Hello, welcome to today's blog
I've been going through some internet reading while watching a long YouTube video on why people can't read a book anymore in 2025. The video was by Mark Manson, the author of the most famous non-fiction book a young man would pick.
Never mind, a few things that I read immediately blew my mind as soon as I opened my editor and started writing this blog.
Firefox Web AI API
This might sound absurd, but this AI will shut down a couple of AI tools and LLM agents as well. Browsers themselves have AI LLM running under the hood to do some AI-related tasks such as getting information from webpages, summarising, rewriting, translating and doing more.
I've this though that we even need n18 type nextjs internalisation tools when we can easily use LLM to translate, summarise and do more.
https://firefox-source-docs.mozilla.org/toolkit/components/ml/extensions.html?ref=blog.mozilla.ai
A nice blog on how to use Tanstack query or react-query to optimise performance, avoid refetching and prevent deduplication, meaning how to use react-query in a better way.
RAG is the old concept in the world of AI development, CAG is the new one, CAG stands for Context Augmented Generation, by it's name it will take the context first before generating text whereas RAG uses LLM model to run everytime for each new query/prompt and needs specific intructions, data ingestion and model traning to work for a specific niche kind of problem and that brings CAG into the picture. CAG takes all the context or so-called training data into it once and then instantly answers based on the context data, quite interesting, read more
Have you thought of having your own Firebase or Supabase alternative or custom-made solution in the backend? Then Pennybase is the github repository to test, check and build one. The file-based backend-as-a-service for pennies backend developer that stores data in CSV files and provides tonnes of APIs for authentication, CRUD, RBAC (Role Based Access Control) and more. PennyBase
Wasm agents for building AI agents for browser automations in the browser using WASM and pyodide, quite an interesting one to check, by still waiting for a JS library to build for me. But we have one for JS lovers by OpenAI called openai-agents-js
Gettemplate.website has a few more new templates that you can use to quickly build your next idea, Blog creation form and Tiptap advanced editor. Do check more because we are consistently adding templates, features like change theme, save the template, etc
We have brought another template called AI React code generator, which takes a prompt and provides simple React, Tailwind, Lucide-icons/react-icons, framer-motion, and gsap as the libraries to provide a single component source code. TRY React AI Code generator
For developers starting their journey in Frontend, Backend or Mobile app development, we provide a roadmap on the website. Check it out.
Do you know the above images are the recently funded companies that are hiring remotely or might hire in the next few months, so if you are looking for a job, email them by finding contacts online.
Quite interesting tweet about new storage or vector DB for semantic search for LLM models storing the data into mp4, who thought about it way back in 2000
That's it for today, see you in the next one
iHateReading
Summary
This blog post covers several recent developments in AI and web development. Key highlights include: Firefox's new Web AI API, enabling AI tasks directly within the browser; a discussion of Context Augmented Generation (CAG) as an alternative to Retrieval Augmented Generation (RAG); Pennybase, a lightweight, file-based alternative to Firebase or Supabase; the use of WASM agents for building AI agents within browsers; new website templates available at Gettemplate.website (including an AI-powered React code generator); and a showcase of recently funded companies hiring remotely. The post also features a Twitter embed discussing novel approaches to vector databases for semantic search using MP4 files.