This week’s system design refresher:
Agent Harness, Explained Visually
What Keeps the Internet Running?
Common Network Protocols Every Engineer Should Know
Why Is Nginx So Popular?
What is a Firewall?
Agent Harness, Explained Visually
An LLM can reason and generate text, but that alone can be unreliable. The reliability usually comes from the harness, the layer around the LLM shown in the diagram below.
The user sends a query into the harness and the context builder converts the query into a prompt using its knowledge base.
The LLM reasons over the provided prompt and proposes some actions (functions to call)
The policy gate checks actions and determines whether they should be allowed or blocked.
The runtime first executes the tools and sends the results back to the LLM.
Finally, the verify step runs tests on the result and includes the feedback in the context for future iterations.
Once the verify step passes, the harness shows the result to the user.
Aside from these, two other layers exist and are shared between all stages. The observability layer tracks things like metrics, and errors. The constraints layer sets the rules and permissions so the system can enforce them.
Over to you: Where do you think most agent failures happen, in context, tools, constraints, or verification?
What Keeps the Internet Running?
When you type a URL on your address bar in your browser and hit enter, you get a web page.
But behind that, there is a whole stack of protocols running to make it possible.
The first thing that happens when you type the URL is a service called DNS, which takes the domain name and finds out the IP address of the website, where the site is hosted.
Also, there is this DHCP that basically assigns your machine an IP address dynamically when it boots up. There is another protocol called ARP that is used to find out the MAC address of the machine where you want to send the packet to go out of the network. That’s the local hop nobody thinks about.
And then most likely, TCP is used for communication. It’s a reliable connection. Reliable means nothing goes missing on the way. And also a TLS protocol that basically wraps the whole thing to avoid eavesdropping and for secure communication. That’s the padlock in the address bar.
If we zoom out and look at the bigger picture, there is this BGP protocol that basically connects thousands of networks which are into the Internet together. And then it decides basically how your packet crosses the Internet. No one’s really in charge of that map, which is kind of wild. Similarly, we have OSPF, which is the protocol used to route traffic within a network.
And then to carry frames on the wire or on the air, you have protocols like Ethernet, Wi-Fi, or PPP.
All of that runs for one web page, we probably forget they exist until the day it breaks.
Common Network Protocols Every Engineer Should Know
Ever wonder what actually happens when you click "Send" on an email or join a video call? Every click, message, and API call on the internet relies on network protocols. They define how data moves, who can talk, and how securely it all happens.
At the foundation are transport protocols: TCP ensures reliable delivery, UDP prioritizes speed, and QUIC brings both worlds together over UDP.
On top of that, HTTP powers the web, TLS secures it, and DNS translates names into addresses.
Need remote access? That’s SSH. File transfers? SFTP or SMB.
Real-time chat and media? WebSocket, WebRTC, and MQTT keep data flowing live.
For identity and access, OAuth and OpenID handle authorization and authentication.
In the backend, DHCP, NTP, ICMPv6, and LDAP quietly keep everything synchronized, addressed, and discoverable.
From simple emails (SMTP, IMAP) to encrypted VPNs (WireGuard, IPsec), these protocols form the invisible language that keeps the internet connected and secure.
Over to you: If one protocol suddenly stopped working worldwide, which one would break the internet first?
Why Is Nginx So Popular?
Apache dominated web servers for 20 years, then Nginx showed up and changed everything. Now Nginx powers some of the largest sites on the internet, including Netflix, Airbnb, Dropbox, and WordPress. com. Not because it's newer or trendier, but because it solves problems that Apache couldn't handle efficiently.
Here’s what makes Nginx so popular:
High-Performance Web Server
Reverse Proxy & Load Balancer
Caching Layer
SSL Termination (Offloading)
Over to you: What’s your primary use for Nginx today, web server, reverse proxy, or load balancer?
What is a Firewall?
Every time you connect to the Internet, a firewall quietly decides what can come in and what must stay out. A firewall is your network’s first line of defense. It filters traffic based on rules you define, by IP address, protocol, port, program, or even keywords. Every packet that tries to enter or leave your network passes through this checkpoint.
There are two main types:
Network Firewall: Sits at the network edge between your infrastructure and the internet. Can be physical hardware, virtualized software, or cloud-deployed service. Operates at Layer 3-4 of the OSI model. Filters traffic based on IP addresses, protocols, and ports before it ever reaches your internal network.
Protects the entire network at once. This is your first line of defense. Internet traffic hits the network firewall before it reaches your router, before it touches any internal systems.Host-based Firewall: This runs as software on individual devices, like your laptop or a server. It works at Layer 3–7, inspecting packets more deeply and protecting only that specific device.
Your desktop has its own host firewall. Your server has its own. Each one is configured independently. It’s your last layer of defense in case something slips past the network firewall.
Together, they form a layered shield, keeping unwanted traffic out while letting legitimate communication flow freely.
Over to you: Have you ever had to troubleshoot a misconfigured firewall rule that accidentally blocked something critical? What was it, and how long did it take to find?