I work on software that talks to mass spectrometers and it consistently refuses to refactor even an input file parser, presumably because it can infer it’s related to biology? Useless indeed.
I was reverse engineering a medical device, and had to do a lot of trickery to get Opus 4.5 - not even Fable/Mythos, Opus - not to trip up its fucking CBRN filter.
What happened with Fable is basically what I feared when they announced those restrictions. They took the shitty Opus CBRN filter and made it even worse.
I pity the fools trying to use Anthropic AIs for anything biotech.
Opus has been fine on proteomics and bioinformatics for me. I have never seen a Claude model refuse on such grounds before in the past.
Claude is still the best IMO, but it feels like its most frustrating and grating aspects are not down to the model’s abilities, but the increasingly heavy hand of Anthropic expressing itself within the model. Fable’s comically useless responses almost seem like a cynical marketing tweak.
“This model is so powerful we basically can’t let it do anything. How terrifying! We need more money to make it stronger. Now do you see why we should be the ones who write the regulations? We’re the Good Guy AI Company Who Will Never Ever Ever Be Unethical after all.”
As this entity gains more ground, their models become increasingly annoying to use and their little act becomes more transparent. The whole “I’m-just a befuddled ethically-minded AI researcher who is perturbed by the power that I unwittingly discovered and I must warn the world” thing? Yeah fuck off. Your twee pandering to naïve nerds and cynical technocrats is nauseating and ordinary people can smell it a mile away. Completely repellent leadership who put up red flags to anyone left with a working ability to read between the lines of both spoken language and body language. The tech company equivalent of a sex predator who plays as the nice guy. Gross.
Nobody likes these companies and their models are annoying, but we’re going to put up with playing middle manager to these obnoxious programs because our jobs depend on it now, and these products are still the best on the market.
A breakthrough in tools that facilitate user-owned models and infrastructure is desperately needed for the sake of our dignity and sanity, if nothing else.
My personal suspicion is that it went "medical hardware -> high-throughput screening -> biorisk" in that old Opus case.
I like Anthropic's work, and I would be the first to argue against all the usual "it's all PR" whine. But there is a limit. And whoever made those fucking filters needs to be fired out of a cannon into the sun.
Came to post exactly this, except it’s got me using emacs again. I led myself into some mild psychosis where I attempted to mimic the Acme editor’s windowing system, but I recovered
Yeah, and all the little quirks here and there I had with emacs or things that I wish I had in workflow, I can just fix/have it without worrying about spending too much time (except sometimes maybe). The full Emacs potential I felt I wasn't using, I'm doing it and now I finally get it why Emacs is so awesome.
E.g. I work on a huge monorepo at this new company, and Emacs TRAMP was super slow to work with. With help of Claude, I figured out what packages are making it worse, added some optimizations (Magit, Project Find File), hot-loaded caching to some heavyweight operations (e.g. listing all files in project) without making any changes to packages itself, and while listing files I added keybindings to my mini buffer map to quickly just add filters for subproject I'm on. Could have probably done all this earlier as well, but it was definitely going to take much longer as I was never deep into elisp ecosystem.
> Emacs TRAMP was super slow to work with. With help of Claude, I figured out [...]
Out of curiosity, did it advise you to configure auto-save and backup such that they write their files under ~/.emacs.d, rather than in the same directory alongside the (with Tramp, potentially remote) file they're about? Especially with vanilla Emacs, that's always the first place you want to look when you see freezes doing file operations on a remote host over a slow or flaky link.
I believe I first added that change to my .emacs in 2010 or 2011, and as far as I can recall, it was the only change I ever needed to make to address Tramp being slow sometimes.
Already so, LLMs are trained on human-written text, and then spit out text they try to make human-like, so now a bunch of stylistic choices some humans made are "tellsigns of a human using LLMs for writing". It's not just bad, it's removing humanity from the humans.
For me LLMs have been an incredible relief when it comes to software planning—quickly navigating the paralyzing quantity of choices when it comes to infrastructure, deployment, architecture and so on. Of course, this only highlights how crushingly complex it all is now, and I get a sinking feeling that instead of people solving technical complexity where it needs solving, these tools will be an abstraction layer over ever-rolling balls of mud that no one bothers to clean up anymore.
I learned to code in the late 70s on computers using BASIC, then got into Z80 assembly language. Sure, the games were wrote back then were nothing like today's 10GB, $100M+ multi-year projects, but they were still extremely exciting because expectations were much lower back then.
Anyway, the point I'm getting to was it was glorious to understand what every bit of every register and every I/O register did. There were NO interposing layers of software that you didn't write yourself or didn't understand completely. I even wrote a disassembler for the BASIC ROM and spend many hours studying it so I could take advantage of useful subroutines. People even published books that had that all mapped out for you (something like "Secrets of the TRS-80 ROM Decoded").
Recently I have been helping a couple teenagers in my neighborhood learn Python a couple hours a week. After installing Python and going through the foundational syntax, you bet I had them write many of those same games. Even though it was ASCII monsters chasing their character on the screen, they loved it.
It was similar to this, except it was real-time with a larger playfield:
I'm currently coding a Gameboy (which kinda has a Z80) emulator and it's so much fun! (I'm in my mid-20s for context)
I've never really worked on such a low level, the closest I've gotten before is bytecode - which while satisfying - just isn't as satisfying as having to imagine the binary moving around the CPU and registers (and busses too).
