Introducing
#AgentGPT
, an attempt at
#AutoGPT
directly in the browser 🤖
Give your own AI agent a goal and watch as it thinks, comes up with an execution plan and takes actions. Try for free now at
The
#AgentGPT
team and I are launching our Beta V2.0 🌟
Its been so awesome to see all of the community support, we've recently just surpassed 15k stars on GitHub.
#AgentGPT
just hit 10,000 stars 🌟
Its been a wild ride this week watching all of our metrics double nearly every day. We've been hard at work with the following:
- Scaling our infra to 100k DAU without breaking our bank
- Local OpenAI account runs
- UI improvements like a view
It works through using models to generate a task list and then iteratively executes tasks, evaluating whether or not tasks are completed or require further sub-actions.
In the future, we'll have long term memory via
@pinecone
and give models the ability to query the web
Super excited to announce
@ReworkdAI
has raised a $1.25M pre-seed ⚡️
We're incredibly lucky to be backed by both
@ycombinator
and
@PanacheVC
in this round
⚡️ Chat with your AI Agents with
@LangChainAI
and
#AgentGPT
in our v0.9.5 release!
Guide agents by transforming and updating results at any point of execution. Heres an example of chatting against an agent that created a travel itinerary ✈️
⚡️
@perplexity_ai
has some of the smoothest text streaming out of all LLM apps
As a weekend hack, I recreated it with the
@vercel
AI SDK and
@nextjs
. Try it for free with
#GPT4
at
Been a wild few days with
#AgentGPT
: 300k impressions across socials, 30k unique visitors, 1k GitHub stars, and more 🚀
Now reaching limits with
@OpenAI
, would be dope if anyone could help us get access to GPT4 or increase our usage limits. Please DM me if you can help :)
Of course its a bit limited in terms of capability at the moment but we're working to metaphorically give AgentGPT arms / legs / and eyes.
You'll soon be able to browse the web, deploy software, and so much more
#AgentGPT
recently hit 20k stars on GitHub 🚀
Its been an incredible ride watching this project grow. A big shoutout to everyone that's helped contribute the past few weeks
#AgentGPT
just got tool support in our beta 0.5.0 release 🚀
Watch as your agents search the web 🌐, generate images 🎨, and write code 💻
Heres a demo of an agent attempting to design a platformer game 🎮
⚡️ gpt-3.5-16K 🤝 Functions ⚡️ in
#AgentGPT
's Beta 0.8.0 release
With
@OpenAI
's latest changes, users can experience blazing fast AI Agents with increased model context and correctness.
Try it for free at
My welcome to YC moment was doing an informal talk during the batch retreat and seeing none other than
@garrytan
himself listening in the audience
Just the other day he did an interview with
@lulumeservey
and talked about our story
#AgentGPT
just hit 20,000 members on our Discord channel ⭐
With this, we're launching Beta 0.4.0 of the platform! Check out the full release notes here:
Update: We're hitting scale limits on both Vercel and OpenAI. We've just added the ability to use your own API key on the platform. Unfortunately, this will be a requirement just until we can get scaling sorted out!
⚡️ AI agents supercharged with your data ⚡️
Introducing Vector databases with
#AgentGPT
's Beta 0.7.0 release thanks to
@LangChainAI
. Currently using
@weaviate_io
by default but we're exploring providers!
Been a while but excited to show off a small AgentGPT 1.0 update ✨
Features:
- A freshly redesigned theme 🚀
- Updated agent handling with a focus on search 🌍
-
@try_sid
for personalized context and queries 🦥
- Increased usage limits 📈
Try it at
When we first launched AgentGPT, I had to beg
@OpenAI
to raise our rate limits everyday to meet increasing demands
They eventually got us to shut up by giving us an absurd limit 😂
Feedback has been immensely positive, thank you to everyone that's reached out ♥ Now that we've alleviated our infra worries, our team will be focusing on the following in the coming weeks:
- Web browsing
- Connection to productivity tools
- Multi agent support
- User accounts
Not gonna lie, ended up subscribing to Twitter blue in hopes of increasing outreach. Twitter has ended up being the best referrer by far, cheers
@elonmusk
If you'd like to help contribute, discuss plans / features, or would otherwise just like to stay up to date with progress, head over to our discord server @
Excited to share the start of Bananalyzer 🍌
A simple first start to our AI Agent evaluation library for web tasks. We're aiming for a rich suite of repeatable web tests for any agent.
