Note on Big Muddy

The definition of "agent"

Sat, 04 Apr 2026 23:59:00 -0400

An interesting exchange between Guido van Rossum and Andrej Karpathy a few days ago on Twitter:

Guido van Rossum: I think I finally understand what an agent is. It’s a prompt (or several), skills, and tools. Did I get this right?

Andrej Karpathy: LLM = CPU (data: tokens not bytes, dynamics: statistical and vague not deterministic and precise) Agent = operating system kernel

The triumph of the data raccoons

Fri, 03 Apr 2026 23:40:00 -0400

My PhD co-supervisor at the University of Toronto, Dr. David Fisman, liked to use the term “data raccoon” to describe the work of using messy, incomplete, hard-to-work-with data to do serious research. Or, as he described it in testimony to the Canadian House of Commons in May 2020 (emphasis mine):

I’ll tell you, my group at University of Toronto call ourselves “data raccoons”, because we’ve sort of managed to thrive for about 15 years on data that most people regard as garbage, so it’s sort of a bit of the normal state of affairs for us with public health data analysis.

It’s an unmistakably Toronto metaphor—the city isn’t called the raccoon capital of the world for nothing!

It occurred to me recently that data raccoons have basically taken over the world. The basis of the AI revolution is vast quantities of text dredged from the Internet, none of which was written for its final purpose of training the deus ex machina. Arguably the most important dataset for training LLMs has been Common Crawl, a mostly uncurated snapshot of the Internet that has been running since 2007. According to a Mozilla report from 2024, Common Crawl was used in two thirds of LLMs developed in the formative period between 2019 and 2023, and the archive also comprised 80% of tokens in OpenAI’s GPT-3. Unsurprisingly, the Common Crawl Foundation has received financial support from AI companies in recent years, all the while being accused of abetting these same companies to train their models on paywalled articles.

Andrew Gelman's blog schedule

Thu, 02 Apr 2026 16:24:00 -0400

Andrew Gelman, professor of statistics at Columbia University, runs one of my favourite blogs on the Internet. He has been writing there for over 21 years, since October 2004. Many of his collaborators also contribute to the blog, but he is the primary author. In a 2024 post celebrating 20 years of blogging, Gelman mentions having over 12,000 posts. This is a cadence of over 1.6 posts/day sustained for two decades!

One of the more unusual things about Gelman’s blog is that most posts are not particularly topical. Sure, many posts are time-sensitive, posting about upcoming events or commenting on recent publications (like doing damage control on deeply flawed papers like to receive attention). But there is generally one non-topical post each day. A line in a recent post caught my eye:

As regular readers know, our posts are usually on a 6-month lag, but this one is so important I had to share it with you right away.

As a regular reader myself, I was aware of the delayed posting schedule, but out of curiosity, I wanted to see how far back this habit went. Here’s the rough timeline I came up with:

In 2011, Gelman wrote that his “non-topical blog entries are on approximately one-month delay”.
In 2012, he referred to “stacking up posts here with a roughly one-month delay”.
In 2014, he said that “most of the posts here are on a 1 or 2 month delay.”
In 2016, he casually mentioned “our 2-month delay”.
Later that year (August 2016), in a post literally titled “My next 170 blog posts”, he said he had filled “the blog through mid-January” and had “170 blog posts in the queue.”
By 2018, he mentioned the blog was “mostly on a six-month delay”.
In 2019, he referred to “our 6-month blog delay.”
In 2022, he wrote: “Usually I schedule these with a 6-month lag, but this time I’m posting right away”.
In February 2026, he said the “current end of the blog queue is in early July”.
Then, in April 2026, came the latest “usually on a 6-month lag” remark.

It seems the blog had about one month of content in the publishing pipeline by 2011, ramped up to one to two months by 2014, two months by early 2016, and finally jumped to six months by August 2016, where it been ever since. Quite the arsenal of scheduled content!

Testing ZeroClaw, Part 2.5: ZeroClaw is alive!

Wed, 01 Apr 2026 23:59:00 -0400

Yesterday, I wrote about how the ZeroClaw GitHub repository had been down for two days with little explanation. Earlier today, the project provided a little more information on Twitter:

They flagged our org which is why we’re down. Code is safe and we’re still working, just waiting for @github

Since March 30 (the day after their repo started 404ing), they project has been promising a blog post to explain the situation. As of now, that post is now available:

Over the past few days, a maintainer used aggressive AI automation to review and merge PRs:

Merges went through that shouldn’t have.

In the process of trying to undo the damage, the maintainer’s GitHub account was flagged, which triggered enforcement actions on the ZeroClaw org itself.

That maintainer has been removed from the project.

