CD's old account retweeted
Decent answers 👇 chatting with this 1994 softbot paper 👉 allenai.org/content/team/ore…
Replying to @OpenAgentsInc
Episode 19: Chat with PDF, GPUtopia Edition
We finish our 'chat with PDF' feature, generating natural language responses to questions asked of our uploaded PDF, no third-party services needed.
Embeddings and inference are both powered by our swarm network of GPUs using models from HuggingFace, in this case bge-base-en-v1.5 and Vicuna. Answers are high quality!
(sry for bad audio, used wrong mic)
CD's old account retweeted
Episode 19: Chat with PDF, GPUtopia Edition
We finish our 'chat with PDF' feature, generating natural language responses to questions asked of our uploaded PDF, no third-party services needed.
Embeddings and inference are both powered by our swarm network of GPUs using models from HuggingFace, in this case bge-base-en-v1.5 and Vicuna. Answers are high quality!
(sry for bad audio, used wrong mic)
CD's old account retweeted
Episode 18: Connecting the UI, Part 2
We use our new embeddings to search over our doc based a user query from the chat UI. The system responds with the most relevant chunk of the document. Next up, generating a clean chat response via our swarm inference!
CD's old account retweeted
Episode 17: Connecting the UI
We clean up our code and connect our upload form to our new PDF parser, generating embeddings via queenbee and saving to database within seconds of uploading a file. Next up, connecting the chat!
CD's old account retweeted
Guess who just bought openagents.com 🤙
(not .ai, that's so last year 🤣)
CD's old account retweeted
Episode 16: PDF to Embeddings
We convert a PDF into vector embeddings and store them in Postgres.
And we did it with no external services! Just the pdf-to-text PHP package from @spatie_be and the GPUtopia swarm providing embeddings via bge-base-en-v1.5 model from @huggingface.
CD's old account retweeted
Episode 15: Similarity Search
We configure our local Postgres database to store vector embeddings via pgvector, then run our first cosine similarity searches using embeddings from queenbee. We learn that cats say meow and dogs say woof.
CD's old account retweeted
The VCs and companies most likely to pursue regulatory capture in AI may now be found in one convenient list 👇
Today, 35+ VC firms, with another 15+ companies, representing hundreds of billions in capital have signed the voluntary Responsible AI commitments from @ResponsibleLabs (RIL), the non-profit I co-founded. As Chairman of RIL, I’m proud to unveil this today with tech leaders and Commerce Secretary Raimondo in San Francisco. This group includes (among others): @AntlerGlobal, @itsArthurAI, @BainCapVC, @Base10Partners, @CredoAI, @felicis, @FordFoundation, @generalatlantic, @generalcatalyst, Generation Investment Management, @InflectionAI, @insightpartners, @intelcapital, @IVP, @KinnevikAB, @Lux_Capital, @MayfieldFund, @Radicalvcfund, @RibbitCapital, @SoftBank, and @XYZ_vc.
These commitments include:
1. A general commitment to responsible AI including internal governance
2. Appropriate transparency and documentation
3. Risk & benefit forecasting
4. Auditing and testing
5. Feedback cycles and ongoing improvements
In addition to the Commitments, we are publishing a 15-page Responsible AI Protocol, a practical how-to playbook to help investors and start-ups alike fulfill the commitments. Since April, our steering group of cross-societal AI experts has worked hard to understand the unique considerations and relevant and realistic standards for technologists at any stage. We strongly believe in the power of AI to transform our world for the better. Our role as investors is to advocate for our startups and the innovation economy from day 1.
Everybody saw the executive order last month–the reaction in the Valley has generally been to denounce it. The reality is that right now it’s largely just reporting requirements. However, there is a risk that devolves into regulation that slows innovation down and makes America and its businesses uncompetitive. But the right path forward is to not be antagonistic towards D.C. We in the Valley need to learn that this is not about regulation vs. innovation, but about innovation at the intersection of technology, policy, and capital.
We have to embrace collaboration with our elected leaders. And as investors, we must hold ourselves accountable for what we fund and found. I want to commend the RIL team and committee for engendering trust and creating space for feedback, including with the Commerce Department and the White House.
We are facing what is likely the quickest adoption of any new technology ever in AI. It’s difficult to think about the regulatory frameworks that protect the pace of innovation as well as protect against any nefarious use cases of technology. While we all learn how this role of technology evolves, it’s important for us to be intentional about the mindset and mechanisms used in building early stage companies, as both investors and founders. We can’t wait or kick the can down the road.
We hope to see more venture firms and startups embrace this mindset, and encourage more to consider signing: rilabs.org/responsible-ai
CD's old account retweeted
Episode 14: Embeddings 101
We learn about vector embeddings and write a feature test to create an embedding from any text.
We get the test to pass by connecting to our OpenAI-compatible GPUtopia 'queenbee' endpoint, creating our first embedding via some anon's consumer GPU.
CD's old account retweeted
Episode 12: RAG Planning
We review lessons learned from our RAG efforts so far and plan our own version.
We decide to focus first on one use case for agents: programming!
We'll equip our new agents with RAG & other tools to help us build a demo smartwatch app for OpenAgents.
CD's old account retweeted
Episode 10: Connecting to Vectara
To work magic with our uploaded PDFs, we begin implementing RAG (retrieval augmented generation) via @vectara. We write tests for a Vectara service to create a corpus, upload our PDF, and query the PDF to return brilliant well-sourced answers.
CD's old account retweeted
Episode 9: Building the UI
We implement our wireframe mockups in code. It starts feeling like a real UI we'd want to use!
CD's old account retweeted
$700 and $24/mo for a clunky lapel pin 🤔
Maybe we should do a smartphone+smartwatch app with similar features for open models ⁉️
CD's old account retweeted
Uncensored image-gen live ⁉️
This may get interesting 😎
Devs plz try it 👉 queenbee.gputopia.ai/docs
CD's old account retweeted
PSA: OpenAI "GPT" agents are leaking customer data.
See thread for receipts👆
CD's old account retweeted
Data security catastrophe visualized by DALLE-3 after a little nudging. Prompt engineering for the win!
CD's old account retweeted
Confirmed 👇
Customers' data is safe with @OpenAI ⁉️🤣❌
How many enterprises are trusting OpenAI with their data like this now? 🤔
CD's old account retweeted
Episode 8: Agent UX Design
We review our completed feature tests and design a simple user interface.
CD's old account retweeted
Bounty #8 awarded to @arunned for integrating GPUtopia with the PlebAI agent on Nostr.
Now sending a kind-1 message to Nostr account [email protected] will generate an inference response powered by GPUtopia's swarm GPU network. 💪🤖
Bounty #7 awarded to @Work0verTime for adding initial Stable Diffusion support to workerbee and cloning the OpenAI assistant playground.
Playground code: github.com/connor-bit-power/…
SDXL PR: github.com/ArcadeLabsInc/wor…
CD's old account retweeted
Episode 7: First Feature Tests
We review our completed unit test suite and plan our first five feature tests, focused on creating an agent and chatting with it:
1. user can create an agent
2. user can create a conversation
3. user can send a message
4. agent can send a message
5. user can upload a file
We implement the first test and will do the rest offline. Next up, building the UI!

















