What's going on in #bioinformatics? Any new cool alignment or mapping algos? Any advances in kmer hashing?
What's going on in #bioinformatics? Any new cool alignment or mapping algos? Any advances in kmer hashing?
When a candidate's CV matches the job description too well
No updates from my side today
No updates from my side today.
look at github.com/PacificBiosc... to get a better idea. We are not longer pushing to GH, but it's still in a good shape
I've been unhappy with my Sonos equipment. Last night, I've decided to take things into my own hands and created a local sonos controller in the browser with LLM help. Once I grok the full API, final frontier will be an Android app. Everything is 100% open source github.com/armintoepfer...
Back in town, this time staying a few nights in SF. Good to be back and looking forward to meeting y'all @pacbio.bsky.social
2025 mood
Testing out #cursor AI, anyone got great tips and tricks?
Anyone got experience with code review bots? Are they any good? #AI
Yeah, but those are tricks to make mm2 faster. I'm thinking of a completely new approach that might be 10-100x faster.
It needs to span larger SVs and handle Q10-Q40 reads.
I don't have the answer.
What I have is a superior ksw2 GPU implementation, but it's closed source.
I wonder if someone will sit down and write a better map-and-align tool than minimap2. If yes, will it (I) still support crappy 90% accuracy long reads or not and (II) offer interoperability being written in C or C++?
North German breakfast
Advent of code, hardcore edition
Day A:
Write fizz buzz in Haskell. You need to create a new DSL. You can only use letters of {A,C,G,T}. The code must be stored in DNA. You need to sequence and assemble it. Maximum coverage of 3-fold. Implicit error correction is allowed.
Hint:
Use HiFi.
There's no traffic jam on the extra mile
Things that happen if you spawn uint -1 threads on MacOS. The system instantly freezes and reboots after like 10 seconds.
Performance review time of the year
Trying out the new QwQ to rewrite gepard:
Step 12: Think about parallelization for speed.
- Given the large size of the sequence, parallel processing will be essential.
- C++ allows for multi-threading, but for更大规模的并行,可以考虑使用GPGPU CUDA。
- 如果能够将计算分解为可以在GPU上并行执行的小任务,那么CUDA将是一个很好的选择。
...yeah
Literally one minute.
Just changed my handle to @armintoepfer.com pretty sweet feature
I'm not used to having a normal work day from 8-4. Thanksgiving break shows me the real European work life balance that I get once per year.
Usually I have 3-5 meetings between 5-9pm, 10 ppl pinging me on slack, verification churn, and chain of command asking me for updates.
AI will help the ambitious to become more productive and successful.
But you need to have a solid background to innovate, hands-on experience to make things really fast, must have suffered to grow, and be able to articulate and explain a hard problem.
I need you to build the nexgen #PacBio. Apply
Tis the season, baking with the kids
What a banger of a disc print.
Tired of chasing streaming providers, I just got this gem, forever.
I made fried rice. Uncle Roger would approve.
True leadership recognizes the unseen effort that prevents chaos — humility begins with appreciating the fires that never had to be fought.
So, #PacBio made the jump. Where are the UK folks? Where are our friends that can only read stories?
PacBio jasmine methylation caller logo
The upcoming SPRQ chemistry for #PacBio #Revio will be accompanied with a software update to improve the on-instrument 5mC methylation performance and introduce 6mA calling 🎉. Every HiFi read comes with both signals – for free. This will also enable Fiber-seq. Talk to your @pacbio.bsky.social rep!