This sounds strikingly similar to the incident that occurred about a month ago, which I also mentioned in yesterday’s post:

Earlier today, during routine maintenance, the visibility of the ‎`zeroclaw-labs/zeroclaw` repository was accidentally changed from public to private and was later restored to public.

After reviewing the GitHub API audit logs and collecting detailed feedback from our engineers, we confirmed that the incident was caused by improper use of an AI agent tool during maintenance.

Obviously, the use agentic workflows in open source projects is an emerging field where best practices have not yet been established. The case of ZeroClaw should be a warning to other projects to keep human review in the loop, or at least to limit the autonomy of agents when a project has numerous contributors. As they say in their blog post:

Testing ZeroClaw, Part 2: ZeroClaw is dead?

Tue, 31 Mar 2026 20:57:00 -0400

Earlier this month, I wrote about setting up one of the many lightweight OpenClaw alternatives, namely ZeroClaw. I had some issues with initial setup, but I got to the point where I could talk with my bot over Telegram.

Some of my initial enthusiasm for ZeroClaw was dampened by the divergence between the docs and the features available in the release build. The release build was quite out of date due to the breakneck pace of development. In the week or two following my initial setup, the release build pipeline was broken, so even when they released a new tag, there were no new precompiled binaries available. Being forced to compile the Rust binary yourself kind of goes against the project’s philosophy of ultra-low resource consumption.

They eventually fixed the release pipeline and I started casually working on a system where I could send notes and ideas for blog posts to my bot through Telegram and have it turn them into structured Markdown files.

But two days ago (March 29), I noticed that the ZeroClaw GitHub repo was 404ing. On the same day, the project posted the following on Twitter:

Our GitHub repo is currently returning a 404 for some users. We’re aware and actively investigating. The repo is public and all code is safe.

One important fact about for-profit plasma donation

Mon, 30 Mar 2026 21:51:00 -0400

For-profit plasma donation is in the news today in Canada. Two people recently died after giving plasma at Grifols for-profit plasma clinics in Winnipeg, Manitoba, although Health Canada has yet to find a link between the plasma collections and the deaths. Today, it was reported that a Grifols clinic in Calgary, Alberta was found non-compliant during an inspection in December 2025:

The inspection found the Calgary centre didn’t accurately assess whether donors were suitable, didn’t collect blood according to its Health Canada authorization, didn’t thoroughly investigate errors and accidents, and didn’t carry out sufficient corrective and preventative actions.

This is obviously a problem for for-profit plasma collection in Canada, where the practice is already controversial. Paid plasma collection is illegal in Canada’s three largest provinces: Ontario, British Colombia, and Quebec, though Ontario allows a few for-profit clinics to operate through an agreement with Canadian Blood Services, Canada’s independent blood authority. British Colombia and Quebec together make up over 35% of Canada’s population; including Ontario, it’s nearly 80%. Besides Ontario, for-profit clinics exist in some other smaller provinces.

Vocal advocacy exists against paid plasma collection, leading to municipal resolutions against the practice in Ontario, even as clinics open. This advocacy is often premised on the fear that paid plasma will undermine voluntary donations. To my mind, the central fact in the for-profit plasma collection debate is that only a handful of countries are self-sufficient in plasma collection, and all of them allow for paid plasma collection. They are: the United States, Germany, Czechia, Austria, and Hungary (Egypt may have also recently joined the list). While other countries, like Canada, may have achieved self-sufficiency for plasma for direct infusion, no other country can meet its own needs for plasma-derived medical products. The world relies on a small number of self-sufficient countries, primarily the United States, to meet the demand for plasma products.

Opt out of very new Python package versions with uv

Sat, 28 Mar 2026 08:43:00 -0400

In light of several recent Python package compromises (litellm, telnyx), here is a useful tip from Hacker News commenter mil22:

For those using uv, you can at least partially protect yourself against such attacks by adding this to your pyproject.toml:

[tool.uv]

exclude-newer = "7 days"

or this to your ~/.config/uv/uv.toml:

exclude-newer = "7 days"

This will prevent uv picking up any package version released within the last 7 days, hopefully allowing enough time for the community to detect any malware and yank the package version before you install it.

Commenter notatallshaw follows up with how to achieve similar behaviour in *pip*:

Pip maintainer here, to do this in pip (26.0+) now you have to manually calculate the date, e.g. –uploaded-prior-to="$(date -u -d ‘3 days ago’ ‘+%Y-%m-%dT%H:%M:%SZ’)"

In pip 26.1 (release scheduled for April 2026), it will support the day ISO-8601 duration format, which uv also supports, so you will be able to do –uploaded-prior-to=P3D, or via env vars or config files, as all pip options can be set in either.

