- Product Hunt Dev
- Posts
- May 15, 2024 - Fake data made easy
May 15, 2024 - Fake data made easy
Today’s newsletter is brought to you by WorkOS
Hey there! Today’s newsletter was easy to write because it’s been a busy couple of weeks. Below, you’ll find all the details on the buzzy launches from OpenAI and Google, a tool for mock data, and of course, the weird stuff.
Let’s go.
CAT NIPS
Notable launches, short headlines
Developers can now access OpenAI’s new GPT-4o model in the API as a text and vision model.
Copilot Chat in GitHub's mobile app is now generally available.
Amazon Q, Amazon’s AI assistant for devs, is now generally available.
Google I/O’s developer updates included Gemini 1.5 Flash, Project IDX in public beta, Fluter 3.22, and more. Our favorite summaries here and here.
LEADERBOARD TOPPERS
Realistic mock data for your relational database
AI has delivered a new way to help developers waste less time on stuff that isn’t coding. This time, it’s seed data.
Snaplet generates realistic seed data using AI to understand your database schema. “Basically, imagine if your seed script could write itself using your schema,” explains the Snaplet team. Default data, like currency codes, for example, is added automatically when it’s needed, and if you have specific data to define, you can do so using Typescript. Snaplet Seed creates a TypeScript client based off your database structure, giving you the power of Typescript and node.js infrastructure. Snaplet supports PostgreSQL, SQLite, and MySQL, with more data sources coming soon.
Let’s look at an example from the team. Say you’re a table-booking startup, and you need data. You’ll need restaurant data, table data, and customer data. Instead of spending a ton of time making up fake data, you’d tell Snaplet to generate ten tables. You’ll get back tables, plus the link data. Snaplet also exposes your database schema in a web UI where you can use plain language prompts to make the data even better. So, for example, all it takes is a few words, and you can make that restaurant data TexMex-themed.
Snaplet was founded by Peter Pistorius, who was also on the cofounding team of RedwoodJS with former GitHub founder Tom Preston-Werner. It was the second most upvoted dev tool of the week last week, coming in after Eraser AI.
At a glance:
• Topic: Product of the Week, relational database, data, Typescript
• Funding: Pre-seed
• HQ: Berlin
• Founder: Peter Pistorius
SPONSORED BY
You want those enterprise customers, but are you ready for them?
The building blocks to become Enterprise Ready are already here, through WorkOS. It's a modern identity platform built for B2B SaaS companies, with flexible and easy-to-use APIs to meet the needs of enterprise customers faster.
WorkOS brings a modular approach to B2B auth, with enterprise features like SSO, SCIM, and User Management. The APIs are designed to provide an effortless experience from your first user through to your largest customer.
It's free for up to 1 million MAUs and hundreds of high-growth scale-ups like Vercel, Webflow, and Loom are already using WorkOS.
BUILT TO LEARN
The case where Llama 3 beats GPT-4
Nearly one month ago, Meta released Lllama 3, claiming it is “The most capable openly available LLM to date.” So how did the model actually stack up?
Kyle Corbitt, the founder of OpenPipe, shared his experience with Llama 3 on the Product Hunt blog — the good, the bad, and the ugly (plus suggestions for Serving Lllama 3). Corbitt was previously a Director at YC working on Startup School and has worked as an engineer at YC, Google, and Facebook. OpenPipe is a platform that helps teams train and deploy their own, specialized LLM models that are faster, cheaper, and more accurate.
In working with the 8B and 70B parameter models, Corbitt found that the fine-tuned Llama 3 8B performs similarly to fine-tuned Mixtral 8x7B, and his previous work had demonstrated that Mixtral can outperform a fine-tuned GPT-4.
The bottom line? “You can serve a fine-tuned model that outperforms GPT-4 for 1/50th the cost.”
WEIRD. WHY? AND WOAH!
Talking points from Planet Internet
🚀🐭 Tesla isn’t the only one with reverse propulsion landing abilities. In fact, Disney has begun using the same theory for some new flying entertainment robots. The difference? They use water sprayed at a high enough pressure to slow down the bots on their descent.
💎🙌 Remember the Game Stop stock squeeze of 2021? Championed by retail investors, it led to a frenzy on Wall Street and even sunk one hedge fund. Well, the trader who started it all is back. Keith Gill, or Roaring Kitty, recently started tweeting cryptic trading memes which in turn sent the $GME price from $10 to $80 over the weekend.
AI CORNER
Highlights from Deeper Learning
“Her” is here. GPT-4o is similar to GPT-4 in intelligence but is much faster and improves capabilities across text, vision, and audio.
Veo is Google’s answer to OpenAI’s Sora – it’s a new generative AI model for video that outputs 1080p video more than a minute long.
Gemini 1.5 is faster and its context window is nearly double its predecessor. It will power big updates to Google Search.
Google showed a demo of Project Astra, a multi-modal AI assistant.
|
MORE TOOLS
DEV TOOLS
Jovu is an assistant that can complete backend services, including data models, APIs, and infrastructure code.
Middleware is an open-source product that offers DORA metrics out of the box.
Baseline JS is an open-source, serverless, full stack framework.
Wegic is an AI web designer and developer assistant powered by GPT-4o.
ThriveStack provides self-service infrastructure for PLG features.
Google’s competitor to DALL•E got an update — to Imagen 3; announced along with Veo and Gemini 1.5 Flash.
MAKER TOOLS
Replica lets game devs, etc. use ethical AI to create unique voices in its Voice Lab.
DataLab lets you talk with your data and features a powerful IDE to tweak and run your analysis.
Productboard is a product management system with AI automated feedback categorization to link insights to features.
Until next week
Written by Sarah Wright.
P.S. Someone awesome forward you this newsletter? Sign up here.
What did you think of today's newsletter? |