Research Intelligence

The sources and topics you follow, in one high-signal feed.

Ethmos summarizes the podcasts and newsletters you follow, sends research agents after the topics you care about, and compounds it all into a living encyclopedia built from your sources alone.

Free · set up in two minutes

This morning Sat · Jun 6
Newsletter Stratechery 05-11

Agentic inference unbundles the GPU

When machines run tasks for machines, latency stops mattering and memory beats speed.

  • Fast “answer inference” for humans, memory-bound “agentic inference” for machines.
  • The largest, most open market yet, where Cerebras, Groq, and even China can compete.
NVDAAgentic inference14 cross-refs
Your agent found

Tracking “AI compute supply & pricing” — across 1,200+ shows, including ones you don't follow.

Three analysts flagged a Q3 GPU oversupply on shows you've never subscribed to.

Newsletter The Diff 06-06

Why aggregation theory bends but doesn't break

Distribution still wins — but the cost of owning demand is quietly climbing.

Distilling All-In · Stratechery · Dwarkesh · Lex Fridman · The Diff · Invest Like the Best, and whatever else you follow.

The problem

The best thinking in your field has never been more available. Catching all of it is the hard part.

Keeping up with all the content buried in podcasts and newsletters is the work of a full-time analyst. Now you have one.

How it works

Pick your sources, set your agents, open your feed.

01

Pick your sources.

Add the podcasts and newsletters you follow. Ethmos listens to and reads every new one.

02

Set your agents.

Tell them what topics to track, and they scan the podcast universe for what matters.

03

Open your feed.

A daily brief of everything worth knowing, with deep dives one click away.

Ethmos summarizes at the depth you choose

It transcribes and reads every podcast and newsletter you follow, turning each into a card you can scan in a minute. You decide what to do with it: take the gist, open a fuller summary, or go all the way to the source.

Podcast Dwarkesh Patel 1h 48m

Scaling laws, and what breaks first

The bottleneck moved from chips to power — and nobody priced it in.

  • Frontier runs are now gated by grid interconnects, not GPUs.
  • Inference demand is compounding faster than training spend.
NVDAPower14 cross-refs

Tracks what you care about, everywhere it comes up

Name a topic and Ethmos puts agents on it, reading past your own subscriptions to surface the conversations that matter from across the field, even on shows you've never heard of.

PodcastBG2 Pod
The capex debate, revisited
Agent · Datacenter buildout

A hyperscaler CFO walked back 2026 guidance on a show you don't follow.

NewsletterSemiAnalysis
HBM supply through 2027

And retains it all in a living encyclopedia.

Everything Ethmos reads compounds into a curated encyclopedia of your field, built from your sources alone. It remembers, so you can search, revisit, and build on it. The longer you use it, the more it knows.

Living encyclopedia 674 sources
AI Compute Infrastructure138 sources
Aggregation Theory58 sources
Semiconductor Supply69 sources

Your analyst is ready.

Get started

Free · two-minute setup