Michael MacCoss
@maccoss.bsky.social
Professor of Genome Sciences University of Washington, Seattle. Interested in proteomics and mass spectrometry.
Same with code. I might not have written any of my students programs but I've helped evaluate them, document, and confirm how the algorithms work.
I totally agree w/ your hammer metaphor. I'm totally worried about LLMs impact but people are still going to use power tools. Hopefully we do it safely.
I totally agree w/ your hammer metaphor. I'm totally worried about LLMs impact but people are still going to use power tools. Hopefully we do it safely.
January 31, 2026 at 5:08 PM
Same with code. I might not have written any of my students programs but I've helped evaluate them, document, and confirm how the algorithms work.
I totally agree w/ your hammer metaphor. I'm totally worried about LLMs impact but people are still going to use power tools. Hopefully we do it safely.
I totally agree w/ your hammer metaphor. I'm totally worried about LLMs impact but people are still going to use power tools. Hopefully we do it safely.
I should clarify. I'm not saying LLMs are like a student. I find I interact with them like I would a trainee. I would never let text from a student get submitted without significant time editing & revising. As a student gets more experienced there is less revision. These tools will get better soon.
January 31, 2026 at 5:08 PM
I should clarify. I'm not saying LLMs are like a student. I find I interact with them like I would a trainee. I would never let text from a student get submitted without significant time editing & revising. As a student gets more experienced there is less revision. These tools will get better soon.
I'm personally much more against kits with mystery reagents than AI methods.
January 31, 2026 at 4:48 PM
I'm personally much more against kits with mystery reagents than AI methods.
We've been through similar transitions in the past. Should we let trainees use kits, calculators, Google searches, Intellij, etc... These tools have the potential of advancing science significantly but we all have concerns about whether we could be creating a gap in the institutional knowledge.
January 31, 2026 at 4:48 PM
We've been through similar transitions in the past. Should we let trainees use kits, calculators, Google searches, Intellij, etc... These tools have the potential of advancing science significantly but we all have concerns about whether we could be creating a gap in the institutional knowledge.
Everything we do is based on building on the existing literature. The responsibility is still on the author for ensuring the content is accurate and appropriately cited. If the content is BS then the community/reviewers, etc... need to call it out. It doesn't matter how the content was written.
January 31, 2026 at 4:48 PM
Everything we do is based on building on the existing literature. The responsibility is still on the author for ensuring the content is accurate and appropriately cited. If the content is BS then the community/reviewers, etc... need to call it out. It doesn't matter how the content was written.
These tools aren't going away and they will get much more accurate and powerful. The real question is what's the best way to use them. Just like how we need to be skeptical with everything we get from Google or even the literature; this doesn't change that.
January 31, 2026 at 4:48 PM
These tools aren't going away and they will get much more accurate and powerful. The real question is what's the best way to use them. Just like how we need to be skeptical with everything we get from Google or even the literature; this doesn't change that.
I'm not sure we need 1000s of AI generated scripts or web apps in the literature but I'm supportive if those help an individual QC, visualize, or interpret their data. We do need to think about the best ways to use these tools and how to train future scientists on how to use them appropriately.
January 31, 2026 at 4:23 PM
I'm not sure we need 1000s of AI generated scripts or web apps in the literature but I'm supportive if those help an individual QC, visualize, or interpret their data. We do need to think about the best ways to use these tools and how to train future scientists on how to use them appropriately.
I find the process of using AI similar to mentoring a trainee. You get drafts that need revision and implementations of algorithms that produce results that don't make sense. That said, AI methods are improving exponentially and becoming very useful. Claude Opus 4.5 appears on a different level.
January 31, 2026 at 4:23 PM
I find the process of using AI similar to mentoring a trainee. You get drafts that need revision and implementations of algorithms that produce results that don't make sense. That said, AI methods are improving exponentially and becoming very useful. Claude Opus 4.5 appears on a different level.
I definitely don't need a TV that big. 😂
January 10, 2026 at 3:38 AM
I definitely don't need a TV that big. 😂
That's what I use my 49" monitor for!
January 8, 2026 at 2:29 AM
That's what I use my 49" monitor for!
We can start the rumor that Skyline is named after the Nissan Skyline if it would help 😀.
December 13, 2025 at 4:46 AM
We can start the rumor that Skyline is named after the Nissan Skyline if it would help 😀.
We're all struggling with this these days.
November 21, 2025 at 10:46 PM
We're all struggling with this these days.
We use this strategy to assess the improvement in the Astral Zoom over the original. We also use this strategy to evaluate the effect of different acquisition strategies. Looking forward to seeing others apply this approach in the evaluation of other instruments. Great work by @chrhsu.bsky.social.
October 22, 2025 at 3:32 PM
We use this strategy to assess the improvement in the Astral Zoom over the original. We also use this strategy to evaluate the effect of different acquisition strategies. Looking forward to seeing others apply this approach in the evaluation of other instruments. Great work by @chrhsu.bsky.social.