Accuracy on its own is a marketing number.
Accuracy against ground truth your customer signed off on is engineering.
Surprising how few people in AI still think about it this way.
AI is accelerating app development and the LLMs are becoming smarter.
But, AI-programming is not without limitations:
1/ Customization and granular control
2/ Hallucinations in AI-generated outputs
3/ Inconsistent designs
4/ Lack of context awareness
5/ What changes were made and why
6/ The mysterious art of prompting
7/ Data privacy and security
Let's dive deeper into these limitations and explore potential solutions to unlock the full power of AI in app development:
๐
Pro tip: Start small. Pick ONE area where you're confident you'll see quick wins. Success there will fuel your next AI project.
Let me know if you've spotted any AI opportunities in your business using this framework ๐
5/5
Finally, evaluate the money side:
โข Can you measure time saved?
โข Will it directly impact revenue?
โข Does it create valuable data you can use later?
4/5
- It's smart. It's on par of intelligence in the benchmarks of GPT4o, while being a much smaller model.
- It has tool-use, google search baked-in into the API.
In my opinion, it's these kind of small models that will change the world and will power the agentic AI revolution.
3/3
Google has finally outdone OpenAI.
The newest Gemini 2.0 release is unreal, here's what it can do live.
I shared my website and I asked it to provide UI/UX feedback.
It worked pretty well as you can see in the video.
The key things with this model is: 1/3
- It can handle a context length of 1,000,000 tokens (GPT4o can do 128,000).
- It's completely multimodal. It can take it video, text, audio, images as input and can give you audio, text and images as output.
- It's super fast and cheap. If it follows the old flash 1.5 pricing, it's around 50x cheaper than GPT4o 2/3
What happens when there's no more friction to using AI?
That's exactly what AI voice bots deliver.
Watch how this one creates instant software cost estimates through natural conversation.
Check this out ๐ (1/6)
It's not just taking calls and answering questions - it's:
Taking action
Booking meetings
Organizing results
Integrating with external systems
It opens up a completely new paradigm. (5/6)