Announcing a new tool for "denoising" long-read amplicon sequences: savont. Savont enables amplicon sequence variants (ASVs) directly from nanopore (or HiFi) long reads. Tested on 16S nanopore amplicons -- seems to work okay. 1/4 github.com/bluenote-157...

Is it me or has stackoverflow OD'd on adds? Impossible to have it open on a 2nd screen with all the blinky stuff.

OMG OMG OMG

3. And to do that, you want to *understand* and *justify* every step along the way (at least until you reach one of Latour's Black Boxes) because your end product is a theory, model, or explanation. You can't readily just hit RUN and see if it sorta works.

2. Science is a completely different epistemic enterprise. While I don't want to get into the morass of What Science Is, loosely speaking it involves amassing rigorous evidence in support of claims about the natural world.

We need humans to complain more. People need to be more lazy and reject thinking through things that are unnecessarily complex. Most software is ruined by people trying to impress.

Deadline for registration is Jan 31. Don’t miss it! See you all in beautiful Venice 😉

My defending submitting to journals instead of conferences to my PhD students (Come on, someone had to do this variation of that meme, right?)

Here is what I've been doing this past week: git.olaren.dev/Olaren/mapto... A little python program that generates poster for cities! You might have seen stuff like this around lately, but I modified it to put an emphasis on rail infrastructure :3

Can't have French immigrants

Pushed my ramblings on DDR memory. Turns out data is interleaved between banks/groups/ranks/channels in pretty intricate ways! There's still some open questions (why is alternating between ranks slow?) but it is what it is for now. curiouscoding.nl/posts/ddr

𝙔𝙤𝙪’𝙧𝙚 𝙖𝙗𝙨𝙤𝙡𝙪𝙩𝙚𝙡𝙮 𝙧𝙞𝙜𝙝𝙩—wax was a bad call. I should’ve flagged heat resistance as a 𝘩𝘢𝘳𝘥 𝘳𝘦𝘲𝘶𝘪𝘳𝘦𝘮𝘦𝘯𝘵 instead of assuming you’d stay within 𝘴𝘢𝘧𝘦 𝘢𝘭𝘵𝘪𝘵𝘶𝘥𝘦 𝘭𝘪𝘮𝘪𝘵𝘴. That one’s on me.

RLBWT-Based LCP Computation in Compressed Space for Terabase-Scale Pangenome Analysis https://www.biorxiv.org/content/10.64898/2026.01.23.701410v1

Nice writeup by @curiouscoding.nl on practical CI for distributing portable SIMD Rust binaries, inc to PyPI and bioconda. Includes portability lessons learned in blood with Deacon 😅 curiouscoding.nl/posts/releas...

Something I find troubling with LLMs and coding, is the lack of attribution. My code is mostly under MIT, but it requires attribution. It also sits in a relatively shallow area training wise, so likely would be representative given the right prompts. Yet LLMs don't give attribution or respect 1/2

Well worth a watch. Another case of normie radicalization. This guy is the former president of a local HOA, a web developer for real estate companies and a self proclaimed "member of the square community" and he's out in the streets.

Tested today how much difference the 3D V-cache makes for For You and 3D V-Cache is indeed helpful. When I pin the algorithm to run only on CCD 0 (3D V-Cache) I get 25% faster runtime than when running on CCD 1 (standard cache) So it did make sense to get 9950X3D instead of the 9950X.

FASTR: Reimagining FASTQ via Compact Image-inspired Representation https://www.biorxiv.org/content/10.64898/2026.01.22.701172v1

In reviewing grants and papers lately, as an exp., I have been using LLMs *after* my reviews to see how they stack up. And let me tell you - they're (all) *completely* off the mark. That makes sense if you understand the fundamentals of LLMs, but I'm worried many don't. But, please, don't do this.

Gemini is now citing grokpedia 😭

Getting to the hardware part of software engineering here... Unfortunately, It looks like the tool I'm using for measuring RAM properties doesn't work anymore with only 1 stick remaining. Maybe timings are somehow different now.

