Hey HN! Wanted to share something cool about an DIY alternative to OpenAI's, Perplexity's and other Deep Research agents
So everyone at my company was hyped about OpenAI's Deep Research when it came out and finally got to try it. They got this massive 26-page report and when I looked at it, I had this "aha!" moment - it's basically just stitching together different search queries and having their model rewrite it. I thought "hey, we could totally build this ourselves!"
I actually made AI Workflow Automation plugin (built it completely with Claude as my programming buddy!) and used it to create this setup in WordPress with Grok 2 and Perplexity's APIs.
Here's how it works: You throw in a research query, it hits multiple research nodes running Sonar Pro (each looking at the topic from different angles), feeds that to AI nodes (I used Grok 2 for its long context window and solid writing), and boom - you get a nicely written well-structured and well-cited research report.
The results were pretty wild:
My setup: Cranks out 5000-6000 words in about 6 minutes
OpenAI's thing: Takes around 10 minutes
Cost? Less than 30 cents to run the whole thing!
I actually did a deep dive comparing OpenAI's Deep Research, Perplexity, and my custom setup. Quality-wise? They're all neck and neck. The cool thing about my approach is you can tweak the output length however you want, or add additional verification or validation steps in between.
What makes this really work is Perplexity's Sonar Pro and Sonar Reasoning - especially now that you can control the time window and search context. It's pretty sweet. If you don't use WordPress or AI Workflow Automation, you could probably cook up something similar in N8N or other workflow software.
If you're using the plugin and want the workflow, just ping me - happy to share! I also wrote a small blog about it that I have attached here.
Hey HN! Wanted to share something cool about an DIY alternative to OpenAI's, Perplexity's and other Deep Research agents
So everyone at my company was hyped about OpenAI's Deep Research when it came out and finally got to try it. They got this massive 26-page report and when I looked at it, I had this "aha!" moment - it's basically just stitching together different search queries and having their model rewrite it. I thought "hey, we could totally build this ourselves!"
I actually made AI Workflow Automation plugin (built it completely with Claude as my programming buddy!) and used it to create this setup in WordPress with Grok 2 and Perplexity's APIs.
Here's how it works: You throw in a research query, it hits multiple research nodes running Sonar Pro (each looking at the topic from different angles), feeds that to AI nodes (I used Grok 2 for its long context window and solid writing), and boom - you get a nicely written well-structured and well-cited research report.
The results were pretty wild:
My setup: Cranks out 5000-6000 words in about 6 minutes OpenAI's thing: Takes around 10 minutes Cost? Less than 30 cents to run the whole thing! I actually did a deep dive comparing OpenAI's Deep Research, Perplexity, and my custom setup. Quality-wise? They're all neck and neck. The cool thing about my approach is you can tweak the output length however you want, or add additional verification or validation steps in between.
What makes this really work is Perplexity's Sonar Pro and Sonar Reasoning - especially now that you can control the time window and search context. It's pretty sweet. If you don't use WordPress or AI Workflow Automation, you could probably cook up something similar in N8N or other workflow software.
If you're using the plugin and want the workflow, just ping me - happy to share! I also wrote a small blog about it that I have attached here.
Let me know if you have any questions!