Twitter Beğeni Hilesi

Would recommend muting "OpenAI" for the time being; nothing new just PR drama, maybe due to weak followup after gpt-4o release and Google IO reveals. Btw really enjoying the hours of quality IO videos being released right now on youtube. Check them out!

I suspect every concern we're having in public about AI, has been debated 100x over behind closed doors by devs working at frontier companies. So why would they willingly continue? Hopefully in exchange for cooperation, they were allowed to implement an AI kill switch of sorts 🙏

OH my son explaining AI: "You see, with programming you need a lot of IF statements, but with AI, you don't need any IF statements." Having built a couple of actually useful AI agents in the past few weeks, this is remarkably accurate.

Hot (half-serious) take: Your devs are slowing down your AI ambitions - whether intentionally or not. AI has outpaced many things, software developers included and it shows. You need product people with vision and a desire for change. Sadly, your dev team are not it.

I put together a mini-tutorial on supporting upserts for vector stores in n8n. community.n8n.io/t/enabling-ups…

Are ppl really building agents but doing the parsing and decision-making upfront? Feel this completely misses the point of being autonomous and wouldn't a basic llm chain work better? My litmus test is: goals rather than tasks and tools can be situational? if not, stick to chain.

As someone who's guilty of throwing LLMs at every usecase, I've been massively humbled by Google's Document AI recently. For OCR, it's a magnitude more faster, accurate and consistent. Hella confusing to navigate setup though, like wtf is up with service design at Google AI?

I wrote another n8n tutorial about integrating cohere's Rerank in your LLM chains. Bit on the advanced side but makes sense once you realise it's all Langchain underneath. community.n8n.io/t/build-a-cust… #lowcode #automation #n8n

AI Agents have total disregard when it comes to rate-limits, especially with web crawl/scraping tools. Bots will go bananas trying to download entire sites if you let them. With the amount of workflows I'm building, I need a frugal agent who can keep me on the free tier😅

PDF Parsing is a hot topic for LLMs and so far I found LlamaParse by LlamaIndex 🦙 to be the best tool for the job - especially for table data. Check out my blog post integrating LlamaParse with n8n.io to automate incoming invoices! #ai #automation blog.n8n.io/how-to-extract…

Having spent some time with the automation community, I feel there's no real urgency for more capable LLMs; if you just need to contextualise pdf, "gpt5" likely performs the same as gpt4. IMO, the real needs are models with (1) cheaper pricing (2) with higher limits.

Join Community member jim (He/Him) as he showcases the power of the new HTTP Request Tool for AI workflows. buff.ly/4dcybgP

Clustering Text Embeddings is incredibly powerful... and suprisingly fun! Learn how with my free n8n template extracting insights from survey responses. #qdrant #n8n #embeddings #ai n8n.io/workflows/2372…

Founding Designer, Max Tkacz, embarks on an ambitious #30DayAISprint to build with AI and I'm excited to see what he comes up with! Didn't notice at first but check out his Youtube Video to get 50% off n8n cloud 🤯 youtube.com/watch?v=s--VFt… #ai #automation #n8n

Qwen2-VL (qwenlm.github.io/blog/qwen2-vl) is seriously blowing my mind right now. Gave a known problematic PDF I use for testing and it handled data extraction and markdown conversion flawlessly... and fast! Now I just need to figure out to optimally self-host 🤔

Many thanks to Bart Veldhuizen, community manager @ #n8n team, for the swag! Really enjoyed sharing at the last hangout and looking forward to what you folks have in store for the next! (which is on 26th Sept btw if you haven't already, you should definitely sign up!)

Looking forward to trying out OpenAI's o1-mini model for automations. The "reasoning" delay is mostly a moot point when it comes to non-chat workflows and the improved response accuracy is sorely needed.

👀 Hilarious looking at o1 pricing and back at this comment I made in July. o1 is exactly the opposite; ridiculously expensive with impracticable usage limits. We need LLMs to serve the needs of the working (wo)man, not the academic elite! /s

This workflow by Jim Le takes customer inquiries, helps them book appointments, and sends follow-up messages to re-engage leads—all automatically! Perfect for small businesses looking to scale their sales efforts. Check it out here and start automating your appointment