Vito Compagnucci

Argentina don’t run

Looks clean, here some feedback: 1. The animation can't be interrupted mid-play. If I try to close it while it is still running, I have to wait for it to finish before it reverses. It should be interruptible at any point. 2. Expanding the last notification clips its content against the container.

Built a Notification Component (using @motiondotdev) LIVE link in comments

wow

@teo_zaragoza Muy Muy bueno

Tuki 🏴

A new era begins. It’s your time to become an investor.

@jakubkrehel Quoting Leonardo da Vinci hits hard

Less is more, more or less jakub.kr/writing/less-i…

@emilkowalski Amazing!

Given how impressed I was with Cursor’s Thermo-Nuclear Code Quality Review skill, I decided to create a similar one, but specifically for animations. /review-animations to review, and /emil-design-eng to build animations. They complement each other well emilkowal.ski/skill

@jackfriks How long does it take you to ship, approximately?

i just hit $37,000/month on my solo startup, so now i can call it a success i think but it got me wondering... how many times have i failed so far? and what's my success/hit rate? 📋 23 projects since 2019 🫙 18 misses ❌ 9 abject failures ✅ 5 HITS!!!

Consumer —> Builder

⚽️

@conductor_build 😍

Even tech companies can name things plainly.

@joshpuckett @F0ODHub

@F0ODHub for me, with that chimichurri, I'm thinking Malbec from Argentina, and would look no further than the '21 Catena Zapata from the Adrianna vineyard (after his daughter). An absolute stunner.

What drink goes best with this?

the internet is the greatest playground ever built. if you’re not having fun on it you’re fundamentally misusing it.

I think the world is moving to agent inputs and agent outputs. Ui is there in the middle as shared source of truth and own way to use the system like it always have. Reality there is that the UI or its functions didn’t disappear. All of it’s still there. What we replaced is the new tab view. When you created a new tab it would open my issues or inbox or whatever was your default which isn’t that useful. Now this box allows you ask anything about the system or codebase, do anything the system can do and more. Create issues, projects. Check status of the whole roadmap, ask it explain how a feature of your product works based what’s in the code. Here is one example of me figuring out a product feature with it.

@nickbakeddesign icon library?

modals

it was amazing, thanks yall! had no pictures, but it was wonderful.

design the arena, let AI iterate.

@santtiagom_ x.com/vitocompagnucc…

¿Cómo funciona un AI Assistant como Claude Code? Primero tené en mente esto: 1) LLM ⇒ modelo que predice texto (por ejemplo, Claude Sonnet 4.5). No puede ejecutar código, ni abrir un archivo, ni consultar una API. 2) AI Assistant ⇒ el sistema que ves e interactuás (Claude Code). 3) Tools ⇒ funciones externas. Por ejemplo: abrir un archivo, consultar una API, editar código, etc. Entonces: ¿qué pasa cuando le pedís a Claude Code que te diga tus próximas reuniones? 1) Claude Code recibe tu prompt: "decime cuál es mi próxima meet". 2) Claude Code le pasa al LLM tu prompt + las tools disponibles (por ejemplo, una función para consultar la API de Google Meets). 3) El LLM no ejecuta nada: predice qué tool necesita, por ejemplo: "llamar get_meets()". 4) Claude Code ejecuta la función real (va a la API y obtiene tus reuniones). 5) Claude Code le devuelve todas tus reuniones al LLM. 6) El LLM genera la respuesta final para el usuario. Lo importante acá es entender que el LLM no puede hacer nada por sí solo. No puede ir a Google Meets y ver tus reuniones. El AI Assistant es el que conecta todo y le dice al LLM qué tools existen. Ejecuta esas tools cuando el LLM las pide y le devuelve los resultados. Sin el AI Assistant, el LLM solo genera texto. Esta arquitectura: AI Assistant + LLM + Tools es la base de casi todos los productos de IA que ves hoy: ChatGPT, Claude Code, Cursor y muchos más.