Memory Leak — #10
VC Astasia Myers’ perspectives on machine learning, cloud infrastructure, developer tools, open source, and security. Sign up here.
ChatGPT is a foundational modal that interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests. Similar to GitHub Copilot, ChatGPT is trained on OpenAI’s Codex that is made up of billions of lines of open-source code.
Why does this matter? ChatGPT has taken the internet by storm. It achieved 1 million users in its first five days. One of the reasons DALLE was so successful was because it provided a visual output that could be shared on social media; in turn, making the magic of generative AI more relatable to the average person. OpenAI learned from this experience and made a front-end interface for ChatGPT that wasn’t present for GPT-3. A key takeaway for founders is to make product launches as visual and engaging as possible to prospective users. Is there a way to present your product in a launch that motivates and enables others to share it?
Fig’s products focus on improving the terminal experience. The company started with Autocomplete and recently launched Scripts, a replacement for a team’s “scripts” folder and internal CLI tools. Fig Scripts provides a searchable collection of a team’s terminal workflows; interactive terminal UIs; automatic provisioning; and out-of-the-box error reporting and usage monitoring.
Why does this matter? About 60 million developers use the terminal, but it hasn’t advanced to the same degree as the IDE. Scripts broadens Fig’s offering beyond a single player use case to team-based value. We’ve seen a few startups including Fig, Textualize.io, Charm.sh, and Warp focus on modernizing the terminal experience.
Cohere Sandbox allows developers to explore, test, and experiment with LLMs via open source. Sandbox is a collection of experimental, open-source GitHub repositories that can be used to build applications with LLMs and to showcase how such applications are built. The initial release includes five services: Conversant, Route Generation, Grounded QA, Topically, and Toy Semantic Search.
Why does this matter? We previously discussed closed source versus open source foundational models and found that most individuals intend to use open source. Cohere, which is best known for it’s closed source API products, is trying to attract open source users with Sandbox. Is this a signal that the open source distribution approach will be the winner?
Definitive Guide to API DevEx Portals
Sagar Batchu highlights an API DevEx Portal is critical for improving API developer experience. He discusses that the key attributes and functionality to build an API DexEx Portal:
- Accessible: API Documentation, Auth Login, and API Key Management
- Understandable: API Request Viewer, API Usage Metrics, and API Status Page
- Usable: Client SDKS, API Runbooks, and API Sandbox
Why does this matter? SlashData found that nearly 90% of developers use APIs. While APIs are ever present, often the tooling around APIs is lackluster leading to a material impact on adoption and usage of an API. An API DevEx Portal can help teams improve onboarding and user experience.
What’s Next for Data Engineering in 2023? 7 Predictions
Barr Moses states her predictions for data engineering including more time spent on FinOps and data cloud cost optimization as well as data warehouse and data lakes use cases blurring.
Why does this matter? Barr is a well-known thought leader in the data space with strong visibility into what’s coming next. We particularly agree with her sentiment around cost. We constantly hear from operators that their cloud data warehouse bills quickly increase. In this macro environment, we can imagine this is a top priority for teams.
Deepgram Completes $72M Series B Round to Define the Future of Speech Understanding
This piece is more than just a funding announcement. It actually walks through the history of Deepgram and technical story behind the business. Speech to text and Natural Language Understanding (NLU) has evolved a lot since Deepgram started in 2015.
Why does this matter? Deepgram has transcribed over 1 trillion words. Speech to text is fundamental for training data generation for use cases including contact centers, speech analytics, conversational AI, and podcast transcription.
⭐️Dragonfly — Developer Advocate (fully remote)
⭐️ Hex — Data Analyst
⭐️ Grit — Founding ML Engineer