Ola Wallerman🇸🇪

4.1K posts

Ola Wallerman🇸🇪 banner
Ola Wallerman🇸🇪

Ola Wallerman🇸🇪

@OWallerman

PhD in genomics, expert in nanopore sequencing and bioinformatics

Uppsala, Sverige Katılım Ocak 2015
523 Takip Edilen898 Takipçiler
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
@nomad421 I use Go all the time now. "If you want me to I’d do these three things next:" Ok, Go!
English
0
0
2
32
James Ferguson
James Ferguson@Psy_Fer_·
@OWallerman but like, how did you get all 3 on the terminal? I don't remember implementing multifigure plots for terminal output (though, it can just take any plot and render it in the terminal, I don't remember implementing that on for the CLI cause it was too complicated)
English
1
0
0
22
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
@Psy_Fer_ Told Gemini 3 Flash to look at the library, figure out which plots to use and tell me how to adjust line height to make it look nicer :)
Ola Wallerman🇸🇪 tweet media
English
1
0
1
32
Raf
Raf@RfSharko·
@OfficialLoganK Look the be honest, after testing Gemini3.1 it's not even close to Claude Sonnet and Opuse 4.6.
English
7
0
112
12.9K
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
Gemini 3 pro did at least give meaningful thoughts. 3.1 is almost unusable in Antigravity
English
0
0
0
32
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
I'm now moving forward to the subsequent stage, and considering how to best make progress. The next step is in motion. Now I am moving forward with the execution. I am considering the next action and how to move forward. SAME% SAME%
English
1
0
0
45
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
@nomad421 Not my experience but might be because I only used Claude in Antigravity, guess it is time to get a Claude subscription. Not too impressed with Gemini 3.1 so far...
English
0
0
0
50
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
@nomad421 Opus talks nice but I threw 5.3 at one of my larger projects and this pretty much sums it up...
Ola Wallerman🇸🇪 tweet media
English
1
0
0
40
𝕐
𝕐@nomad421·
Claude Opus 4.6 is an absolute beast at Rust. You still have to watch it carefully to ensure the implementation, and not just the output, are what you intend. However… wow!
English
2
0
0
326
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
@nomad421 @AnthropicAI Even better - ask it to build a bioinformatics framework in rust for agentic programming. No more slow python tools. For ⭐️ - can send some stats if you have a good testset
English
0
0
1
170
𝕐
𝕐@nomad421·
@OWallerman Well, now it has to be built! It has a logo!
English
1
0
0
18
James Ferguson
James Ferguson@Psy_Fer_·
@kirk3gaard @ATinyGreenCell @The__Taybor Every hospital I've sold a computer to for runing a P2S was so they could get hands on with the tech, test some pilots, so they could move to clinical testing. That stuff takes time to happen, and it pipelines into P24/48 sales. Removing low price option to try it is a mistake.
English
1
3
10
2.1K
Clive G. Brown
Clive G. Brown@The__Taybor·
As the person who conceived both devices and led their development- as I stated at London Calling in public - P2S is targeting pioneering users who want flex. P2i is targeting more conservative users who want reliability- so sell appropriately. genomeweb.com/sequencing/oxf…
English
6
3
21
2K
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
@vsbuffalo What do you think will be the biggest win for the field? Will we see faster code for general purpose tasks or "just" more specialized tools?
English
0
0
0
22
Vince Buffalo
Vince Buffalo@vsbuffalo·
Don’t get me wrong. This frontier is great in many ways. People who are better designers than coders will create amazing things. Testing is cheaper! Eventually better software will emerge as a result. The upper frontier: we can use AI to lower the barrier to excellence.
English
2
0
0
201
Vince Buffalo
Vince Buffalo@vsbuffalo·
IMO there are two frontiers of AI & coding. The bottom frontier: barrier to entry is now close to zero, software is easy to vibe code. Abundance, with slop: we’re going to see bad bugs; risk of scientific papers invalidated when AI inserts fake data or bunk code. New tail risks.
English
1
0
1
388
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
Should be M pairs/h, not reads. Star is absurdly fast, wonder why it is not used for DNA. It is time to kill fastQ
English
0
0
0
45
Ola Wallerman🇸🇪
Ola Wallerman🇸🇪@OWallerman·
@nomad421 Ok it was not trivial to create an aligner from scratch with STAR speed and functionality, but at least I have some alignments. :)
Ola Wallerman🇸🇪 tweet media
English
1
0
1
33
𝕐
𝕐@nomad421·
@OWallerman I think that there is much room for efficiency improvement (several open issues / PRs in the STAR repo already), but the first goal would just be to produce the same output. I think having STAR written in Rust would make it more approachable & raise interest from maintainers.
English
1
0
1
47
𝕐
𝕐@nomad421·
Any bioinfo friends out there have a huge load of AI credits & want to try smth crazy (but amazing if it works)? What if we ask Claude to port STAR to Rust? Current dev & maintenance on this central tool has largely stalled, but I think a lang change would reinvigorate it!
English
2
1
11
1.7K