured out Apple's VR play.

Or rather, how VR will actually take off, and why Apple is perfectly positioned to build Jarvis in real life.. Alright so lets get the facts straight:
- Vision pro is $3.5k
- Comes out 2024
- M2 + R1 chip
- Their WWDC announcement was VERY different for the vision pro (generally they have everyone using it)
- No specific use cases (replacement for tv? taking a video of your kid? work?)

Microsoft just announced they are adding Windows Copilot, their AI operating system.

It looks great at a first glance, and you'd think Apple is in trouble. But that's far from the truth.

While Windows AI OS might be cool, Apple's MacOS may have already won.

Here's why: Ok first up the new Windows AI OS. From the demo video it
- copilot like os, basically clippy
- lets you edit, summarize and create documents across windows
- can use a chatbot to change settings like dark mode or turn bluetooth on
- send images to teams

The medical industry will never be the same again.

Google has released Med-PALM2, their LLM specifically for medical purposes.

It scored higher than humans on a medical exam.

Heres the breakdown, and a glimpse of future of healthcare: Alright so what' is med-palm? It's google's version of palm-2, but fine tuned and trained specifically for medical purposes.

That means its much MUCH better than a generic LLM at
- understanding users questions
- understanding various health scenarios
- synthesizing answers

Quick comparison of Claude vs GPT-4.

Speed on GPT4 is a huge issue, and its still slower despite it not being peak times.

How do you think the outputs compare? At first GPT4 seems more comprehensive, but is it really? Ex:

I didn't even ask for that? It kinda made it up with random numbers

Made agents shareable so its easier to share them with other people (instead of screenshotting it)

Uses Vercel OG image to dynamically create the preview image.

Hoping it increases social aspect of it a bit more (will be doing more cool stuff around this later too) Heres one of it scraping the news sites and creating an outline for a podcast using the latest news from tesla app.cognosys.ai/s/gruPwp0

Spoke with a few tech-savvy friends who heard of chatGPT but barely use it.

Kinda surprised me, so I asked them why they don't use it more often.

The answer? They didn't know what to use it for.

And when I think about it, it makes a lot of sense:.

Let me explain: They basically told me that they didn't know what it was capable of other than "write me a paper" or "write my marketing copy".

I think as these models get more capable, they're going to be more confusing for the average person to use.

Google's new model, Bison, is here.

I decided to compare it to GPT4 and benchmark them in 4 real world scenarios, looking at: output, reasoning and cost.

Here's what i found: First up Coding: Here's the prompt I gave it:

Create the backend api for a flask application. The application should support the basic endpoints needed to create a netflix-like application.

First pic is gpt4, second is bison.

Googles Keynote 🤯 for AI. Here's a brief overview:

About 25 different AI products. Biggest LLM news

Palm 2
- 4 different sizes, gecko (runs on mobile) , otter, bison, unicorn

Palm med
- tuned for Medical purposes. Can analyze x-rays to help radiologists.

- Palm Security Bard getting A HUGE upgrade.

Its now fully running on Palm 2:
- Much better at coding
- Can search github urls
- Will cite code from sources
- Export python to Colab
- Export code to @Replit
- Interact with partnered tools/google apps directly in bard
- Support image prompts

Vector databases & embeddings are the current hot thing in AI.

Pinecone, a vector DB company, just raised $100M at ~1b valuation.

Shopify, Brex, Hubspot and others use them for their AI apps

But what are they, how do they work and why are they SO crucial in AI? Let's find out Ok so first, what are vector embeddings? You've likely seen this word thrown thrown out a billion times all over twitter

The simple explanation is:

Embeddings are just a N-dimensional vectors of numbers. They can represent anything, text, music, videos, etc. We'll focus on text

GPT4 32k is insane.. I gave it a 23 page congressional hearing and here's what it said: Ok but first, what even is GPT4 32k? Well it just means that it supports 32k tokens.

In simple terms thats roughly 24000 words, or 48 pages of text.

No embeddings, uploads, weird hacks needed. Just copy paste.

If you're using the GPT4 API and not using system prompts, you're doing it all wrong.

Why? Let me show you with an example, I am a cat: In regular chatGPT you're limited to a specific number of tokens, ~8k tokens.

I tested this with
1) Asking it to always reply with "i am a cat" at the start
2) Then sending it a bunch of aaaaaaaaaaaaaaa's

It failed on the 4th message.

Everyone is talking about OpenAI, Microsoft and Stability as a huge AI player.

But there is one company that arguable the best positioned in the space, and they dont even have an LLM offering yet.

Apple has the potential to shift the entire AI landscape, here's how: Ok so first, let's talk about a couple of important developments in the AI space.

Namely a few days ago a 19 year old genius literally made a version of chatGPT that can run on your browser, called webGPT.

Yes it runs inside of chrome. No API needed. It runs any GPT Model

https://twitter.com/998575387067142144/status/1649147091573432321

OpenAI's discord is filled with amazing prompts.

Unfortunately it's closed (full).

Here are some of the more interesting ones i've found (GPT4 - prompts in alt text) Codepup 3.0 - AI Developer

Update on cognosys: it’s been absolutely insane

We’ve hit nearly 70k visits in 4 days. Nearly maxed out everything in the paid version of @vercel. Had to frantically switch our endpoints to the edge (700gb hours for the first two days).

Fortunately edge is a lot cheaper.

AutoGPT, AI Agents are all the buzz. The hype is only getting stronger,

But you know what everyone isn’t talking about?

The implications it has to our society.

Let’s take a look: So by now sure you’ve played with a variation of AutoGPT, and you understand how powerful it is.

But we’re just scratching the surface.

For the last few days I've been building Cognosys - it's a web based version of AutoGPT/babyAGI

And I'm happy to finally launch the beta version today!

It's 100% free:
cognosys.ai I built this to make things easier for folks struggling to set it up locally

I'm hoping this makes AI Agents a bit easier to use

By 2025 Autonomous AI Agents will be in every aspect of life.

They've been out for a week - and already the possibilities are mind-blowing.

But what are they? How did we just figure them out and whats next for us?

I'll try my best to explain. Buckle up, its gonna be a douzey Ok so what exactly are agents? On a high level they are:

An AI system that, given a task, runs in a loop until the task is solved

Basically the AI gets assigned a goal, figures out what it needs to do to accomplish that goal (on its own), and then spawns more AI to do it.

AutoGPT is taking the internet by storm. Its everywhere.

They’re basically AI agents that run by itself, and complete tasks for you.

The best part is you can set it up yourself, and have your own 🤖 doing your bidding for you in less 30 minutes.

Heres how: Ok first up: You will need to know how to run terminal commands. Thats about it.

Then go and do these things :
- git setup
- download python
- download docker desktop
- Get an OpenAI Api Key

Here are the links:
sullyo.notion.site/Quick-links-fo…

Whoa.. still not convinced of AI Agents? This might change your mind...

I pretended to be a fake shoe company and gave AutoGPT a simple objective:
- Do market research for waterproof shoes
- Get the top 5 competitors and give me a report of their pros & cons

Here's how it went: First: It went straight to google to find the top 5 waterproof shoes reviews.

Once it found links, it created questions for itself like

"What are the pros and cons of each shoe"
"What are the pros and cons of each top 5 waterproof shoe"
"Top 5 waterproof shoes for men"