Flowith Neo: A Promising Yet Somewhat Bare-Bones AI Workflow Power Tool

Neo: The Supercharged Oracle

Hey everyone, today I want to chat about my recent experience with Flowith's new feature, Neo!

I'm sure many of you have seen the teaser by now. In a nutshell, Neo is a supercharged upgrade of the previous Oracle ✨. It integrates more tools, has its own backend sandbox, and most impressively, it can run tasks periodically! It's like it gives you a massive toolbox 🧰, helps you automatically set up workflows, and can even make those workflows run on their own schedule. It's a bit like helping you build a "Kouzi" platform (or say, Coze), and then letting it run autonomously!

Why Is Neo Worth Paying Attention To?

Why do I think this direction is great? 🤔 Before, when using Oracle, I'd typically run some simple tasks. For complex operations, I still had to download things locally, modify them using Cursor or something similar, and then run them. Now Neo is powerful. The workflows you set up can run directly in its own sandbox, offering a one-stop service that covers everything from ideation to continuous execution! We all know that platforms like "Kouzi" or Dify require manually building workflows, which is incredibly time-consuming 😫. They have many rules, and troubleshooting is a hassle. Unlike writing code where you can check logs when errors occur, with these platforms, you might not even know why something failed. Although "Kouzi" has many users now and is okay for simple setups, complex ones are truly time-consuming and labor-intensive. So, the idea of Neo automating workflow creation for you is really interesting! I've tried before to have a large language model generate a workflow and import it into Dify, but it basically never worked 😂. Those Dify files shared online often fail to import successfully too.

Some Real-World Pain Points

BUT! The reality is a bit leaner than the ideal... 😅 From my usage, I've discovered a few issues to share with everyone:

1. Success rate isn't that high: Looking at the Discord group, many folks report that generating million-word articles or running complex tasks yields inconsistent results. The more complex the task, the higher the chance of failure seems to be. It feels like the feature is still in a relatively early stage.
2. Difficulty pushing data: The teaser mentioned sending emails, but many in the group have reported it doesn't work 📧. I asked about sending via DingTalk bot, and was told that's not possible either. Currently, results can only be displayed on the webpage. The official word is that an APP will be developed later, introducing more tools and enabling push notifications through the APP.
3. Task modification is super troublesome: Once a task is generated, if you're not satisfied with a particular step and want to make a slight adjustment? It's extremely difficult, almost impossible! Flowith staff have also admitted they don't want users modifying already generated tasks. So, a lot of the time it's like "opening a blind box," relying on luck 🎰.
4. Major subscription model changes: I was previously on the monthly unlimited membership. Now there's a Professional tier for $20 per month, but it only gives 2000 credits, which are reportedly used up very quickly. Many in the group are not happy with this pricing.

Hands-on Testing: From Expectation to Reality

Sharing a few of my own small tests 🧪:

• 24/7 Continuous Task Test: Search Twitter for AI startup-related content and display it on a frontend webpage.
  • Pros: Executed the search, called tools, generated the frontend webpage.
  • Cons: The scheduled task didn't actually produce output.
• Data Push Test: Attempted to send data via DingTalk bot or email; both failed.
• Pleasant Surprise Test: Asked for help building a frontend/backend product.
  • Generated a detailed search strategy document.
  • Provided detailed suggestions for frontend/backend integration, deployment, etc.
  • Code generation is now much more detailed than before.

Final Thoughts

Overall, Flowith's changes this time are really significant, and you can see they have grand plans. It's normal to encounter issues and receive criticism right after launch, but I think the overall vision is quite good and worth looking forward to!

Flowith Neo's progress is clearly visible, especially in rapid web frontend/backend development—it's much more impressive than before! It's just not as perfect as the promotional video yet, so it's good to manage expectations a bit when using it. I believe they will continue to optimize, and I'm very much looking forward to Flowith's future developments!

That's all for today's share. I hope it helps those of you interested in Flowith! 💕

#Flowith #AITools #NeoReview #WorkflowAutomation #AIAgent #ProductivityTools #AIMoneyMaking #TechEarlyAccess #MyAIExperience