BB Digest: Gemini 2.0 thinks for free

Digest #530 → Build with 20k+ others - Become a Pro member

Hey folks,

We’re gearing up for Build Month - a month (ish) focused on helping people build apps with ai without writing code—we’ve got workshops, sessions & resources coming out soon.

I’m hosting two sessions;

• Cursor co-working tomorrow (we’ll chat all things cursor and help each other out) and;

• Office hours on wednesday (come chat about your projects, get help).

• Plus, Amie is hosting a Bolt office hours today to show how to prompt→app effectively.

To register, they’re linked in Slack here. (you need to be a pro member to access)

TLDR; inside today’s newsletter

• Gemini 2.0's new models and Replit’s amazing AI agent

• new blog on finding the sweet spot for AI projects

• 3 expert workshops coming up

• conversations from our community on building apps

• 6 AI tools including a new AI ad maker

• Why Replit's mobile-first agent caught my eye

• 6 posts including OpenAI’s secret move in Japan

• to-dos

Alright, let’s get to it!

 🔎 Trends & news

• Google released a batch of new Gemini models - The Gemini app got access to Google’s “thinking” model with a special variant. Gemini 2.0 Flash Thinking with Apps (like SEARCH, Youtube, Maps etc.) is better than DeepSeek and available for free.

  • For developers, Gemini 2.0 Flash is now generally available with simplified pricing ($0.10/$0.40 for 1M input/output tokens). A new model, Gemini 2.0 Flash Lite is in preview with even (slightly) lower pricing. If you consider intelligence per $, Gemini Flash models are the best in the world, with no one near them.

  • Gemini 2.0 Pro (Experimental) battles hard with 3.5 Sonnet on benchmarks and is arguably the strongest non-reasoning model out there. Gemini Advanced users also get access to this model in the app.

• Replit’s AI Agent is now free to try. You can also use it right from Replit’s mobile app. Replit has all the features to get started with app creation, develop it seriously and put it live for others to see/use—all of it without writing any code yourself. (more below)

• OpenAI in the meantime: rolled out Deep Research in the EU, UK and more (yayy, costing me $200), made ChatGPT Search available for users without any signup and schemed with California State University to give 500,000+ students and staff access to ChatGPT Edu.

• Learn how to create powerful, reliable AI agents with Galileo's in-depth eBook! Enjoy 100 pages of expert content on agentic frameworks, proven techniques for evaluating and improving agent performance, and tips for identifying failure points and production issues. Get your copy now*
  *sponsored

👀 ICYMI: New stuff we’ve launched

• You can unlock ROI by focussing on mid-scale projects for getting AI to help you. These aren’t your typical moonshot projects or just low-level AI assistance and hence they don’t get much attention. we put together 4 examples of mastering this middle ground of AI with resources to get you started too.

Upcoming Workshops

• 11th Feb - Create full-stack apps and deploy them with Bolt. (sign up)

• 14th Feb - Get working code from your Figma designs with Builder. (sign up)

• 19th Feb - Quickly bring your AI app ideas to life with Create. (sign up)

💬 Inside the community this week

Become pro to join us building and learning together.

• Wyatt shared results from ChatGPT Deep Research and what it means for public policy analysts in DC. (link)

• Members are building apps with Replit and Cursor. Justine made one on her iPhone, while I'm figuring out localhost stuff. (link)

• Alexandra's building a Lovable app to estimate presbyopia. 🤯 How cool is that (link)

Join the conversation, plus full access to courses and workshops by becoming a pro member today!

⚙️ Top new tools

• Jam AI – your AI that writes bug reports for you.*

• Icon - Think ChatGPT + CapCut, but for making winning ads with AI in minutes.

• Tana - Put your notes to work with voice and AI.

• ChatBase - The easiest way to make AI agents that work with your documents.

• Riffo - Let AI name your files to fit the contents.

• CubeOne - Leave slides and scripts to AI as you get ready for a presentation.

More tools here →
*sponsored (book an ad here)

👀 What caught my eye… Replit’s AI agent

Replit made their AI agent free to try and it’s taken over my X feed. The strange part is that most of these examples are people creating apps when travelling or even sat on the loo, while on their phones—which is crazy.

Here’s an example from one of our members, Justine: She built an app to track if her dog Tilly has been fed or not? With the features it has, this app would have raised a few million dollars in 2021. I bet someone did 👀

Now, yeah, those businesses are not sustainable. But we all often need “personal software” like this. With Replit, (and I assume an increasing number of tools) this category of software will see a massive uptick.

 📜 Interesting posts

• How I use AI in early 2025 - A good post without being overwhelming.

• Generating image descriptions and alt texts for 10,000 images with AI. (sidenote: something like this will cost <$0.25 in AI costs with Gemini 2.0 Flash Lite)

• Tyler Cowen is using Deep Research to write small papers.

• Why AgentOps is picking Clay for CRM automation along with Hubspot.

• SoftBank makes a joint venture with OpenAI in Japan. It’ll spend $3B/yr on OpenAI’s tech.

• The end of programming as we know it by O’Reilly.

📌 To-dos

That’s it for today. Feel free to hit reply and share your thoughts. 👋