Ayush Kumar is an experienced Data Scientist. In his seven years in the industry, he’s developed expertise in many data roles; most recently, he’s been working with generative AI, “because there has been a lot of hype and also it is quite promising technology,” he says.
While Ayush is usually reluctant to add new tools to his workflow, in the weeks since he added Pieces to his stack, he’s noticed big changes. “I don't use a lot of new tools generally, but I have been quite hooked to Pieces.” Even though Pieces has a bunch of great features to choose from, Ayush identified a few favorites as explained below.
- Centralized Code Repository: Offers a unified place for storing and accessing previous code, enhancing organization and retrieval efficiency.
- Time-Efficient Code Reuse: Simplifies the process of reusing code, replacing the manual and unorganized approach with a streamlined, search-based system.
- Innovative Web Extension: The Pieces Web Extension integrates with sites like Stack Overflow and Medium, allowing easy saving and reusing of code snippets directly from the web.
- Automated Metadata Generation: Automatically stores useful metadata about saved snippets, including related links, for easy reference.
- Local LLMs for Enhanced Security: Addresses security concerns with local language models (LLMs), ensuring code stays private and secure.
- Integrated Query Support: The ‘Ask Copilot’ feature in JupyterLab notebooks provides immediate assistance for coding queries, saving time and effort.
- Effective for All Developers: Tailored for programming needs, Pieces Copilot delivers more relevant and efficient results.
Ayush’s Tech Stack:
- IDEs: JupyterLab
- Other Tools: Pieces, Chrome, GitHub, AWS, Anaconda
Searching Saves Time
Auysh’s former code-reuse strategy wasn’t exactly efficient. “I have been mostly referring to my older use cases by opening the full code and copying & pasting a certain section, but it's all manual,” he explains. “I would have to first think about where I have done some similar work, then open that Python file, and copy and paste it into my new work environment. I would say there was no organized approach in terms of reusing; it was just all over the place.”
Once he installed Pieces, that disjointed approach to finding and reusing code was replaced by a far more useful process. In fact, it’s one of Ayush’s favorite Pieces features. “It’s quite amazing to have a central repository and a search functionality to quickly get to my previous code,” says Ayush. “I don't think I had that ability before in any fashion. I would say that Pieces has been a huge efficiency game-changer for me.”
The Revolutionary Pieces Web Extension
Practically every developer runs into bugs and searches for solutions on the web. Typically, a developer would copy that code, edit it in their IDE, and then head back to the web in a few weeks to solve the same problem. Once Ayush installed the Pieces Web Extension, that cycle became far less repetitive.
“Pieces also has inbuilt functionality on third-party websites like Stack Overflow and Medium. Wherever we have a good code snippet in the article, the Pieces Web Extension gives the copy and save button,” Ayush explains. “Once I save it, I go back to JupyterLab and then I can just start to use the code. The best part is whatever I saved recently pops up first, so I don't need to search for it. If I have a snippet that is already saved from the past, I search for it right there and then I can build on it. As I have any questions, I don't need to search Google— I can just use the Pieces Copilot.”
Beyond automatically adding save buttons below online code blocks, Ayush appreciates that Pieces generates and stores metadata about those snippets, from titles to related links. “Pieces gives me related links, which are quite powerful,” Ayush remarks. “I don't need to go to Google and find a link again; a snippet already has the most appropriate links because Pieces is meant for programmers.”
Ayush + Pieces Copilot = 🖤
As a data scientist, Ayush finds Pieces Copilot practically and academically fascinating. “There have been a lot of security-related concerns with LLMs where organizations aren't sure whether their data is secure or not. That's why there has been a lot of interest in leveraging local models,” notes Ayush. “Everyone is trying to also use local models, because everything is proprietary and code is not leaving your device.” The convenience and security of using Pieces’ local LLMs is just one of the reasons Ayush loves Pieces Copilot.
“The ability to be in a JupyterLab notebook and just hit the ‘Ask Copilot’ button has been a game changer for me because it saves me a lot of time,” Ayush says. “I might hit on a bug, and I can just ask right then and there on the same screen, ‘What is the issue with this code?’”
Ayush thinks the benefits of Pieces Copilot extend to all developers. “Pieces Copilot has become much more efficient for any developer to ask any question and get a particular result,” says Ayush. “The LLMs in Pieces are sensitive to programming, so I think that gives better results.”
Ayush’s final takeaway? “I would definitely recommend using Pieces because it's a great tool to have in your toolset,” he declares. “It allows you to be a lot more efficient, as you can start to reuse your code again in different projects. I'm going to use Pieces quite often.”