Gustavo Araujo Dunhão’s Post

🚀 Proud to share my latest project: PayFlow Pro! 📊✨ Managing payroll manually from messy Excel worksheets can be a headache. That’s why I built PayFlow Pro—a modern desktop application designed to automate project-based pay calculations and streamline HR workflows. Built entirely in Python, this tool bridges the gap between raw data collection and final reporting with a clean, user-friendly dark-theme UI. 💡 Key Features: 📂 Seamless Integration: Instantly upload and parse Excel sheets (.xlsx, .xls) using pandas. ⚙️ Dynamic Filtering: Map date ranges, filter project statuses, and assign individual pricing logic on the fly. 📊 Automated Excel Export: Generates multi-sheet professional reports, including a high-level "Global Info" overview and a structured "Summary Pay Sheet". 🖼️ One-Click Payslips: Automatically renders and exports individualized, presentation-ready PNG payslips into a dedicated folder for easy distribution! 🛠️ Tech Stack: GUI: CustomTkinter & Tkinter Data Processing: Pandas & OpenPyXL Image Generation: Pillow (PIL) Date Management: tkcalendar I’m really excited about how this combines data manipulation with a clean user experience to solve a real-world administrative bottleneck. Check out the source code on GitHub below! 🔗 GitHub Repository: https://lnkd.in/gscE-K8S I'd love to hear your thoughts or feedback in the comments! 👇 #Python #SoftwareDevelopment #CustomTkinter #Pandas #Automation #OpenSource #DataScience #PayrollManagement #GUI #PythonProgramming

Have you discovered the magic of "disposable code" yet? 🪄 As we roll out tools like Claude Code to our teams, I’m noticing a common hesitation. A lot of employees are intimidated because they think using these tools means they have to build a permanent, complex software application. They think they have to become software engineers. But that couldn't be further from the truth. There are actually two very different ways to build with AI: 1️⃣ Building an Application: You are creating a permanent tool or agent that you will use every day. (This takes planning and maintenance). 2️⃣ Task Completion (Disposable Code): You have a one-off problem. Maybe you need to extract data from 100 messy PDFs, or rename 500 files. You ask Claude to do it. Claude writes a quick script in the background, runs it, gives you the final answer, and the code is thrown away. You never have to read the code. You never have to save it. The code was just a temporary vehicle to get your task done. If you are intimidated by AI coding tools, start with a one-time task. You aren't building software—you are just getting your work done faster. #AIatWork #SAPTaulia

We are in the golden age of personal software. I am going to try and highlight projects that make me excited about how people are using LLMs to create exactly what they need. https://lnkd.in/gAMnZZ-c #software #ai

Claude Code just got a lot more autonomous. 🎯 New feature: /goal Instead of babysitting your AI coding assistant turn by turn, you set a completion condition and Claude keeps working until it's met. Real examples: ✅ "All tests in test/auth pass and lint is clean" ✅ "Every PR from this week has a CHANGELOG entry" ✅ "Split this large file until each module is under the size budget" How it works: → You type /goal [your condition] → Claude starts working immediately → After each turn, a fast model checks whether the condition is met → If not, Claude keeps going — automatically → When done, the goal clears itself The clever part? The evaluator is a separate model from the one doing the work. So completion isn't decided by the agent hoping it's done — it's verified independently. Pair it with auto mode and you've got fully unattended coding sessions that stop exactly when the job is done. This is what "agentic" actually looks like in practice. 🔗 Claude Code docs: https://lnkd.in/dCnGBVPi #AI #DeveloperTools #ClaudeCode #Anthropic #AIEngineering #Productivity

As someone using A.I every day, I find it hard to understand how careless a developer can be to let A.I do everything without reading or reviewing. You hear stories like "A.I mistakenly dropped my entire database." How?? Did you read or review the commands it gave? Or were you just relaxing? Because personally, I have used A.I at a time when I was doing some server optimization work and A.I almost wiped an entire server of about 50GB of data. Not even once, this even happened today too. Almost wiped my entire personal server hosting a lot of useful data. You can't be that careless as a developer. #tech #development #AI #devlife #techwithmartony #softwareengineering

SQLite shipped an AGENTS.md file this week, an explicit contract for how agents should interact with the database, not just an API reference. A document that says: here is how to use this if you're a model, not a human. Infrastructure is acknowledging it: the agent is a first-class consumer of your data and needs different affordances than a human does. We ran into this building Podium. The enrichment layer powers both Sage and Familiar, two products with completely different access patterns. Sage needs deep, personalized queries against a user's skin diary and product history. Familiar needs fast, broad queries across creator storefronts. Both are agents consuming product data. Neither pattern looked like anything we had built for human-facing interfaces. We rebuilt the data model before we rebuilt the agents. Every team I see struggling with agent reliability built the agent first and retrofitted the memory layer. When the consumer is an agent, the data model is the primary design constraint, not an implementation detail you get to later. If you're scoping an agent project right now, the question I'd start with is: what does the agent need to know, and how fast can it find out? That answer determines your schema, your memory layer, and which model you even need to call. #agentinfra #aiengineering #founders

Will AI augment or replace developers? I argue neither. The whole class of software developer will dissolve. Why? Because with AI as the author, we no longer have to cater for human desires to expressively write and understand code and instead can focus on rigorous mathematical foundations. That's what I've been build at Lume, the first AI native programming language and platform.

We just released a video companion to the first article in our FileMaker AI Foundations series. If you've been curious about where AI-assisted development is heading, this is the place to start!

Knowing how FileMaker handles code under the hood changes what's possible with AI. We published a video companion to the first article in our FileMaker AI Foundations series. It covers FMXML, the format FileMaker uses when copying and pasting scripts, fields, tables, and layouts, and the FM Clipboard Tool that puts it to work. The foundation for everything in the series that follows. Watch here: youtu.be/8vISOTOjvJs #FileMaker #Claris #FMXML #AI #SoftwareDevelopment #FileMakerDevelopment

Stop babysitting Claude Code. OpenAI just open-sourced Symphony — and it kills the "one human, one terminal" workflow most teams are still stuck in. The model: → Open task appears on the board → An agent picks it up automatically → Work runs in an isolated workspace → Code written, tests run, proof of work produced → Humans review the outcome, not the process → Work gets merged The shift nobody's pricing in: Teams stop managing terminals. They start orchestrating outcomes. Inside OpenAI, some teams hit 500% more landed PRs in 3 weeks. That's not a productivity bump. That's a different operating model. Same logic applies to Claude Code. Stop sitting next to it. Give it a ticket, an isolated workspace, and review the output. More execution. Same headcount. Clearer visibility. Faster path from idea to shipped. 📖 OpenAI: https://lnkd.in/gZFY4S55 💻 Repo + spec: https://lnkd.in/gsHk-myw