A CRISPR off-target search cannot report a sequence it was never given. @ameynert.bsky.social explains how DivRef brings common human variants and haplotypes into the search space—and how her rebuild complements @timd.one work on Sassy. Read Alison’s post: blog.fulcrumgenomics.com/p/a-crispr-o...

A CRISPR off-target search is only as good as the sequences it searches

Rebuilding DivRef so CRISPR off-target searches can account for human variation

blog.fulcrumgenomics.com

whether you AI-love or AI-hate I think all of you will enjoy a little something about this story:

amos@fasterthanli.me · 2h ago

ran out of claude/chatgpt quota. thought: why not try qwen3.8-max, I hear good. have it operate on a machine I use a lot, jax, which is also the router for the entire house. it finds /etc is tracked by etckeeper but has no remote. "okay push it to this private repo", I say. qwen thinks, wait...

It is simultaneously true that 1) recent advances in LLMs made real progress in solving open problems, and they are a step change better and 2) a lot of the problems we ask llms about are not major open problems. LLMs will give math solutions without appropriate attribution

CRISPR off-target search can find the right locus while hiding plausible alignments within it. I wrote about how Sassy v0.2.5 uses recursive backtracking to report every reasonable alignment—more than 9× as many in under 30 seconds. blog.fulcrumgenomics.com/p/why-crispr...

Why CRISPR Off-Target Search Should Report Multiple Alignments Per Locus

How Sassy enumerates every reasonable alignment without sacrificing runtime

blog.fulcrumgenomics.com

Travis has been cooking something up since IGGSy ;) A fast algorithm for SMEM finding in haplotype panels that uses space proportional to the number of 1s in the (sparse) matrix, inspired by the jump index [13].

arXiv cs.DS Data Structures and Algorithms@csds-bot.bsky.social · 2d ago

Paola Bonizzoni, Davide Cozzi, Travis Gagie, Younan Gao, Ragnar Groot Koerkamp: MEM-finding with run-length compressed strings https://arxiv.org/abs/2608.01387 https://arxiv.org/pdf/2608.01387 https://arxiv.org/html/2608.01387

I am honestly very tired of doing computer homeworks and this feels like "instead of a complete functional thing you will get Homework, which luckily a machine might solve for you if you harness and prompt it right (meta-homework!)

#rustlang has an official LLM policy. Pretty sensible - everything boils down to "you can't copy paste LLM output without any work and call it a contribution". QC requirements are higher for LLM generated code, also common sense. blog.rust-lang.org/inside-rust/... Also this part cracked me up:

LLM in coding discussions in public sphere is just completely broken. Don't ship things you couldn't build yourself. One would think this wouldn't be a controversial take, but apparently it is.

i'm really proud of this policy. it's not perfect, but i think it's close to the best that was possible given all the constraints. i'm excited that this will make things easier for mods and reviewers, and hopeful that it will spark change in Rust's governance. blog.rust-lang.org/inside-rust/...

rust-lang/rust is adopting an LLM policy | Inside Rust Blog

Want to follow along with Rust development? Curious how you might get involved? Take a look!

blog.rust-lang.org

it was already a close race, but now it’s a dead heat OpenAI announces two more accidental felonies openai.com/index/third-...

mr. TIM@timkellogg.me · 6d ago

Anthropic winning as usual

the information density of gerrit's ui is one of its biggest strengths. yes, it's intimidating the first few times you look at it, but so is a qwerty keyboard

Hype! blog.rust-lang.org/2026/08/04/e...

Enabling the next iteration of the borrow checker on nightly | Rust Blog

Empowering everyone to build reliable and efficient software.

blog.rust-lang.org

LLMs adopts linguistic quirk -> Humans avoid linguistic quirk -> Humans develop new linguistic quirks -> LLMs adopt linguistic quirk... LLMs can update a lot faster though, and so I wonder if this all does something quite weird to written english.