Editors hate this one weird trick

Thu, 05 Mar 2026 20:05:00 -0500

Given my recent posts on AI in academic publishing, I just wanted to share this joke from Prof. Arthur Spirling on Twitter:

Actually you cant run my paper through Claude to desk reject it because Claude is a regular coauthor of mine. Conflict of interest. Checkmate, editors

Testing ZeroClaw, Part 1: Setup

Mon, 02 Mar 2026 19:15:00 -0500

As mentioned last week, I’ve been meaning to test out a personal agent from the Claw-like ecosystem. I settled on testing out Zeroclaw, a popular and lightweight OpenClaw alternative that should run well on my Raspberry Pi 4 4GB.

I wanted to harden my setup as much as possible and opted to running everything in Docker. I started with the official Docker compose file and added my OpenRouter key. I brought up the pre-built container image and tried sending the basic “Hello” message to the agent using the CLI. However, I got error because the automatically generated config file defaulted to a version of Claude Sonnet 4 that wasn’t available on OpenRouter. I switched to claude-sonnet-4.6 and then gpt-oss-20b (for much cheaper testing).

The Zeroclaw web gateway was a bit of a mess. Of the features I tried, only memory management and the basic status dashboard worked. Trying to talk to the agent through the web interface would give me a black screen (here’s someone complaining about the same error). I’m still being charged for the tokens, though! The cost tracker always displayed zero, even as I sent CLI and Telegram messages (more on that soon). The configuration editor gave me an error and so did the diagnostics tool.

The project docs/wiki were helpful for figuring things out, but development is running so far ahead of releases that a bunch of the features referred to aren’t available in the current stable version (v0.1.7, from last week). This includes getting and setting specific config options from the CLI and resetting the gateway pairing token. To use these features, you have to compile yourself.

Some examples of just-build-things-ism

Sun, 01 Mar 2026 11:58:00 -0500

The best mantra to come out of the AI era is: “You can just build things”. (So good OpenAI ripped it off for their Super Bowl ad.)

I’ve been pretty inspired to see how many people are now building all kinds of incredible tools thanks to advances in AI coding agents, even if they have no previous background in coding (see my post on Havelack.AI from a few days ago).

Here are a few more examples I’ve been following:

Canadian journalist Alex Panetta writes about his AI-augmented workflow at A.I. For You. I first came across his work with his debut article “I killed my doomscrolling habit with AI. You can too”. In it, he explains how to vibe code an automated, personalized daily news digest. I’ve tried to build something for myself but I haven’t gotten it quite right yet. A great follow for big news consumers.
Economics professor Scott Cunningham, author of the great textbook Causal Inference: The Mixtape, has a presentation explaining how to encourage AI adoption among academic faculty. This starts with faculty experiencing a killer use case for AI, which he suggests is building slide decks. He shares his tools/agent skills for this use case and more on GitHub.
Another economist, Chris Blattman, built a website to share the productivity tools he developed with Claude Code. He provides a tutorial and code on Claude Blattman.

And of course, Simon Willison has been building and sharing tools habitually for years now.

Big Muddy turns one month old

Sat, 28 Feb 2026 10:57:00 -0500

It’s been one month since my first post on Big Muddy. There were a few factors driving my decision to start this project:

I wanted to get into the habit of writing every day.
My “random interesting links” folder was overflowing, but I wasn’t doing anything with these links.
My admiration for Simon Willison’s work and his suggestion for everyone to start a blog to share what they learn.

As the saying goes, writing is thinking. Instead of allowing interesting articles, tools, and bits of knowledge to languish in a “temporary” bookmarks folder, I could actually engage with and learn from the material by writing something about each item and make it easier to re-find later. This also forces me to curate the links and ideas that are actually worth saving, since writing a post, even a short one, takes a lot more effort than just throwing a link into a folder.

I figured I might as well share the results with the world, since someone else might find this information useful. And I’m helping to write my ideas and preferences into the next generations of LLMs, I guess.

I’ve made exactly one post per day since starting this blog, which was my goal when I set out. A handful of these posts are pre-written the day before (if I know I won’t have the opportunity to write a post the next day), but most are written the day of. Most are short (Bash tells me just over 190 words on average, though this is slightly inflated by Markdown formatting). Some are very perfunctory, just a link with a few words, when I really needed to get a post out for the day. At the start of this project, I cut my “temporary” bookmarks folder to zero. It has now been replaced with a backlog of links I want to write about on this blog.

It's incredibly easy to game Twitter's trending news algorithm

Fri, 27 Feb 2026 20:30:00 -0500

Twitter’s “Today’s News” section is a mix of real news, very minor stories (usually discussion of a random AI-related post), nonsense trends, and barely disguised marketing.

The algorithm behind it seems pretty easy to manipulate.

