Devin: The First AI Software Engineer
Cognition Labs has introduced Devin, a groundbreaking new AI as said to be the first autonomous software engineer!
Devin can code just like a human programmer.
It completed real jobs on a freelance site and solved complex coding problems that stumped previous AI models. When tested on an open-source coding benchmark, Devin fixed over 13% of issues all on its own, way more than systems before it.
Here’s why this is a big deal.
Programming is a key job that takes humans a long time to learn, refine their craft and over-time master. Take it from someone who’s spent over 20 years coding.
AI has already been able to provide code output but Devin proves that AI may one day work alongside people as coworkers instead of just assisting. Think of all the extra projects teams could take on if they had helpful AIs to share the workload, handle small bug fixes, or even start bigger projects while they hire for larger ones.
The future looks bright for Devin too – its creators said this is just the beginning of what it can do with code.
Who knows, maybe one day we’ll all have our own personal Devin bots to help with coding our apps and websites. I for one, am on the waitlist.
Around the web
📸 Bring your panoramas to life with this Vision Pro Use Case
😎 Fun experiment: Add yourself to famous movie scenes
🪄 Gemini gets personal with it’s writing
🔍 Custom GPT to help evaluate your search rankings
🔈 This AI Generated Reggae hook is actually good
xAI/Elon Open Source Grok
The debate with OpenAI + Elon continues to heat up, and the latest update is a big one.
xAI & Elon have opened up their flagship LLM, Grok, as open source and engineers are having a blast diving in and tearing it apart to understand what makes it tick.
While incredibly complex and technical, Andrew Gao, a researcher at LangChain AI, shared his findings from analyzing the architecture and here’s a summary for you (hang with me ‘till the end, it’s worth it).
Grok has over 300 billion parameters, which is massive for an AI model, and it works using a technique called “mixture of experts” where different parts of the model specialize in different tasks. It’s designed to be helpful, harmless, and honest.
Grok uses a method of positional encodings called “rotary embeddings” which allows the model to learn. It has a huge vocabulary of over 130,000 words to understand language nuances, and the model structure is optimized for both performance and privacy with techniques like quantization.
Ok, enough tech talk – but here’s why this is important.
Just like we appreciate knowing the ingredients of our food, and where our clothing is made, we’re moving towards a world where it’s important to know how our data is being provided.
OpenAI is closed.
Google will never release it’s search algorithm.
Instagram is a black box.
Yet, Twitter, and now Grok (the LLM available inside of a paid Twitter account) are now open for all.
Say what you will about Elon, but he’s challenging Status Quo in every way you can think of and I’m here for it.
