Menu
 Sign Up
Journal Club:

The latest academic research, explained in plain English.

Journal Club is a premium daily newsletter and podcast authored and hosted by Malcolm Diggs. Each episode is lovingly crafted by hand, and delivered to your inbox every morning in text and audio form.

Journal Articles
Join the Club Listen to Samples
Scroll down to listen to samples.
Journal Articles

Free Samples

(A few previous episodes)

A Low-Cost Deep-Learning-Based System for Grading Cashew Nuts

In Computer Science we don't talk about cashews very often, but maybe we should. A team of researchers in Vietnam just trained a computer-vision classifier on top of YoloV8 that can grade cashew nuts as they're coming down a conveyor belt. It can perform the grading in milliseconds, with less than a 3% error rate, running entirely on cheap off-the-shelf components. Impressive!..

0:00 0:00

TCP Stratos for Stratosphere Based Computing Platforms

A few companies are planning on putting their datacenters in the stratosphere. Yes, the stratosphere! Their theory is that the sheer amount of free cooling would justify the costs and headache. For a moment let's put aside how they're going to get the machines all the way up there, and then let's put aside how they're going to keep them afloat. There's still one big problem: TCP isn't going to work very well in the stratosphere. Not well at all. That's according to a new study in which the researchers are proposing an alternative new variant of TCP for use exclusively in the stratosphere. They're calling TCP Stratos...

0:00 0:00

Inferring Trajectories of Psychotic Disorders Using Dynamic Causal Modeling

The field of Psychiatry doesn't always lend itself well to data-science. The field of Computational Psychiatry, on the other hand, thinks we can do better. In this new study, researchers built a proof-of-concept model that should be able to take a new patient (who is admitted for depression, mania or psychosis), review their history, and then project forward the ultimate progression and possible treatment of their condition. Far fetched? Let's dive in and find out...

0:00 0:00

The Next Episode

(sneak preview)

Bell icon
Bell icon

The next issue of this newsletter will be sent out at 6:00 am (PT).

If you don't sign up before the release time, you'll miss that episode forever. There is no archive, you only get access to the issues that are released while your membership is active.

  • 00

  • Hours

  • 00

  • Minutes

  • 00

  • Seconds

The Topics We Cover

This is a high-level overview. To see the list of past episodes, go to the archive.

Machine Learning (A.I.)

We'll keep you up to date on the latest training and implementation techniques within Machine Learning, Computer Vision, Expert Systems, Natural Language Processing, Robotics, Autonomous Driving, and all forms of applied AI.

Compute Infrastructure

As the hyper-scalers grow and evolve, so do their compute offerings. As new chips and materials are developed, new types of computing become possible.

Programming Language Evolution

Whether it's the latest GPGPU programming experiments, new dev-tools for Rust, or new packages for Python, we keep up with the latest in programming language evolution.

Distributed Systems

From farm equipment to satellites, the world is increasingly comprised of interconnected distributed systems. We analyze and explore many of them.

Algorithmic Advancements

The sea of algorithms only seems to get wider and deeper. Whether it's new algos for hashing, search, bin packing, or anything else: we'll keep you up-to-date on the latest and greatest.

New Data Stores and Structures

The ways we can store and retrieve our data are changing all the time: Just in the past few years CRDTs, vector databases, streaming queues, novel blockchain storage engines, and realtime dbs have changed the way applications are built.

New Programming Paradigms

The best-practices for application development are in constant upheaval: Recently durable-execution environments, quantum computing, serverless state machines, reactive systems, gRPC clusters, and multi-cloud environments have shook things up.

Security Improvements

It's never been more difficult or more complex to secure an application. The rash of recent breaches and leaks affirm this fact. As applications get more complex, their attack surface only grows. We'll help you keep track of the latest best practices in securing your stack.

Join the Club

The next issue of Journal Club will be sent out in: 00: 00: 00.

Sign up to get the next 365 issues of Journal Club. One issue will be delivered to your inbox every morning (at 6am Pacific Time). You will also be able to pull one additional past-episode from the archive each day. If you aren't 100% satisfied just let us know in the first 30 days and we'll refund all your money. Here's the refund policy.

1 Year of Journal Club

(prepaid, not a subscription)

$4.99/mo billed yearly at $59.88 total.

  • Get the next 365 issues of Journal Club. One issue will be delivered to your inbox every morning.
  • Every issue comes in text and audio form, and includes a link to the paper's original source/pdf.
  • You will also be able to pull one past episode from the archive every day.
  • This membership can be gifted to someone else (during checkout)
  • Does not auto-renew next year. This is not a subscription.
  • If this isn't the best newsletter you've ever joined, let us know in the first 30 days for a full refund. Here's the refund policy.
Add to Cart
30-Day Money Back Guarantee

Frequently Asked Questions

We create a new issue of our newsletter (in text and audio form) every morning, and email it to all our current members. Every issue is a breakdown of a recent academic research paper from a peer-reviewed journal. We try to pick the most interesting/impactful article each day. When an issue arrives in your inbox you can read our breakdown from that email, or you can click the provided link to listen to the audio version (narrated by Malcolm). We'll also provide the link for you to download the original paper in PDF. If you scroll up on this page you'll be able to listen to some free samples to get a feel for the episodes.

Yes, you can pull one episode from the archive per day. Just go to the Episode Archive and follow the instructions to have any episode resent to you via email.

Anywhere from 5-15 minutes from beginning to end. Or 750-2500 words for the written version. It depends on the paper.

Every issue provides the direct link to the download page for the article being discussed (the DOI url).

Every day we browse DOAJ and other article databases for new manuscripts. We manually sift through them until we find something interesting. If an article makes it to the "interesting" pile, then we'll read and scrutinize it, and try to determine if it's the kind of paper that would make a good episode. Some are good candidates, but most aren't. Eventually we find the winner-of-the-day and manually write, record and edit the episode.

Nope, everything is made the hard way: manually, by hand. By the nature of our content, you can probably guess that we're not opposed to AI or technological advancements. But, they're not good at everything. As of this writing, it's our position that Large Language Models (LLMs), CNNs and all other Expert Systems are still really bad at making sense of novel information (information that doesn't comport with the system's prior understanding of the world). By its nature, academic research is always novel. So its exactly the kind of content that LLMs (and others) would have trouble grappling with. Generally speaking, we believe in using AI when it can do a job objectively better than a human can. But at this point that's still not the case. A human expert reviewing a research paper and delving into the nuance and implications of that research still (in our opinion) vastly outperforms AI at the same task. Maybe that will change, but until it does, we'll be making these episodes the old fashioned way.

Also, AI voice synthesis (text-to-speech) is almost there, but still not quite. The voices sound nice for demos, but when you're explaining a complex topic we've found that a human narrator will naturally emphasize certain words in a way that TTS can't do yet.

We're happy to take a look at it, but we can't guarantee we'll make an episode out of it. Just email: [email protected] and please include the DOI. Please note: we're not making episodes about pre-prints or whitepapers at this time. And we can only make episodes about Open-Access (cc-by) papers published in peer-reviewed journals.

We're using Stripe Checkout. So, it'll accept cards and a few other country-specific methods depending on where you're located. If you're using an Android device, you'll probably see a Google Pay button, and if you're on iOS you'll probably see a Apple Pay.

Sure just email [email protected] and we'll email you back a Paypal checkout link.

Yes of course. If you're not happy, let us know in the first 30 days and you'll get a full refund. Partial (prorated) refunds are available after that. Our full refund policy is here.

No.

No.

Shoot us an email: [email protected]