iiinova — where your digital self keeps evolving

Reposted by Tim Fennell

Nils Homer @nilshomer.com · 05/10/2026

bwa-mem3 is now 6-8x faster than the original, and now as fast as minibwa. What's stopping you from using it?

static.klipy.com

Zoom Zoom Mazda Commercial

ALT: Zoom Zoom Mazda Commercial

051

Tim Fennell @tfenne.bsky.social · 28/09/2026

I'm winding up to generate a mostly large joint callset (about 10k samples) through GATK's HaplotypeCaller to make gVCFs, and then on into GenomicsDB. I'm curious - are folks still actively using HaplotypeCaller / GATK for germline joint genotyping? Or are folks moving on to something else?

000

Tim Fennell @tfenne.bsky.social · 28/09/2026

New release of chelae over the weekend. The fastest adapter and qual trimming software you've probably never heard of 😉 Still measures the fastest of the bunch, though AdapterRemoval v3 isn't too far behind. Now with support for streaming interleaved FASTQ! 🧬 🖥️ github.com/fulcrumgenom...

github.com

GitHub - fulcrumgenomics/chelae: Fast, highly accurate, read-trimming for NGS data.

Fast, highly accurate, read-trimming for NGS data. - fulcrumgenomics/chelae

030

Tim Fennell @tfenne.bsky.social · 24/09/2026

And if you use it for the off-label purpose of filtering out reads from not-your-organism, also check out k2tools for faster fastq filtering: github.com/fulcrumgenom...

000

Tim Fennell @tfenne.bsky.social · 24/09/2026

Do you use kraken2? If so, don't skip the new release yesterday, now available on bioconda: github.com/DerrickWood/... Includes a PR from me that speeds up both database load (up to 3X) and classification (24-39% improvement)!

github.com

Release v2.17.2 · DerrickWood/kraken2

Minor release to fix bugs with k2 wrapper, speeds up merging using multi-threading. We have another release pending that will see significant speed ups to classification. What's Changed Multi-data...

100

Reposted by Tim Fennell

Nils Homer @nilshomer.com · 14/09/2026

bwa-mem3 v0.12.0 is out 🧬 Since v0.10.0: faster on both Arm & x86 (--fast is now ~2× minibwa and stock is within ~15% on x86) while staying a byte-identical (--compat). It also brings ~30–40% faster methylation and a big memory-safety pass. github.com/fg-labs/bwa-... #bioinformatics #genomics

github.com

Release v0.12.0 · fg-labs/bwa-mem3

0.12.0 — faster on both architectures, a faster --meth, and a large safety pass A broad optimization release that speeds up both Arm and x86 (unlike 0.11.0, which concentrated on Arm), plus a methy...

0132

Tim Fennell @tfenne.bsky.social · 06/09/2026

Also .... how is it that your bsky tagline doesn't include necropoet???

000

Tim Fennell @tfenne.bsky.social · 06/09/2026

I think she enjoyed “Rescue” best.

110

Reposted by Tim Fennell

Ashland Poetry Press @ashlandpoetrypress.bsky.social · 01/09/2026

@ktlandon.bsky.social 's debut collection, Abide, launches today! 2025 Richard Snyder Prize Winner, selected by @kimaddonizio.bsky.social Abide is available at @bookmobile-itasca.bsky.social , Amazon, & elsewhere now! #bookrelease #poetry #poetrycommunity #newbook #itasca #buyabook #poems

073

Reposted by Tim Fennell

Jess Smith @nanojess.io · 31/08/2026

Every time I read a paper like this I just think about what a PITA it's going to be to design off-target NGS assays that are suitable for regulatory review.

011

Tim Fennell @tfenne.bsky.social · 26/08/2026

I think this is actually a first - I'm unaware of any streaming, query-grouped duplicate marking tool that estimates sequencing dupes and gives an adjusted library size estimate.

100

Tim Fennell @tfenne.bsky.social · 26/08/2026

Sequencing duplicates are detected by writing a small amount per read-pair (~16 bytes) to a temp file - costs nearly no runtime. After output is closed (unblocking downstream), the temp file is read and the rate of sequencing dupes calculated. Adds maybe 2-3% to CPU used during the active phase.

100

Tim Fennell @tfenne.bsky.social · 26/08/2026

dupblaster 0.3.0 is released! The new release adds: - optional tmp file compression with zstd - sequencing (optical) vs. library duplicate detection - incremental performance improvements On top of existing: - fastest and most accurate query-grouped duplicate marking - comprehensive QC stats 🧬🖥️

130

Reposted by Tim Fennell

Nils Homer @nilshomer.com · 25/08/2026

fgumi v0.7.0 is released: github.com/fulcrumgenom... 1. sort is 28% faster; 2-3x faster than samtools 2. improved CODEC consensus calling 3. dedup now outputs metrics closer to picard MarkDuplicates and dupblaster 4. retag is a new tool that can copy/move/delete SAM tags

github.com

Release v0.7.0 · fulcrumgenomics/fgumi

For those running the fgumi command line tools v0.7.0 makes sort faster, hardens CODEC consensus, and brings dedup metrics to Picard/dupblaster parity. Sort is up to 28% faster — and 2–3× faster th...

