Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
42 contributions to AI Marketing Hub
🆕🍌 Banana Claude v3
I rebuilt banana-claude for the current Gemini image stack. What it does: - Generates, edits, compares, continues, and reviews images - Uses Nano Banana 2 Lite, Nano Banana 2, and Nano Banana Pro - Shows the exact prompt, privacy settings, and estimate before any paid call - Creates consistent characters and bounded multi-model portfolios - Adds exact copy locally when generated text is not reliable enough - Can generate A LOT of images in parallel and FAST! Repo: https://github.com/AgriciDaniel/banana-claude Release notes: https://github.com/AgriciDaniel/banana-claude/releases/tag/v3.0.0 Open source. Test it, break it, and tell me what needs improving.
2 likes • Aug 30
Thanks @Daniel Agrici 🔝
2 likes • Aug 28
Thanks @Daniel Agrici 🔝🥷🔥
this AI agent uses chrome like I do
One voice call. Three browser agents working at the same time, in real Chrome windows. What runs in the video: - Keyword research on DataForSEO - A calendar event created end to end - A GEO audit across three sessions - All three in parallel, from one voice line The only thing I type in 16 minutes is one button. Underneath it is the Secretary skill. It freezes the task and its bounded evidence before anything runs. The child model never gets file tools. Every result is validated locally. It stops before any action that needs your approval. The brain it reads is manifest first. 13 files and about 6,800 tokens per run instead of 58 files and 58,000 tokens. That is the difference between an agent that works and an agent that runs out of context. Watch: https://www.youtube.com/watch?v=hFILyRFRFdk Secretary: https://github.com/AgriciDaniel/secretary Phone: https://github.com/AgriciDaniel/nokai Free and open source. Build with it. Break it. Tell me what you find.
3 likes • Aug 17
Really interesting approach @Daniel Agrici , especially the manifest-first brain — that context math is the part most agents get wrong. Saving this to try it out this week. I'll let you know what I find.
Anthropic watermark and marketing
NOTHING CHANGES. Google's synth-ID has been doing this since past 2 years. Nothing here will change if you're using best practices and the right tools. F#ck Neil, even if he is popular..."marketers who rely on AI content won't do as well" or "ai content does not perform so well" is unnecessary bullshit, that's just his opinionated message dressed up as fact. be careful of human slop, this guy is one big part of it 🗑️
Anthropic watermark and marketing
2 likes • Aug 12
Totally agree @Daniel Agrici . AI isn’t the problem. How you use it is. Good judgment + the right tools + best practices. Everything else is noise. 🎯
Claude Code + Computer Use
Spent time digging into how Claude Code works with computer use, browser automation, and MCP servers. Covers official support, sandboxing, security risks, failure modes, and headless setups. Ends with a concrete stack for marketing automation. Full research here: https://claude.ai/public/artifacts/58df8b85-75fc-4e5c-951b-f17a2b68b0d4 Worth a read if you're building anything with browser agents.
Claude Code + Computer Use
4 likes • Jul 4
Thanks @Daniel Agrici ✨
1-10 of 42
Joan Marquez
4
67 points to level up
@joan-marquez-5213
🚀 Co-fundador de 🧰 Tool Box | IA, Automatización y Crecimiento 📈 | Transformamos procesos manuales en sistemas inteligentes.

Active 33m ago
Joined Dec 18, 2025
Madrid
Powered by