Jul '25 • 🛠️ dev-help
Stuck Building My First AI Agent - Need Help
Hey everyone,
I’ve been working on a simple AI workflow using n8n and DeepSeek R1 (the free model), and I'd definitely need some help.
Here’s what I’m trying to do:
- A user submits a YouTube video link
- I use Apify to transcribe it
- Then I send the transcript to the LLM
- The goal is to generate a LinkedIn-style post based on the video content
It worked once. One of the early posts actually came out great, sharp, and on point. I was happy.
Then I tried a different video… and everything fell apart.
Now I’m getting posts that are vague, generic, or completely miss the point.
It’s been over a week of trial and error. I’ve rewritten the system prompt like 30 times - changing wording, tone, structure, nothing sticks.
Sometimes I get decent output. But most of the time, it’s way off.
So I’m stuck, and I’m starting to wonder if it’s the way I’m prompting or if this model just isn’t capable for this kind of task.
If anyone here has built a content-writing agent that works reliably or in production right now, I’d love to know:
- How do you write a system prompt that holds up across different inputs?
- Is there a framework or method that helped you make it consistent?
- Is switching to GPT-4, Claude, or another paid model would solve this?
- And how do you actually test/improve these prompts without losing your mind?
Not looking for theory or general AI advice, just real-world experiences from people who’ve actually built something like this and made it work.
Would really appreciate any input. Thanks 🙏
4
9 comments
Stuck Building My First AI Agent - Need Help
skool.com/learn-ai
Learn and become genuinely proficient in AI to get paid for it. From the creators of the AI Automation Agency model.
Leaderboard (30-day)
1
+387
2
+368
3
+357
4
+271
5
+247
Powered by