🍔🧠 How Dropbox Built AI Search Using RAG and AI Agent Techniques
PLUS: 7 Best API Practices 🔥, Server Sent Events Explained ⚡, Video Game Blurs? 🎮
Happy Monday! ☀️
Welcome to the 179 new hungry minds who have joined us since last Monday!
If you aren’t subscribed yet, join smart, curious, and hungry folks by subscribing below.
📚 Software Engineering Articles
Learn 7 best practices for API design to build robust interfaces
Why you should use one big server instead of microservices
How to develop high agency and become a top performer
Master code reviews with these essential guidelines
From zero to hero in AWS cloud computing
🗞️ Tech and AI Trends
Amazon's satellite internet lands first airline customer
Atlassian acquires Browser Company for $610 million
Anthropic settles with authors for $1.5B in copyright dispute
👨🏻💻 Coding Tip
Use NGINX's proxy cache to serve stale content during failures
The next level experience to read newsletters
Meco is a distraction-free space for reading newsletters outside the inbox.
I love it because:
It lets you forget about all the emails you get and focus on what matters to you.
It’s super simple to connect to gmail/another email provider.
It has great recommendations and over 30k users.
Clean your inbox and try Meco for free here!
How Dropbox Built an AI Product Using RAG and AI Agents 🤖
Dropbox created Dash, a unified search and knowledge layer that sits on top of various business tools like emails, files, chats, and calendars. This AI-powered system helps teams cut through information overload by providing intelligent search, organization, and secure sharing capabilities across different data sources.
The challenge: Building an enterprise AI system that can handle diverse data formats, fragmented sources, and multiple modalities while maintaining security and accuracy at scale.
Implementation highlights:
Hybrid retrieval strategy: Combined lexical search with semantic reranking for fast, accurate results
Custom interpreter: Built a secure environment for AI agents to execute validated code plans
RAG pipeline: Grounded LLM responses in real documents to prevent hallucinations
Multi-phase agents: Implemented planning and execution phases for complex workflows
Strict validation: Added static analysis and runtime checks to ensure reliable operation
Results and learnings:
Fast responses: Achieved sub-2-second retrieval for 95% of queries
Flexible architecture: Created a model-agnostic pipeline supporting different LLMs
Enhanced accuracy: Successfully combined RAG for lookups with agents for complex tasks
The key takeaway is that enterprise AI requires a thoughtful balance between speed, accuracy, and security. Dropbox's approach shows how to combine RAG and AI agents effectively while maintaining strict control over execution.
7 Best Practices for API Design 🔥
Written by
What are Server-Sent Events (SSE)?
Written by
How to develop High Agency
Written by
Operating Principles That Guided Me to Staff Engineer (Part 1: Driving Impact)
Written by
Code Review Guidelines ✅
Written by
Video Game Blurs (and how the best one works)
How to build realtime blurs on the GPU and how the best blur algorithm works
- "Dual Kawase"
blog.frost.kiwi/dual-kawase
ARTICLE (chrome’s glow-up story)
A history of Chrome
ESSENTIAL (rainbow science for nerds)
What is a color space?
ARTICLE (prompting like a boss)
Attention is the new Big O
ARTICLE (DIY Google vibes)
Let’s write a search engine
ARTICLE (handbook hustle)
How to Create Engineering Handbooks
ARTICLE (5-minute power move)
This 5-minute daily habit made me a productive software engineer in Big Tech
ARTICLE (cloud climber’s cheat sheet)
Resources to go from Zero to Hero in Cloud in 2025
ARTICLE (terminal magic tricks)
Terminal sessions you can bookmark
ARTICLE (bigger is sometimes better)
Use One Big Server
ARTICLE (Go brrr in containers)
Container-aware GOMAXPROCS
ARTICLE (AI truths lol)
The Bitter Lesson is Misunderstood
ARTICLE (Swift goes rogue)
How we built an interpreter for Swift
Want to reach 190,000+ engineers?
Let’s work together! Whether it’s your product, service, or event, we’d love to help you connect with this awesome community.
🛰️ Amazon’s Satellite Internet Unit Lands First Airline Customer (2 min)
Brief: Amazon's Project Kuiper scores its first airline partnership, bringing satellite-powered Wi-Fi to flights as it prepares to compete with SpaceX's Starlink service.
🚀 Atlassian Acquires The Browser Co. for $610M to Boost Web Productivity (2 min)
Brief: Atlassian snaps up The Browser Co., maker of AI-powered Arc and Dia browsers, for $610M, aiming to enhance enterprise workflows amid competition from Google and OpenAI.
📱 iPhone 17: Thinnest iPhone Ever & Everything to Expect from Apple’s Event (2 min)
Brief: Apple’s upcoming iPhone 17 is set to be the thinnest iPhone ever, with rumors hinting at major design changes, performance upgrades, and new features to be unveiled at the hardware event.
⚖️ Anthropic Reaches Settlement in Landmark AI Copyright Case (4 min)
Brief: Anthropic settles a high-profile copyright lawsuit over its AI models, setting a potential precedent for how generative AI systems handle training data and intellectual property.
📈 Google Stock Surges 8% as Antitrust Ruling Spares Chrome Divestment (2 min)
Brief: Google avoids worst-case antitrust penalties—no Chrome sale required—allowing it to retain search dominance and sending shares soaring, while Apple gains 4% on eased default-search payment terms.
💸 Tempo: A Payment-First Blockchain Backed by Stripe, Visa, and More (2 min)
Brief: Tempo, a new blockchain optimized for payments, launches with support from Stripe, Visa, and major financial players, offering 100K TPS, near-zero fees, and built-in privacy for remittances, payouts, and embedded finance.
This week’s coding challenge:
This week’s tip:
Use NGINX's proxy_cache_use_stale directive to serve stale content during backend failures or updates, implementing a robust circuit breaker pattern. This provides graceful degradation by serving slightly outdated content instead of errors when your backend services are struggling.
Wen?
High-traffic applications: Maintains service availability during traffic spikes or backend failures
Zero-downtime deployments: Serves cached content while new backend instances warm up
Microservices architectures: Prevents cascading failures when dependent services experience issues
"The art of programming is in the inspiration of the programmer, and in the tact of the machine."
unknown
That’s it for today! ☀️
Enjoyed this issue? Send it to your friends here to sign up, or share it on Twitter!
If you want to submit a section to the newsletter or tell us what you think about today’s issue, reply to this email or DM me on Twitter! 🐦
Thanks for spending part of your Monday morning with Hungry Minds.
See you in a week — Alex.
Icons by Icons8.
*I may earn a commission if you get a subscription through the links marked with “aff.” (at no extra cost to you).
Thanks for including my article, Alex
And… happy birthday 🎉