Read more here:
@MomenticAI
@ReworkdAI
- web data extraction using multi-modal agents. many use-cases including data ingestion, relevant to chroma. seems like their agents work!
⚡️
@perplexity_ai
has some of the smoothest text streaming out of all LLM apps
As a weekend hack, I recreated it with the
@vercel
AI SDK and
@nextjs
. Try it for free with
#GPT4
at
Finally, we've added some web searching capabilities. This is however in limited preview so expect it to be unavailable, but get a lot better shortly. Here's a sneak peek
@LOFI911
Hey unfortunately just a limitation of OpenAI throttling our server. The project is completely open source so if you want, you can get it running locally (for free) and use your own key.
Our infra bill is projected to be incredibly high so we can't have huge runs on our own key
With that being said, we're still proudly open source and you're free to run locally for free! (Though we do need to do some work to fix up local setup)
Heres our GitHub release notes:
We're also launching a pro subscription with access to GPT-4 to finally help cover our OpenAI bill 😅
Use code TWITTER25 for 25% off! (Only available for the first 100 users)
Of course AgentGPT is still very much in beta. We have a lot planned and your support means a lot.
Companies you didn't know were
@ycombinator
a 🧵
1.
@9GAG
👀
They we're founded in 08 and joined YC in S12. A large part of early meme culture- wasted a lot of hours here during high school but feels largely forgotten now
@SagoLabCom
This was actually originally the idea the team and I explored. The space is definitely saturated and we ended up wanting to solve more ambiguous problems that would be hard to map into a structured flow
Introducing
#AgentGPT
, an attempt at
#AutoGPT
directly in the browser 🤖
Give your own AI agent a goal and watch as it thinks, comes up with an execution plan and takes actions. Try for free now at
Making LLMs return YAML responses is super nice for UX 🙏
Unlike JSON, their intermediate representations are easily parsable during streaming. The same request with JSON would typically force users to wait for the entire duration of the call to complete
@mr_8_bit
24 year old recent-ish grad here. I'd say if you just build cool enough stuff the companies worth pouring your life force into will come to you
Wouldn't focus on LC that much. In terms of prioritization, just depends on what you want to optimize for. Feels like finding balance
@ronawang
While it shouldn’t be looked at as a status symbol, I’d still recommend it to any young first time founders
Valuation at that early a stage (perhaps in general) has no bearing on the success of your company. The mentorship and network effects of YC however can actually make a
@chipro
We're essentially working on no 4.
Systems to extract structured web data at scale with LLMs. We work with companies today to help them collect foundational data for data constrained product offerings
@Suhail
We did. Previously a single user could cost us up to $40 a day if they hit our site's limits but we were ok with just eating the cost
@perplexity_ai
and
@phindsearch
definitely have too
Thanks to our users for bearing with latency issues. Below is a look at our sentry metrics across our completion routes 📉
Some users waited up to 60 seconds for some completions - far beyond what's really acceptable ⏰
Shoutout
@ajndkr
+ lanarky and
@FastAPI
@DYtweetshere
@scale_AI
@evanon0ping
@theappletucker
Congrats on 1st place! I certainly think LLMs are going to completely disrupt the software engineering flow.
Security for a backend as malleable as what's proposed would be a bit concerning though given how much power this gives to agents performing prompt injection.
Beyond that, we :
- Introduced support for
@supabase
's vecs library
- Migrated our Vector store to
@pinecone
which has worked like a charm
- Styling / markdown formatting updates
For the full change log, check out our post on GitHub
Welcome to YC,
@SrijanSubedii
,
@asimdotshrestha
, and
@ReworkdAI
!
Reworkd AI (YC S23) allows companies to automate core business workflows with the help of AI Agents, all while integrating seamlessly with their existing tech stack.
It's still in its early stages - we're starting by using Vector search to filter out repetitive tasks 🗑️
However, you'll soon be able to create agents with context over whatever data you want 🗄️
My girlfriend has something like a 95%+ win rate in scrabble
Because of this, last Christmas I made an app where she could play against the best scrabble player she knows (herself). Try it at
⚡️
@perplexity_ai
has some of the smoothest text streaming out of all LLM apps
As a weekend hack, I recreated it with the
@vercel
AI SDK and
@nextjs
. Try it for free with
#GPT4
at