183

Reposted by Tim Fennell

Fulcrum Genomics @fulcrumgenomics.com · 24/08/2026

A good case for ensuring scientific software is safe. @science-goblin.bsky.social shares what the pixy project reinforced about building code scientists can trust, maintain, and keep extending 👇

031

Tim Fennell @tfenne.bsky.social · 24/08/2026

Did whoever coined "dovetailed" as the term for an NGS read-pair where the 3' ends extend past their mate's 5'-ends have *any* knowledge of woodworking? In a field of terrible names, this one wins for me because it invokes the exact opposite the originators' intent/usage.

010

Tim Fennell @tfenne.bsky.social · 21/08/2026

You mean "Delete and Report Spam" season?

100

Tim Fennell @tfenne.bsky.social · 05/08/2026

Awesome forensics work by @ameynert.bsky.social to figure out all the issues and build a more robust pipeline and more complete and accurate DivRef!

010

Tim Fennell @tfenne.bsky.social · 01/08/2026

So if you're writing rust and want to move I/O off the main thread, check it out!

020

Tim Fennell @tfenne.bsky.social · 01/08/2026

Along the way I* learned about miri (github.com/rust-lang/mi...), and loom (github.com/tokio-rs/loom), both super-useful tools for hardening a library like this. rawb-io now uses both in CI, has far better testing, and provably no possible deadlocks or data races!

210

Tim Fennell @tfenne.bsky.social · 01/08/2026

Maybe three weeks ago, I* refactored some code out of dupblaster and methylsieve for using dedicated threads with large buffers for read-ahead and write-behind. Why? Because it turns out, even in Rust, it's still hard to write threading code that "just works". github.com/tfenne/rawb-io

github.com

GitHub - tfenne/rawb-io: Read-ahead / write-behind byte IO: threaded Reader/Writer that decouple a pipeline stage from the kernel pipe via a background thread and a byte ring buffer

Read-ahead / write-behind byte IO: threaded Reader/Writer that decouple a pipeline stage from the kernel pipe via a background thread and a byte ring buffer - tfenne/rawb-io

242

Tim Fennell @tfenne.bsky.social · 01/08/2026

Woah, 11 years since last commit? Also, "Opal (ex Swimd)" - the old name is so much better! I really hope your rust port is a play on "swim[d]ers".

100

Tim Fennell @tfenne.bsky.social · 01/08/2026

I have not - but I did log a separate issue a few days ago. I’m not actually looking to rewrite it, just trying to decide if I should even include it in a benchmark. Same question, different purpose.

010

Tim Fennell @tfenne.bsky.social · 01/08/2026

Ooh, ooh, can I get a ruling on one? I got an issue request in dupblaster, to add FastDup to the benchmark. v1.0 is 10 months old, only commit since then is to add paper citation. Issue open since January that it segfaults(!) if there are no RGs in the header. (see github.com/zzhofict/Fas...)

github.com

GitHub - zzhofict/FastDup: Identifies duplicate reads. This tool locates and tags duplicate reads in a coordinate ordered SAM or BAM file.

Identifies duplicate reads. This tool locates and tags duplicate reads in a coordinate ordered SAM or BAM file. - zzhofict/FastDup

310

Tim Fennell @tfenne.bsky.social · 23/07/2026

I’m curious, is this just different SIMD giving different reduction paths, causing floating point drift?

110

Reposted by Tim Fennell

Nils Homer @nilshomer.com · 23/07/2026

bwa-mem3 v0.7.0 is out. 5–24% faster than v0.6.0, output byte-identical; upgrade and it's free. If you do methylation: reworked bisulfite path with TAPS support too. github.com/fg-labs/bwa-mem3/releases/tag/v0.7.0

github.com

Release v0.7.0 · fg-labs/bwa-mem3

0.7.0 — methylation accuracy, a sharper --fast, and broad speedups This is primarily a methylation release. The bisulfite path was reworked to a chemistry-aware contract — TAPS support, a NEUTRAL -...

151

Tim Fennell @tfenne.bsky.social · 22/07/2026

Thanks again to @curiouscoding.nl et al for developing sassy in the first place, and for collaborating with us on this addition!

020

Tim Fennell @tfenne.bsky.social · 17/07/2026

I mean .. this also applies to humans. We just can't drop our guard when talking to an AI either. If I believed everything every human at work ever told me, with no skepticism, things would look really different.

220

Tim Fennell @tfenne.bsky.social · 17/07/2026

Oh man. 7am is early? 7am is second cup of coffee territory.

