- unwind ai
- Posts
- 10x Faster Web Scraping for AI Agents
10x Faster Web Scraping for AI Agents
PLUS: ChatGPT Go for $5/month in India, Google's image editing model
Shubham Saboo & Gargi Gupta
August 20, 2025
In partnership with
Today’s top AI Highlights:
Opensource multi-agent memory engine for LLMs and AI agents
Firecrawl v2 API just made web scraping 10x faster
OpenAI launches ₹399 ($5) per month ChatGPT Go in India
Google’s anonymous image editing model nano-banana
Opensource serverless vector database for AI agents and RAG
& so much more!
Read time: 3 mins
AI Tutorial
Building targeted B2B outreach campaigns is one of the most time-consuming aspects of sales and marketing. The challenge isn't just finding companies; it's discovering the right decision-makers, researching genuine insights, and crafting personalized messages that actually get responses.
In this tutorial, we'll build a multi-agent AI email outreach system using OpenAI GPT-5, Agno for orchestrating agents, and Exa AI for intelligent web search. This system automates the entire outreach pipeline - from company discovery to personalized email generation - delivering professional, research-backed outreach emails in minutes instead of hours.
Our multi-agent system conducts real research on each company using website content and Reddit discussions and ensures every email feels genuinely personalized.
We share hands-on tutorials like this every week, designed to help you stay ahead in the world of AI. If you're serious about leveling up your AI skills and staying ahead of the curve, subscribe now and be the first to access our latest tutorials.
Latest Developments
AI agents now remember like humans do, with working memory for immediate context and deep storage for long-term insights.
Memori is an opensource stateful memory engine that gives AI agents the ability to recall past work, maintain preferences, and coordinate multi-step processes without losing context.
Memori uses a multi-agent architecture with three distinct modes: Conscious Mode for essential short-term context, Auto Mode for dynamic long-term retrieval, and Combined Mode that blends both approaches.
The framework is SQL-first, supporting SQLite, PostgreSQL, and MySQL with built-in versioning and optimization. What makes this particularly compelling is how the same memory can be shared across multiple agents, creating workflows where one agent's learnings directly benefit another's performance.
Key Highlights:
What It Processes - Three specialized AI agents work together to capture conversations, extract people and projects, analyze roles, patterns, facts, and preferences, and intelligently decide which memories to keep active for future interactions.
Dual-Mode - Conscious Mode promotes essential conversations to immediate access like human working memory, while Auto Mode dynamically searches the entire memory database based on current context.
Cross-Agent Sharing - The same persistent memory can be accessed by multiple agents, enabling workflows where agents build on each other's knowledge and maintain consistency across different tasks.
Auto-Records All Chats - Works with any LLM library and records conversations with a single
memori.enable()call that automatically captures and contextualizes all interactions without code changes.
Firecrawl just hit two major milestones: raising $14.5M in funding and shipping their most ambitious product update yet.
The team has shipped /v2 - a complete overhaul of their endpoint that delivers 10x faster scraping through intelligent caching, introduces semantic crawling that understands natural language prompts, and adds news and image search capabilities that integrate seamlessly with AI applications. The release also includes a completely redesigned playground and dashboard, plus developer experience improvements that eliminate common pain points like manual timeouts and complex API parameters.
When foundational tools that power 1000s of AI agents get this level of investment and engineering focus, it makes the entire ecosystem more powerful and accessible.
Key Highlights:
Lightning-fast performance - Intelligent caching keeps frequently accessed pages fresh in the system, delivering most scrape results in about 1 second instead of the previous longer wait times.
Semantic crawling with prompts - Simply add natural language instructions like "get the blog pages" to any crawl request, and the system automatically configures the right parameters to find exactly what you need.
Expanded search capabilities - New support for news and images alongside web results, with direct integration into their Fireplexity demo showing how AI applications can seamlessly pull relevant visuals.
Developer-first improvements - Streamlined API parameters, better SDK types, enhanced Python support, and clear error messages that actually help you fix issues instead of leaving you guessing.
Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.
Quick Bites
AWS Labs just shared their Vibe Coding guide, and it's quite thoughtful, covering how to craft better prompts, when to use different AI clients for different tasks, and how to keep yourself in the driver's seat while letting AI handle the heavy lifting. The core philosophy is that you own the architecture and vision, and AI accelerates the implementation. Check out the Hacker News thread too - folks are sharing some genuinely helpful tricks of their own that go well beyond what AWS covered.
OpenAI has launched a new subscription tier, ChatGPT Go, in India for just ₹399 ($4.75) per month. This tier offers 10x higher message limits, image generations, file uploads, 2x longer memory compared to the ChatGPT free tier, and access to Projects and custom GPTs. We might see more geographic pricing tiers not just from OpenAI but also other AI companies as they compete for emerging markets.
A mysterious AI image model has been dominating LMArena, and nobody knows who built it. nano-banana came out of nowhere with image editing abilities better than other models like OpenAI’s gpt-image model. While everyone’s just speculating, all signs point to Google being behind this stealth drop. The model can do complex edits, keep characters looking consistent, and generate images crazy fast. It’s available on LMArena’s Battle mode only, popping up a few times randomly.
Here are some demos that might force you to try it out:
That’s not all in AI image editing, however. Alibaba Qwen just dropped Qwen-image-edit model, an opensource 20B model that can edit images from natural language prompts. The model handles both semantic editing (style transfers, object rotations, IP creation) and appearance editing (precise element addition/removal with unchanged backgrounds). We’ve been playing with it too, and it’s really good! You can try it out on Qwen Chat for free. It’s also available on Hugging Face, Modelscope, and via API.
India’s fastest-growing YouTuber, Dhruv Rathee, launched his startup AI Fiesta, a platform that offers all the top models, including GPT-5, Gemini 2.5 Pro, Claude Sonnet 4, and more under one roof, for a cheap ₹999 ($11.5) per month subscription. As interesting as it may sound, the catch is not hard to miss - just 400K tokens per month. That limit is way too small for serious users who are already paying for Pro or Plus plans across OpenAI, Anthropic, or Google, and ironically, the people who don’t need those subscriptions can already rely on free tiers of GPT-5, Gemini 2.5 Flash, or Claude with very generous usage.
Anyway, because the internet never waits, someone just built a Chrome extension that removes chat limits on AI Fiesta by blocking message count tracking requests. Guess the only fiesta here is in the GitHub repo.
Tools of the Trade
Chroma Cloud - Opensource serverless, cost-effective, Cloud-based, fully-managed version of Chroma vector database with the same API as local Chroma. It lets you create a database in seconds, scale without configs, and pay only for usage.
Graphite Chat - Integrates AI chat directly into GitHub pull requests, letting you highlight code sections and ask questions about functionality, request refactors, or get fixes that can be applied instantly within the same interface. Serves both PR authors and reviewers.
Claudia - A desktop companion for Claude Code with a beautiful GUI. It wraps Claude Code command-line functionality in a visual interface with features like searching through your Claude Code sessions, creating custom AI agents, running background tasks, managing MCP servers, monitoring token usage and costs, and more.
Awesome LLM Apps: Build awesome LLM apps with RAG, AI agents, MCP, and more to interact with data sources like GitHub, Gmail, PDFs, and YouTube videos, and automate complex work.
Hot Takes
My house has 27 GPUs.
Before AGI arrives:
Acquire GPUs.
Go into debt if you must.
But whatever you do, secure the GPUs. ~
Ahmad OsmanFunny how people think AI will replace software engineers, while Amazon still employs over 100,000 customer support reps. ~
Yuchen Jin
That’s all for today! See you tomorrow with more such AI-filled content.
Don’t forget to share this newsletter on your social channels and tag Unwind AI to support us!
PS: We curate this AI newsletter every day for FREE, your support is what keeps us going. If you find value in what you read, share it with at least one, two (or 20) of your friends 😉
Reply