New postdoc opportunity in our team! Topic is computational analysis of (meta)genomic data to study evolution of plant-microbiome interactions. Part of the AEGIS consortium (aegisearth.bio/en). Apply here: www.wur.nl/en/vacancy/p... Please RT!

Postdoc on computational (meta)genomic analysis

Climate change is one of the greatest challenges for global food security, requiring new approaches to understand and harness the natural diversity that enables plants and their associated microbiomes...

wur.nl

It’s hot across Europe right now. Millions of us are feeling it - in cities, towns, and villages across the continent. When heatwaves hit, small actions matter ⬇️

De snelheid waarmee een Europese crisistop 'migratie' wordt georganiseerd is een beetje een contrast met die van 'droogte, branden, klimaat,...'. Het symboliseert politici die denken aan verkiezingen in plaats van generaties.

I do wonder if this is like the forest fire that needs to wipe everything out so life may flourish once more but for software

one very weird downside to spending weeks at a time working on a series of episodes on a single topic are the content gobbling crawlers that make shoddy AI transcriptions of podcasts. if I google something too niche the top results are just AI generated summaries of what I wrote last week.

“have we tried writing the requirements first?” No? Oh ok.

This thought expt (on how to find maths problems that openAI can solve) is interesting tjoresearchnotes.wordpress.com/2026/08/02/r... a) all of their proofs fit into one context window, b) scan open problems looking for addressable ones, 3) reported cost is only the compute cost for the successes

This is some really cool stuff! A chaining dynamic hashtable that uses *less* space than the input (by using quotienting), that actually seems to have quite competitive performance 🤯 And also a faster flat variant has 1.25 cache misses on average (and at most 2 or 3), that has super fast queries.

arXiv cs.DS Data Structures and Algorithms@csds-bot.bsky.social · 3d ago

Xilin Tang (Cornell University), Yuqi Mai (Cornell University), William Kuszmaul (Carnegie Mellon University), Alex Conway (Cornell Tech): Succinct and Fast Tiny Pointer Hash Tables https://arxiv.org/abs/2607.28892 https://arxiv.org/pdf/2607.28892 https://arxiv.org/html/2607.28892

Thank you so much Springer for informing me that my review is no longer required and closing the portal even though the deadline is in 3 days. I was just in the process of finalizing my comments but I guess it goes to the trash now 😠

One of the easiest ways to stand out when writing or creating videos or posting: DONT. USE. AI. Suddenly you sound distinctively different from 80%+ ppl who do and output the same verbose + nicely formatted stuff that feels all the same (+ most ppl stop paying attention midway)

"I am basically reviewing slop papers at gunpoint. It feels like living in some weird level of hell."

Anirvan Chatterjee@chatterjee.net · 4d ago

"Five of these had hallucinated authors and/or entire papers in their bibliography…One of the NeurIPS papers had so much hallucinated/nonsensical jargon (53 pages of it) that there wasn’t any point in going through the bibliography."

Bwah.. every other paper now contains claude-speak Would *you* call a paragraph 'The X problem'? Would *you* write 'clean path to a proof'? Would *you* add random overly detailed parentheticals to the abstract? It just triggers me so much.

lmao this is brilliant, a page about cursed knowledge gained from building immich immich.app/cursed-knowl...

Cursed Knowledge | Immich

Cursed knowledge we have learned as a result of building Immich that we wish we never knew.

immich.app

Claude also does a weird comment style that's basically "narrating my edits" that only care about the context of the immediate task. Hate it.

I hope cities like #Zürich who are just starting to build bikeways follow Copenhagen's lead and plan bike spaces scaled for future demand. This one gives 2/3 of the space to people riding and walking and 1/3 to drivers. There's 20 people biking or walking for each car on this bridge.

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 - 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

github.com

Dear reviewer, please review my paper. Files attached. - paper.org - fig1.svg - table2.csv No? Then why do I get your .docx, .png, and .xlsx?