Joined January 2019
- Tweets 94
- Following 360
- Followers 872
- Likes 1,770
15 Photos and videos
Pinned Tweet
7 Dec 2021
I spent months working on domain-specific search engines and knowledge discovery apps for biomedicine and eventually figured that synthesizing "insights" or building knowledge graphs by machine-reading the academic literature (papers) is *barely useful* :
markusstrasser.org/extractin…
18
85
346
Markus Strasser retweeted
24 May 2023
Excited to share what I’ve been up to with the mech interp team at Anthropic!
transformer-circuits.pub/202…
24 May 2023
Our Interpretability team is experimenting with “Updates” – small, informal research notes in between our major papers.
ALT Screenshots of the figures from the May 2023 interpretability release artistically stacked on top of one another
3
60
10,672
25 Apr 2022
"why would any machine learning researcher want to work at something that isn’t at the forefront of generalization capability? ". Great read
25 Apr 2022
Job search is over. I've joined @HalodiRobotics as the VP of AI! Also, some things on what I learned talking to a bunch of ML companies in the last month:
evjang.com/2022/04/25/rome.h…
4
31
31 Jan 2022
I've bio-pilled a few SWE friends and know a dozen more that would take a pay cut to go into bio/research software if it was seen as a valid "cool" career with a clear path and enough salary to build a family eventually
30 Jan 2022
I've spent several years building software tools in genomics.
While the work couldn't be more exciting, there are unique challenges and tensions in the world of scientific software.
How does software in the life sciences actually work (and not work)?
newscience.org/how-software-…
5
2
40
Markus Strasser retweeted
3 Jan 2022
my gut is that productivity software for the life sciences, particularly R&D, will turn out to be a failure.
there's 10 companies in this space, and none have cracked a GTM. The simple truth is R&D is cash strapped, doesn't buy software, and has impossibly long feedback loops
4
4
85
31 Dec 2021
was wondering about the angry nerd mob with torches gathering outside my apartment @slatestarcodex
17
9 Dec 2021
apparently this hit #1 on HN in the morning and spawned some really solid discussions and personal reports ( the usual heckling). Thanks to @kev_in_hu for posting it to HN .. totally forgot that's a thing
news.ycombinator.com/item?id…
7 Dec 2021
I spent months working on domain-specific search engines and knowledge discovery apps for biomedicine and eventually figured that synthesizing "insights" or building knowledge graphs by machine-reading the academic literature (papers) is *barely useful* :
markusstrasser.org/extractin…
2
1
9
7 Dec 2021
I spent months working on domain-specific search engines and knowledge discovery apps for biomedicine and eventually figured that synthesizing "insights" or building knowledge graphs by machine-reading the academic literature (papers) is *barely useful* :
markusstrasser.org/extractin…
18
85
346
7 Dec 2021
The idea of semantically structuring research findings (text) is a seductive siren call and there's many shipwrecks. So if this report helps to prevent just a single one from happening in the future I'm happy with it.
2
1
42
7 Dec 2021
thanks to @Ben_Reinhardt @bgavran3 @mattsclancy @rmhsilva @rmeinl and @an1lam for their thoughts on the draft
3
9
18 Nov 2020
More people should learn to program but less people should become software developers
1
9
12 Nov 2020
Textbook quote right here:
12 Nov 2020
physics is capable of freaky stuff. evolution's job is to find it.
1
6
30 Oct 2020
didn't expect a post about category theory to make the link between slave labor and bad infrastructure tooling for me
"[...] the Greeks didn’t need technology because they had access to cheap labor through slavery."
bartoszmilewski.com/2020/02/…
1
2
30 Oct 2020
From my own experience: there's no "pubmed on steroids" yet because big pharma can just hire 100 cheap PhDs in Hyderabad to parse papers into JSON.
1
1
4
30 Oct 2020
Even Oxford postdocs (bio/chem) make only about 35k£ a year so just use 30 of them and deploy them as a moist web-crawler fleet. Postdocalypse ftw
1
1
25 May 2020
I made a musical pitch detection algorithm when I started out to code 3 years ago. It detects notes and chords in <20ms.
I built this interactive spectrogram with it:
github.com/mkstra/polyphon
It's great for music apps. If there's interest I can refactor it into a clean library
1
5