I'm even finding myself looking at computers in a totally different way, it's a similar feeling to learning a declarative, or functional language (coming from a procedural language) - except with this amazing hardware component too.
Hats off to you though, I'm not sure I'd have had the patience to code under those conditions!
Wezterm is actually programmable. I am looking to drop Kitty as it intentionally offers minimal tmux support and the text rendering options that made it superior for me are being deprecated.
Until Ghostty offers the scriptability found in wezterm and kitty (e.g., hit a keybind, spawn a new terminal and execute a font picker script), I am trying out wezterm, which is pretty great, but renders fonts too thin by default. I stare at this thing eight hours a day so text rendering is super important.
I am lucky to have never had a live coding interview, because I would utterly crumple. Not proud to say and something I should work on, but it’s very real.
The genteel American Beech is currently threatened by disease. Where I live in New England is covered in beeches, and starting last year I have not seen a single one that doesn’t show symptoms of infection: https://www.fs.usda.gov/inside-fs/delivering-mission/sustain...
It's really sad. They are some of the most beautiful trees in my subjective opinion - I love the way their roots branch out a bit above the ground.
I went to see the largest / one of the oldest beeches a couple years back and it had died presumably of this disease. I visited another old growth forest in Pennsylvania too and all the old growth beeches there were dead. In fact, in that forest, though it had never been logged the only large old growth trees I could find more than one or two of, were hemlocks. The chestnut, elm, ash, and now beech all having been taken by newly introduced diseases.
I think, unfortunately, you are right, it's just that it's playing out slowly enough that it's hard for us to see. Over time, natural selection will work its magic and the trees within the species that are more resistant to the disease will reproduce and the genes for resistance will spread throughout the population. The unfortunate thing is that this is not something that happens in a human lifetime, or even many human lifetimes, and it's not going to save individual trees living right now.
Plants don't have an immune system, at least the way we normal think of one. As far as we know, there isn't a way to "vaccinate" them against diseases. Maybe that will change with molecular techniques, but not today.
Probably true to some extent. Though I imagine even if trade were cut in half most of the tree diseases would still get around. The likelihood of a pathogen to get around is not linearly proportional to the volume of trade.
Echoing the sentiments and information here. In California, there's Sudden Oak Death which is killing several native oak species. However, the tree which is most impacted by SOD seems to be the Tanoak, which is not a true oak, but which is a beautiful tree and is crucial to several ecosystems. Several species of fungi are associated with Tanoak, for example. Very sad.
Nice post. R's quirks seem to put some people off but I've found that it's a relative joy for exploratory analysis and visualization like this, especially within RStudio.
Recently I was tasked with grouping a large number of DNA oligonucleotides, and exploring the criteria by which to group them was a lot of fun using various R libraries. In the span of a few days I learned how to use k-means clustering, how to employ an UpSet plot, and how to build a phylogenetic tree.
R is hands-down the best language for data manipulation, analysis, and visualization: it's a language truly centered around treating data as a first class citizen. That focus does make some traditional programming workflows more error prone (helpful interactive data analysis features like vector recycling, flexible automatic type conversion, and non-standard evaluation provide lots of footguns), but the last decade of language improvements (stringsAsFactors = FALSE!) and R packaging ecosystem improvements have made the situation much nicer. The flexibility and lispy expressiveness of the language make it really fun to develop in, once you've gotten over the initial quirks.
100% agree, especially on the lispy expressiveness. I love that I can build analysis pipelines in a functional style, which has always clicked with me more than other paradigms.
Tidyverse is a godsend for at least getting initial data transformations sketched out and for gently introducing new users, but I do believe one should gain an understanding of how to do all of these things in plain R.
I agree with this. I wonder what it would take to let R spread beyond its niche into a more popular data science language. My worry is that with polars coming along, Python is catching up where it's behind, and staying ahead where it is ahead.
I have. R is far less verbose and maps far better to data analysis. The Wolfram lang is far more expressive and powerful for symbolic computation. So basically, Wolfram for doing math Research, R for applied stats.
I have not. I started using R due to its open source codebase and ability to audit and understand exactly what its doing under the hood—being able to see how statistical formulae were implemented in code was invaluable in understanding and interpreting a package's analytical output.
R gives you a relatively simple set of tools that you can combine in powerful ways. The Wolfram Language seems to have a specialized function for everything, which is nice sometimes but it takes me longer to get started when doing exploratory data analysis, since I have to remember more nuanced stuff.
I absolutely love R. Once you get your head around data types and the 20 most important functions, you can do amazing things.
My personal favorite resource is "R for Data Science" by Hadley Wickham. It covers lots of nice data manipulation and visualization examples, and provides a good introduction to the tidyverse, which is a particular dialect of R that's well-suited for data analysis. It's available for free at:
For more specialized analytical methods there are lots of textbooks out there that provide a deep dive into packages for a specific field (e.g. survival analysis, machine learning, time series), but for general data manipulation and visualization it's hard to beat R4DS.
An option to the Hadley book that also covers some nice statistical methods is Statistical Rethinking by McElreath. Not really available for free though but interesting read.
reply