ansonhw

chatsheet.com

no embeddings yet, but soon

Ya this is nice that it’s infra agnostic instead of requiring tons of dependencies.

down again for me

not sure why i didn't see any status updates

was down (messages not going through) for me, then web was ok, then all went down, now back up

excited for this to run sandboxes

we're working on this at chatsheet.com

what features are you looking for?

Agree that's what we're trying to do at chatsheet.com

Program in "plain english" but as powerful as writing code

i stopped manually updating airtable and excel. all doc/web/pdf flows auto update.

Doesn't this mean 45% of workers already do use AI chatbots? Glass half full vs empty

Save Worker 17 o7

tech world: rag, llm, prompting, agents, bla bla

normies: chatgpt

Hmm i'll check out the pagerank stuff, IMO RAG isn't super great. I think RAG overall has been oversold since embeddings KNN hasn't proven to be super accurate.

Go is the best for scale out, concurrent agents

Gemini

Hmm i think the issue with any large data set is indexing + querying. At https://usefind.ai/ this is a fundamental problem we've been trying to tackle. What kind of structure do you think would work?

I'm confused why search + LLM wouldn't work for this?

NGMI

I agree actually that chat is overrated overall as UX. It works really well for chatgpt but creates the wrong ux expectations for users where more precision or constraint is needed. Also not good for large processing. It was magic though for chatgpt.

Does it support pgvector over weaviate?

why is bill gates so obsessed with bringing back clippy

Check the screenshot below re: training on OpenAI Outputs. They've fixed this since btw, but it's pretty obvious they used OpenAI outputs to train. I mean all the Open AI "mini" models are trained the same way. Hot take but feels like the AI labs are gonna gatekeep more models and outputs going forward.

https://x.com/ansonhw/status/1883510262608859181

Feedback -> paying Scale AI to babysit/label tons of these sessions.

It will just be like the post training that turned GPT3 into the original ChatGPT.

When: 1/ More accurate + less supervision 2/ Faster 3/ Background + Parallel

Will be the magic moment. Currently like supervising a grandparent using the web. But there's huge potential if ^^^ happens. Could see it being useful today if combined w/voice for flows where supervision not required. Example: asking w/voice to make reservations while driving.

sick

good batching and tensor parallelization prob

Hope not - we just rewrote our entire stack w Go . We're massive fans of it.

Earlier this year, we launched Find AI v1:https://news.ycombinator.com/item?id=40801494. It was a tool for searching LinkedIn-style data with natural language queries. It worked well for tech-focused searches, but the dataset was small and queries were limited.

With v3 (yes a 2 version improvement), we’ve overhauled everything. Our new algorithm uses deep research agents combining multiple data sources and can search over 1B+ profiles (more than LinkedIn). The new version also handles more nuanced searches, like:

- For selling GPU Cloud services: “Companies that have done a Series A hiring for Pytorch”

- For SDR’s: "AI companies that raised $20M+ Series A."

- For VC Partners: "Ex-OpenAI employees founding 'stealth' startups."

- For IR Professionals: "Head of alternatives investing at large private wealth firms."

- For Recruiters: "Fullstack engineers experienced with Go and Svelte."

We deleted all our old code (Ruby on Rails) and rebuilt the entire stack with Go + Svelte 5 (which we love) for speed and scalability. It’s now 10x faster, with a new UI and features like threaded searches.

We’re also launching an API, so you can integrate Find AI into your own apps. The API has also launched as an integration in Clay—learn more here: https://www.clay.com/university/lesson/find-ai-integration-o....

We started building Find AI to help founders, recruiters, and investors find the right people faster. Try it out here: https://usefind.ai/.

Would love your feedback and questions!