Intro

I’ve been seeing a few folks talking about Amp recently, which is an agentic coding tool built by Sourcegraph, the folks who had built Cody. I have worked with Cody before and had a great time with it before it got overshadowed by Windsurf/Cursor etc, so I wanted to give Amp a try as well. Now, this is NOT a full fledged review but only my first hands on impressions from implementing a mid sized feature over the weekend. Thanks to Quinn (CEO of Sourcegraph/Amp) who gave me a bunch of extra credits to try it out.

AdTechGod recently posted an interesting question on twitter “How many Super Bowl advertisers return year after year, and how much churn is there? I assume the ROAS isn’t strong enough to justify long term commitments.”

This got me thinking about the Super Bowl ads and how they have changed over the years, and I set out to do some analysis on it.

First steps

At the outset, I needed to get some information about the ads shown during Super Bowl in the past years. I found a few data sets which were fairly incomplete but a good starting point nonetheless, and I did a quick and dirty pivot table to get an initial insight.

Running Pi-hole or AdGuard Home provides great network-wide ad blocking, but creates a single point of failure. When your DNS server crashes or needs maintenance, your entire network loses DNS resolution. I had this problem and thus far had chosen to do a sub-optimal solution of using both AdGuard DNS as well as an upstream Google DNS server in my internet gateway. dnsmasq running on the gateway that manages DNS requests from all the clients in my network has no concept of a failover DNS server, or server priorities. It dispatches requests to all the servers it has configured, and ideally should honor whichever servers responds first.

Around an year ago, I spent a few weeks setting up the “perfect” workout tracking system. Custom spreadsheets, progress graphs, integration with my calendar, and so on. I was proud of this beast of a system. But you know what I didn’t do all those weeks, and several following ones? Actually work out.

Sounds familiar?

We’ve all been there - caught in the trap of over-engineering solutions to problems that could be solved with a simple notebook and pen. We’ve become so good at planning, organizing, and systemizing that we’ve forgotten to just…do things.

Aravind Srinivas (Perplexity) recently talked about how advertising might work in a future dominated by AI. His vision? A world where we never see ads because our AI agents handle them all behind the scenes. Sounds neat, but I’m not entirely convinced.

Here’s the thing. What Aravind describes isn’t really advertising. It’s more like super powered comparison shopping.

Taking his example of “Book me two nights at Rambagh”, the “human” equivalent of this is a human agent (Either yourself or a physical agent). One who goes to let’s say MakeMyTrip, or Google Flights, or Booking/Agoda/Airbnb etc to see and compare deals and then booking the best one. Maybe the human agent will also take decisions around the location, and the view, and the services/reviews besides just the cost to make it a well rounded decision. But essentially, this is a filtering of offers, not advertising.