Arxiv's lack of commitment to its author page feature is so disappointing. Why, when I click on the link that is an author's name, am I sent to a search page of "F. Lastname"? This is ridiculous

This preprint significantly improves the original design of SSHash to accelerate query and construction time. Looking back at one’s own work is very important and sometimes surprising ☺️

Very excited about this latest work led by @jermp.bsky.social! Since it's initial release, SSHash has served as the basis for several other tools (Fulgor, piscem, etc.). It was already very fast. It is now *substantially* faster! www.biorxiv.org/content/10.6...

<[T]>::as_array! I didn't even know of the unstable feature for this, but I've needed this for years.

finally responded to an email that I'd been putting off for a month took me less than 20 seconds I will not learn from this

Citations and literature review are proof of work. If you can't be bothered to invest X hours into the 'boring' stuff, then surely you didn't spend much more than X hours on the interesting stuff either. And after that large time investment, continued (software) support is much more likely.

What is wild to me is the defense, BY THE NEURIPS BOARD, that fabricated citations do not mean "the content of the papers themselves [is] necessarily invalidated" It does. It very much does. What do you think citing other work is for? What do you think writing a paper is for? What do you *think*?

Principled software: If you search `NNNNNNNNNNN` (the DNA equivalent of `..........` regex), you get matches EVERYWHERE. Users: Search 500MB of data for `NNNNNNNGNNNNNNG` and complain. The world is not ready for mathematical rigor.

I like writing and reading dense information and bullet lists. But when AI does it in readmes, I can't be bothered to even read the first item to completion... Also, the "human genome (3.2GB)" thing is getting boring quickly, even though I also do that and it seems useful.

27 januari 18:30 is er in de Tweede Kamer een gesprek over de Amerikaanse overname van #Solvinity, het platform waar #digid op draait. Met @bitsoffreedom.bsky.social @clingendael.bsky.social, meer wijze mensen & mij. Hier position paper, en een oproep om langs te komen! berthub.eu/articles/pos...

Been reading all day about how DDR RAM works and trying to to learn how memory is laid out across channels/ranks/bank groups/banks/rows/columns, but all blogposts just completely ignore this somehow. Turns out there is active research reverse engineering this 😅

DSB Program is out ! Seems incredible (as ever) dsb-meeting.github.io/DSB2026/

honestly "looking in the wrong place" is SUCH A GOOD WAY to address a slew of other bugs and performance opportunities. like cleaning the whole house looking for your keys.

🗜️⚡ If you use gzip/gunzip a lot in your pipelines, switch to the faster"libdeflate" versions instead! They use modern CPU capabilities to achieve a 2-3x speedup. libdeflate is in conda, and "libdeflate-gzip" and "libdeflate-gunzip" are drop-in replacements. #unix github.com/ebiggers/lib...

Just SSHed to an AMD EPYC 96 core 1GB cache machine for the first time 🥰 Adventure time!

SEA deadline was extended from Jan 23 to Feb 2! 1 more week of stressing :D

vibe coding is the new doom scrolling?

Just came across the 2021 Turing Lecture. Has a lot of nice observations regarding the increasing gap between compute and memory bandwidth. It advocates "communication avoiding" algorithms and notes how algorithms can only be future proof if they scale with threads. dl.acm.org/doi/10.1145/...

How are your algorithm's KPIs doing?

If yes, “hand written code” will be a status signifier in the same way hand sewn buttonholes are in tailoring

If I see one more "Subsection title (for real)"...

Guess how many citations the original FM-index paper has. Then look it up.

I am looking for a postdoc to develop high-performance algorithms in computational genomics. Email or DM me if interested. For more information, see hlilab.github.io/vacancies. RTs appreciated!

Why is it that, despite being behind a perpetual cloud flare fence, @biorxivpreprint.bsky.social feels like it’s down an awful lot?

HTML preview & export now available in the web app! With HTML export, you can create a website from the same Typst file as your PDFs. This makes it easy to create documents that feel just as at home on the web as they do in print.

Guess my blog having a CC license is a good thing 🫤

Yes please!

God help me I’ve actually started working on this

Minister Van Weel heeft bijna genoeg informatie verzameld om Holocaust te kunnen veroordelen