Week 7 : Get feedbacks sooner as you can
From AWS to hackathon, it was better to beta first here's why
Hi Madshippers,
Week 7 just wrapped up, and if there's one thing I’ve learned the hard way, it’s this :
👉 Get feedback as soon as humanly possible.
But before to start, let me introduce you quickly what you’ll find on this edition :
From AWS Summit 2025 highlightings to Hackathon momentum
The pricing and cost reality behind AI
Why beta is better than perfect
The next AI Agent Hackathons in 2025
Ready ? Let’s g(r)o(w) !
I used to think I needed to get everything “right” before showing anything. Polish the edges. Design the perfect flow. Build the thing in silence until it sparkled.
But this week reminded me : progress loves friction. And friction comes from real-world collisions with real people.
From AWS Summit 2025 highlightings headaches to Hackathon momentum
Watch the replay of keynotes here
This week, I got my FastAPI backend running locally for my first AI agent : the Content Analysis Agent. Think of it as a little brain that scans educational content and flags what’s missing, outdated, or redundant.
Building locally felt cozy… until I needed to connect it to a frontend and actually use it. And of course, nothing works until everything works.
At first, I was hesitant to open it up, imposter syndrome mixed with first project thing.
But as soon as I shared a demo, feedback poured in. Not just "cool project" but real, actionable suggestions. From interface quirks to performance lags, people helped me see what I couldn’t from inside my tunnel.
Same thing happened at the Hackathon : I wanted to build the perfect adaptive learning path engine.
But magic didn’t happen until I stopped tweaking alone and started co-creating. Shared the deck. Got roasted. For good. Iterated. Start to rebuilt better. At the point I did an incomplete repo on the hackathon. Also for good, theres others coming !
And ask myself questions I don’t even thought about in a scalable vision as
The pricing and cost reality behind AI
Let’s talk money and rentability, because AI doesn’t run on fairy dust.
Even before you launch, the cost of building and running AI tools stacks up fast. Between hosting, inference, storage, APIs, and time, you can feel the drain before you even get your first user.
Take this week for example: I had a backend working locally with FastAPI, but as soon as I needed to host it for real users (or even for myself on a public link), I had to start evaluating:
Render for backend hosting
Lovable for the frontend
Possible add-ons for OAuth, storage, database, etc.
Plus monitoring and logs if I wanted to sleep at night.
Even using open-source models, compute is never free. And if you're plugging into commercial APIs (OpenAI, Claude, ElevenLabs, etc.), the pricing becomes a real part of your build strategy.
Lesson? Your tech stack is a budget.
Every feature, every endpoint, every decision has a cost.
Not just in money, but in time, complexity, and sometimes… your peace of mind.
So I’m learning to be strategic, frugal and honest with myself :
👉 Which alternatives free existing to limit cost at a scalable perspective ?
👉 What’s essential now?
👉 What’s nice to have later?
👉 What’s just shiny?
Sometimes the most powerful AI product is the simplest one that’s cheap to run, easy to debug, and fast to test.
Because if your infra bankrupts your runway, the smartest agent in the world won’t save your project.
Why beta is better than perfect
You know what testing early does?
It saves your energy.
It corrects your illusions.
It gives you speed.
Every moment you wait is a moment you delay your growth curve. Every silent day is a missed opportunity to be challenged, inspired, or supported.
I used to be scared of feedback before entrepreneurship. Now I crave it. Because the truth is, feedback isn't judgment—it's a gift.
Bonus : if you want to accelerate and learn by doing fast fast fast here’s the
next AI Agent Hackathons in 2025 :
👉 Microsoft AI Agents Hackathon 2025 (Apr 8-30, Virtual)
Learn from 20+ expert-led sessions and build cutting-edge AI agents using frameworks like Semantic Kernel and Azure AI Agents SDK. Submit by Apr 30 for a chance to win cash prizes !
https://lnkd.in/e_md8xTz
👉 Global Agent Hackathon (April 15th-May 30th 2025 - Online)
Welcome to the May 2025 edition of the Global Agent Hackathon !
Join for a month-long, open-source AI Agent Hackathon, open to all builders and dreamers working on agents, RAG, tool use, and multi-agent systems.
https://lnkd.in/eWEDcQgZ
👉 DeveloperWeek [AI + ML] Hackathon 2025 (May 12-29, 2025 - Hybrid)
Join the nation's largest challenge-driven in-person and online hackathon AI DevSummit 2025 - AI/ML innovation with over 300 participants !
https://lnkd.in/erhi9E6v
Will you participate to one ? See you then :)
Angéline
Ps : I’m testing the right day to send you this, don’t hesitate to tell me what’s best for you in the Substack chat of Madship :)