This trending topic revolves around an explosive DM warning of imminent 25% layoffs at a FAANG company:

Here is the original post, which comes from an account called Tech Layoff Tracker (@TechLayoffLover):

There is no reason to believe this post is real. The account, created this month (February 2026), made its first post 7 hours ago. The post in question was made 5 hours ago, or 2 hours after the account’s very first post. Of course, the account carries an utterly meaningless blue “verified” checkmark.

But despite all this, the news summary puts “Tech Layoff Tracker” right in the headline, as if it’s a known reliable source and not an account (most likely) created the same day as the summary itself!

These academic journal AI policies aren't going to last

Thu, 26 Feb 2026 16:51:00 -0500

I recently came across the following policy on the submission page of an academic journal:

Use of Artificial Intelligence (AI) tools: One of the goals of Spectrum is to stimulate critical thinking and skill development among authors and reviewers alike. Spectrum discourages the submission of content generated by artificial intelligence (AI)-assisted technologies (such as chatGPT and similar tools). This includes tools that generate text, data, images, figures, or other materials, as well as tools that are used to summarize and synthesize sources. Authors should be aware that such tools are vulnerable to factual inaccuracies, biases, and logical fallacies, and may pose risks to privacy, confidentiality, and copyright.

If authors choose to submit work created with the assistance of AI tools, such use must be disclosed and described in the submission. The disclosure must include: 1) what system was used, 2) who used it, 3) the time/date of the use, 4) the prompt(s) used to generate the content, and 5) the content in the submission that resulted from use of AI tools. The output from the AI system should also be submitted as supplementary material. Authors must accept full responsibility for the accuracy and integrity of the submission. AI systems do not meet the criteria for authorship, and should not be listed as a co-author.

Film recommendation: Bugonia

Sat, 21 Feb 2026 23:27:00 -0500

I watched Bugonia (from director Yorgos Lanthimos) blind tonight, and I highly recommend it. The film is centred on a broken man who loses himself in conspiracy theories to cope with his tragic circumstances, but it’s also so much more than that. It features outstanding performances by Jesse Plemons and Emma Stone, as well an absolutely kidney shredding score.

Looking up the film for this post and I see it was nominated for Best Picture this year. I’m not surprised. Definitely a great watch, having known nothing about the film going in beyond the one sentence description.

Bugonia is available to stream on Amazon Prime in Canada and probably elsewhere.

A handful of composers created most classic RPG soundtracks

Sun, 08 Feb 2026 22:36:00 -0500

I’ve always been a big fan of soundtracks, and video game soundtracks are no exception. Buying games on GOG.com usually nets you the soundtracks as well, so recently I’ve been enjoying a lot of classic RPG music. What struck me was how few composers were responsible for creating the ambiance of so many beloved classics. Look at how many series are covered by just the following six composers:

Inon Zur (Icewind Dale II, Dragon Age: Origins, Dragon Age II, Fallout series starting with Fallout 3 plus Fallout Tactics, co-composer for Baldur’s Gate II: Throne of Bhaal and Pathfinder: Kingmaker, additional music for Neverwinter Nights)
Jeremy Soule (Neverwinter Nights, Icewind Dale, The Elder Scrolls series starting with Morrowind, Star Wars: Knights of the Old Republic)
Justin E. Bell (Pillars of Eternity series, Tyranny, The Outer Worlds)
Mark Morgan (Fallout, Fallout 2, Planescape: Torment, Torment: Tides of Numenera, Wasteland 2, Wasteland 3)
Kirill Pokrovsky (Divinity series up through Divinity: Original Sin)
Borislav Slavov (Divinity: Origin Sin II, Baldur’s Gate 3)

Of the above, I highly recommend the truly excellent Divine Divinity soundtrack (terrible title, great music!), as well as Baldur’s Gate 3, particularly the vocal songs like “Down by the River”, “I Want to Live”, and “The Power”.

To me, it emphasized just how hard it is to break into this industry commercially, as these famous names and a handful of others will (deservedly!) continue to get work on the small number of major projects that get published every year. I worry that the less prestigious work that helps pays the bills/build experience for the large majority of composers who have yet to achieve name recognition will increasingly go to AI, impoverishing the pipeline for tomorrow’s great video game soundtrack composers.

Welcome to Big Muddy

Wed, 28 Jan 2026 23:00:00 -0500

Hi, I’m Jean-Paul R. Soucy, a data scientist working in healthcare in Montreal, Canada. Welcome to Big Muddy, my spin on a Simon Willison-style links-and-notes blog. Here I collect and share things I’m learning across technology, science, politics, and whatever else catches my interest. You’ll find interesting links, brief write-ups, quick experiments, and the occasional deep dive.