110

Reposted by Tim Fennell

Nils Homer @nilshomer.com · 14/07/2026

unum: a pure-Rust HLA/KIR genotyper It's a port of github.com/mourisl/T1K, offering significant speedups, ergonomics, and opinionated improvements. I am looking for folks who want to give it a try and give constructive feedback. github.com/fg-labs/unum

github.com

GitHub - fg-labs/unum: Fulcrum-owned Rust HLA/KIR genotyper (strangler port of T1K)

Fulcrum-owned Rust HLA/KIR genotyper (strangler port of T1K) - fg-labs/unum

192

Tim Fennell @tfenne.bsky.social · 14/07/2026

I see you have fastp in there ... chelae (github.com/fulcrumgenom...) is quite a bit faster, and in my hands, also more accurate fwiw.

110

Tim Fennell @tfenne.bsky.social · 14/07/2026

Oh, I don't know. Probably?

020

Reposted by Tim Fennell

yaoi gagarin @mel.bzky.team · 12/07/2026

well okay so the english won against the norwegians but in only a few days theyll have to face the french. anyway thats quite enough about September 25th 1066, lets look at some football news

111312332

Reposted by Tim Fennell

Andrew Whitwham @whitwham.bsky.social · 09/07/2026

Release 1.24 of HTSlib, SAMtools, and BCFtools is now available from GitHub. See htslib.org/download/ for links to tarballs and release notes. 🧪🧬🖥️

htslib.org

Samtools

074

Reposted by Tim Fennell

Nils Homer @nilshomer.com · 09/07/2026

🚀 ferro-hgvs 0.7.0 is out: our biggest release yet. Major upgrades to HGVS normalization, parsing & projection: spec-compliant 3′ shifting, mosaic/compound alleles, multi-axis projection (g/c/n/p/r), Ensembl support, plus ~1.7× faster parsing. github.com/fulcrumgenom...

github.com

Release v0.7.0 · fulcrumgenomics/ferro-hgvs

Added (reference) validate manifest schema/version at load, fail loud on an incompatible reference (#1003) (mosaic) parse predicted-wrapper and whole-entity-LHS =/ forms (#992) (protein) parse ins...

131

Tim Fennell @tfenne.bsky.social · 09/07/2026

The main use case right now, is not static fwiw. It's a Rust tool that builds off samblaster, which packs paired-end position information into u64s (contig/pos/strand x 2) for streaming duplicate marking. So the sets ends up being hundreds of millions deep in total.

110

Tim Fennell @tfenne.bsky.social · 09/07/2026

This is amazing! It was quite literally on my list to go searching for a better/faster HashMap/HashSet for u64 keys today! Should I try it out yet, or wait for a 2.x release to crates.io?

110

Tim Fennell @tfenne.bsky.social · 09/07/2026

Lol, it's true. When opening PRs I now instruct Claude "write a concise/terse PR description to a MarkDown file so I can edit it before opening the PR" for this exact reason.

110

Tim Fennell @tfenne.bsky.social · 09/07/2026

So TL;DR is fable for sub-agents is even worse - silently falls back to Opus with nearly no trace. CC @nilshomer.com and @robp.bsky.social.

142

Tim Fennell @tfenne.bsky.social · 09/07/2026

Opus found our questions about aligners and "teach me basic infosec" *so* benign it concluded the only reasonable answer was that the sub-agent system did not support fable at all. We then asked some truly benign questions, and Opus was shocked:

Innocuous questions truly route to fable; anything even vaguely bio or infosec gets silently downgraded to Opus.

111

Tim Fennell @tfenne.bsky.social · 09/07/2026

The switcth to Opus happened with no notification to the parent (Opus), nothing in the output that would indicate it happened. And happened even though I have Claude Code configured to STOP instead of auto-switch to Opus when the safeguards are hit 😱

111

Tim Fennell @tfenne.bsky.social · 09/07/2026

Opus and I did a little experimenting and some forensics, and determined we could figure out what model a sub-agent used to render a response:

What model did a sub-agent use to generate a response?

200

Tim Fennell @tfenne.bsky.social · 09/07/2026

More fun with fable yesterday. A coworker thought they'd cracked the code by asking opus to pose questions to a fable sub-agent in Claude Code. Dear reader, they had not.

151

Tim Fennell @tfenne.bsky.social · 09/07/2026

Thanks for getting release 0.4.0 out the door and all the additional fixes. Can't wait to pick it up in my projects. 😍

010

Tim Fennell @tfenne.bsky.social · 09/07/2026

Most of my commit messages are authored by Claude, and are now far more detailed and useful than anything I used to write by hand. And I too rarely read commit messages directly - but I do frequently ask Claude to review history and summarize for me, where detail helps over fixed structure.

110

Reposted by Tim Fennell

Randall Munroe @xkcd.com · 08/07/2026

Offside xkcd.com/3268/

Comic. [chessboard with red dotted line and flag at second line. Highlighted rook is positioned next to King behind the red line.] [caption] I thought I had checkmate, but I didn’t realize my rook was offside.

464601816