Let’s connect! 🤝
Looking to meet more people in:
💻 Software development
🤖 AI & automation
🚀 Startups & SaaS
🛠️ Indie building
If you’re building something interesting, I’d love to hear about it.
Tell me what you’re working on below 👇
@KDope958 Always plan for what happens when something fails. A workflow might retry after a timeout and end up creating duplicate records or sending the same notification twice.
Getting automation to work is one thing. Making it reliable is another.
@AdamHoltererer I use both quite heavily, so I’m curious how this plays out. If using ChatGPT for complex tasks means less Codex time, that’s going to be frustrating. I’d honestly prefer separate limits.
@eandualem That’s a smart approach, Elias. Having one main agent per repo makes sense.
I’m curious though, when parallel workers make overlapping changes, how do you handle conflicts? Does the main agent review and merge everything?
Let’s connect! 🤝
Looking to meet more people in:
💻 Software development
🤖 AI & automation
🚀 Startups & SaaS
🛠️ Indie building
If you’re building something interesting, I’d love to hear about it.
Tell me what you’re working on below 👇
@buildbyharshit Hey @buildbyharshit! I’m building software products and exploring AI agents too. Looks like we’re in the same space. Let’s connect bro 🤝
@eandualem This is interesting, Elias! I’m working on something in the AI agent space too. Curious how you’re handling coordination between multiple agents. Definitely good to connect bro 🤝
Quick question for devs:
If AI can write code in seconds, why does shipping reliable software still take so much time?
What’s actually slowing us down?
@omarvvvr Python isn’t really doing the heavy lifting. Libraries like PyTorch use optimized C++ and CUDA code to run the actual computations on GPUs.
Python mostly controls the process, while the GPU does the hard work.
Wait, when starting a new SaaS, which approach do you take?
A. Build the MVP first, then find users.
B. Understand potential users’ problems, then build.
I’ll be honest, I’ve mostly been doing A.
But building something nobody needs is a real risk.
What’s your pick?
@theo That 51% false-claim rate is wild. There’s a big difference between generating impressive code from scratch and actually understanding an existing codebase.
If the model keeps inventing things that aren’t there, you’re spending more time correcting it than building.
@Techorbitt I like the ambition, but there’s nothing wrong with being a software engineer either. You don’t have to own the next big tech company to make an impact. What matters is building something that genuinely makes people’s lives better.
@AdamHoltererer Interesting approach, Adam. I like the idea of letting people create their own rankings instead of relying on one general benchmark. Curious to see how different the results get when people set their own priorities.
@jurlycat I get preventing direct copies of existing games. But there’s a big difference between copying a game and building something inspired by it.
Hopefully Claude can tell the difference without limiting creativity.