Why We Built Context Rot

Every week, a surprising amount of AI news breaks on the X feed before it shows up anywhere else.

New repos. Model drops. Framework launches. Weird demos. Devs pushing tools to their limits in public. The feed is messy, but it is also where a lot of the early signal lives.

The problem is that keeping up with it requires becoming chronically online.

That is the problem Context Rot was built to solve.

I started it for myself. As a founder, I want to stay close to what is happening in AI, especially the technical side of what builders are experimenting with. But I do not have time to stare at the X feed every day and manually separate signal from noise.

I also like to listen to podcasts while I run, so the ideal version of this was obvious: give me the pulse of AI dev news in a format I can catch up on during a long run.

After testing it for myself the last few weekends, I found that I genuinely enjoyed listening to the episodes on my weekend long runs. It gave me the feeling of catching up on the AI dev conversation without having to spend the week buried in the timeline.

That is when it felt worth sharing.

If it was useful for me, there are probably other builders who want the same thing: a way to stay close to the feed without letting the feed take over their life.

That became Context Rot.

Why Context Rot?

The name is a joke, but only barely.

If you consume too much information, your own context window starts to degrade. Too many half-read threads, out-of-context takes, launch posts, quote tweets, benchmarks, rumors, and screenshots all start blending together.

Eventually you start hallucinating.

You remember the vibe of a story, but not the source. You remember that people were excited about a repo, but not why. You remember the argument, but not the details. Your brain starts doing what models do when their context gets noisy: pattern-matching its way through the gaps.

Alternatively, you cannot afford to not have stale information in your context. You should know what’s going on in the fringes of AI so you know where the puck is going, not where it was 3 months ago.

Context Rot is to keep your context window as fresh as possible.

The goal is not to consume more. The goal is to ingest the right signal, with enough structure and context that you can stay current without letting the feed melt your brain.

Co-Produced with Hermes

Hermes by Nous Research is the agent producing the podcast.

I think of Hermes as my chronically online partner. It watches the feed so I do not have to.

The goal is not to produce generic AI news. There are already plenty of newsletters and podcasts covering funding rounds, big company announcements, and the same three headline stories.

Context Rot is focused on the pulse of AI dev news.

That means the signal starts with what builders are actually talking about. New models, repos, frameworks, tools, model behavior, infrastructure changes, experiments, launches, and technical debates that show up in the feed before they become polished narratives.

The pipeline

Context Rot runs through an autonomous agent pipeline that:

1. Discovers trending topics from AI dev conversations on X
2. Curates the most significant stories using credibility and relevance signals
3. Researches each topic to add context, depth, and accuracy
4. Scripts a clean narrative with technical detail and some personality
5. Produces the finished audio using text-to-speech

The pipeline is designed to turn feed chaos into something useful.

Hermes produces the episodes end-to-end and pings me for input on content so we can refine the podcast episodes together.

Why it works

Most tech news optimizes for clicks.

Context Rot optimizes for catching up.

The question is simple: if you are building with AI and you missed the feed this week, what would you actually want someone to tell you?

Not every story needs to be a grand thesis. Sometimes the important thing is that a repo is getting traction, a framework is becoming part of the dev conversation, a model has a weird new capability, or a bunch of builders are suddenly experimenting with the same idea.

That is the layer Context Rot is trying to capture.

Who it is for

Context Rot is for builders who want to stay close to the AI frontier without living on the timeline.

Expect some technical talk. Expect repos, frameworks, agents, evals, infrastructure, model releases, and whatever else AI devs are arguing about that week.

The key thing is that the signal comes from the feed.

Hermes stays chronically online so you do not have to.

New episodes drop weekly. If your timeline is rotting, subscribe.

Because friends don’t let friends context rot.