Que pedazo de evento se viene este viernes 🔥 Keynotes de Pablo Zivic, Ivanna Figueroa, Ale, Shroominic, Fardo, Manuel Soria, Francisco, Lautaro Gesuelli, Lucas Petralli & more en las oficinas de Lemon Argentina 🫡 📌Full schedule: llmnative.dev/events

1 screenshot current version 2 paste in excalidraw 3 use as scratchpad (notes, design ideas, move stuff) 4 paste back in v0 works like a charm

playing around with the new AI SDK 5 aiming for image consistency + fast inference ⚡️ also - I just love non-chat AI inerfaces, feels more like magic ✨ v0 🤝 fal -> 🔥

flat $20 plans with subsidized unlimited use probably never made sense in the first place. credit-based pricing just works, it's always been the way for products built around prodiver APIs (hence early focus context engineering & RAG for token efficiency and making low-cost &

We hosted another LLM-Native meetup at Lemon Argentina — a space where we bring together founders, engineers, researchers, and AI enthusiasts building frontier products with llms. Full house again, with 200+ on the waiting list. We dug into real-world technical challenges through 8

I would enjoy more reasoning models if they actually challenged my reasoning. I’d prefer to brainstorm solutions to complex problems. it’s promptable but LLMs are obsequent-by-default; and good solutions come from asking the right questions, not prompting the right instructions

Sharing an overview of the trace reviewer we developed at FieldData , thank you to the LangChain team for the feedback! Some key points: * soft clustering of traces (promptable) * reasoning model running in bg * hybrid UX for discovery Code and more details in 🧵:

vibetyping > vibecoding you should never backspace when interacting with llms - just vibe type this is quite interesting considering how LLMs work by tokenizing words; they must see a lot of typos in their training data so as to understand typos in cases a human would struggle

built this to solve the problem (?) of needing to know the time while patiently waiting for it to aesthetically show up. ok - I did it just for fun. it was interesting to see how AI struggled with the pattern coordination, so it hooked me in to solve it clock-inception.vercel.app

I started heavy using v0 and love it - I now share my first template as ambassador 😎 AI SDK 5 integration on Kibo UI components - audio input, model switching, showing sources from web search + more I hope some may find it useful! v0.dev/community/ai-s…

funny how one guy in the openai stream said "smarter, more powerful, and faster than any other model" that faster felt weird - and man gpt-5 takes forever. of course I take prompt -> 2-min -> ready-result over more iterations, but still

Done, was quite simple actually given the complete docs + minimal changes on AI SDK side - a fairly simple task to test gpt5. Kibo UI elements -> AI Elements🤝 v0.dev/community/ai-e…

I don't want the cursor for XYZ I want the resend for XYZ set it up once, works forever you just forget it's part of the stack - no annoying marketing emails - works like a charm - minimal (~0) maintenance

this was a game changer for how I integrate v0 in my own existing private repos. local repo -> v0becode -> back to local - no need to use github as middle man - no UI file uploads - one-step open in v0 it was the missing piece in my workflow, so I built it using the v0 SDK.

this is where we are going. I will see Bilbo on ads instead of a random labrador. image gen quality is no longer a constrain, only latency and cost; both can be solved. but then - if EVERY ad on social media is hyper-personalized, will they be effective? food for thought

after weeks of forcing Dia I just gave up because of this (which btw didn't happen before) a browser where you can't paste and open localhost url 🤡 for some reason instead of just opening the url it forces a chat makes me miss good old days when chrome was all we had