Can AI drop a summer banger, or is it better off sticking to data pipelines?

In this episode of Data Party, we pivot from serious architecture talks to put Google Flow Music to the ultimate test. Watch as we attempt to create a SAAS tool into a summer banger!

While AI can produce impressive one-shot generations, fine-tuning and editing tell a completely different story. Join the breakdown of:

• One-Shot Magic vs. Editing Reality: Why AI models excel at generating initial concepts but struggle with iterative tweaks.

• The Regeneration Dilemma: Why tools like Google Flow Music insist on re-creating whole tracks instead of editing specific sequences.

• Broader AI Limitations: How these music editing friction points parallel common LLM pitfalls in slide creation, graphic rendering, and code generation.

• Product Opportunities: Where the massive white space lies for developers to build smarter, sequence-aware AI editing tools.

Whether you're building AI products or just curious if an algorithm can handle the mixing console, tune in for a fun, real-world reality check on the limits—and future—of generative creative tools!

Welcome back to Season 2 of Data Party! 🥳

As a small business owner, preparing taxes is one of those tedious, time-consuming tasks that nobody looks forward to. Too many forms, too much sorting, and way too much manual work.

In this season premiere, we’re fixing that by building an autonomous Tax Agent designed to handle the heavy lifting!

In this episode, we cover:

• The 2-Flow Architecture: Setting up quarterly checks and sweep confirmations.

• Connecting External Services: How we integrated Plaid to fetch financial transactions smoothly.

• File Tree & Claude MD: Structuring project files and instructions for clean execution.

• Real-World Demo: Watching the agent categorize expenses, update from human feedback in real-time, and calculate tax estimates.

• Iteration & Prompt Evals: Honest lessons learned from building and debugging AI agents.

🐙 Grab the Code:

Want to set this up for your own workflow? Grab our open-source GitHub repository here:

https://github.com/Happychick

⚠️ Disclaimer: This podcast is for educational and building purposes only. We do not provide financial or tax advice—always consult a certified CPA for official filings!

Data Party: Season 2: A New Hope 🚀Ok, ok ok. It's been a while, and I have not being posting consistnFor the last two years, we've talked about data trends and industry movements. But the world has changed. Thanks to large language models and advanced evaluation tools, data professionals don't need to just provide insights and try to "influence" stakeholders anymore. We have the power to build end-to-end products ourselves.This season is all about moving from spectators to builders. No more endless frustrations—just pure execution.What to Expect This Season:Building E2E: How to take your data skills and turn them into functional AI products.Mastering LLMs: Practical approaches to leveraging language models in production.Rigorous Evaluation: Building the guardrails and evaluation metrics that make AI products actually work.Hit subscribe, get ready to build, and let’s get this party started. 💻✨New episodes drop every week!