iiinova — where your digital self keeps evolving

Brett Phinney @ucdproteomics.bsky.social · 01/10/2026

Maybe so ! It’s like me and taking photos with my phone, take enough of them and some will turn out really good !!

Brett Phinney @ucdproteomics.bsky.social · 01/10/2026

I think ai will force all colleges to go back to the 50s or earlier . Eveything will need to be done by hand and tests taken in person orally or with pencils.

010

Brett Phinney @ucdproteomics.bsky.social · 01/10/2026

I definitely feel barriers to entry have lowered for a lot of people that had ideas and never had a way to implement them

110

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

Maybe I should be a greeter ! ! I’ll just hang out outside my core facility and greet people as they drop off samples and then Have the agents do all the work

120

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

One of the best surprises about having accessible code is that I can have the agents read it and explain exactly how it works . I know so much more now than I did before

110

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

I’ve basically stopped using spectronaut because because the licensing and code make it too difficult to use on my cluster . And when radient starts working well on timsTOF data (Claude hasn’t cracked it yet ) I’ll switch to that becuase the code is open and modifiable

110

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

One of the things that actually is surprisingly awesome is making the code available so the agents can study and modify it . That seems to be the best/biggest thing .

110

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

This was from one of my Claude code sessions today .

110

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

Btw sagerecon caught something with a set of samples that I didn’t know . Some were alkylated with acrylamide!! So it works !!

120

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

The sagerecon looks awesome ! Dda only as it’s based on sage I take it ?

110

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

Yes , it spawns a suite of specialized sub agents to check the results . I think I can build in a separate llm to also judge the results as i have deepseq and qwen running on the cluster , but i dont think they are as good the normal frontier models yet

010

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

The skill is much more powerful than the gui. I think gui’s are dead honestly and so is analysis software that’s not agentic . It’s crazy how powerful it is . And if done right, it’s reproducible.

230

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

Yes ! And I was thinking of adding those two grumpy dudes form the muppets that just yell criticisms and talk shit about the results . Maybe for the next version !!

120

Brett Phinney @ucdproteomics.bsky.social · 30/09/2026

I asked Claude to make podcast where a statistician and biologist discuss the proteomics results run through my claude proteomics skill and while it's not as good as @benneely.com podcast, but it's not bad either :) it's in the newest version of the skill github.com/bsphinney/DE...

github.com

GitHub - bsphinney/DE-LIMP: LIMPA DAVIS Proteomics Pipeline

LIMPA DAVIS Proteomics Pipeline. Contribute to bsphinney/DE-LIMP development by creating an account on GitHub.

194

Brett Phinney @ucdproteomics.bsky.social · 17/09/2026

You can still buy IPG strips ? Amazing!!

210

Brett Phinney @ucdproteomics.bsky.social · 16/09/2026

Honestly I've had it with commercial products, you lose all your data eventually 100% guaranteed. This is why open weight AI models and things I can run locally that people can't take away from me really are appealing

020

Brett Phinney @ucdproteomics.bsky.social · 16/09/2026

thanks to Gert in the Metabolomics Core here we have an open weight AI #LLM gateway that I can test current open weight models for analyzing #proteomics Data. #deepseek-v4-flash was not too shabby!! app.notion.com/p/Can-local-...

app.notion.com

Can local AI models analyse proteomics data? — benchmark and prompt fix (FINAL REPORT) | Notion

The question

150

Brett Phinney @ucdproteomics.bsky.social · 05/09/2026

I remember when I was in grad school many labs actually researched mass spectrometers . Crazy huh ?

110

Brett Phinney @ucdproteomics.bsky.social · 05/09/2026

I wonder how many people inside the companies understand them these days. Probably 3-4 tops

130

Brett Phinney @ucdproteomics.bsky.social · 05/09/2026

I really wonder if I’m the only one google Dreambeans sends stuff like this to 🤣https://labs.google/dreambeans

110

Brett Phinney @ucdproteomics.bsky.social · 02/09/2026

Interesting . I have a feeling we will be able to account for more of these kind of things soon .

010

Brett Phinney @ucdproteomics.bsky.social · 02/09/2026

Bruker actually does some surprisingly sophisticated things . For example the Postgres db that records eveything the instrument is doing. They dont give out the password , but Claude can easily read the automated backups that are made everyday. I’ll show some examples soon

010

Brett Phinney @ucdproteomics.bsky.social · 02/09/2026

Plus Bruker’s crappy file format was pretty useless before AI , but now holy smokes ! Claude can read everything about the files easily without needing a file parser like with Thermo’s raw files . I’ve completely changed my tune ! Yes they can be corrupted more easily , but now , worth it !

110

Brett Phinney @ucdproteomics.bsky.social · 01/09/2026

I had claude do it lol. Say what you will about Brukers shitty software, some things they do are just godsends. Like always appending the plate location and a continuous injection number to the file names. Why can't thermo do that??

130

Brett Phinney @ucdproteomics.bsky.social · 01/09/2026

Haha, e don't use PRTC standards. Do you find that they help? DIA-NN and Spectroanut seem to do fine without them . We usually put samples in those locations I think. Maybe they are easiest to reach ?

110

Brett Phinney @ucdproteomics.bsky.social · 31/08/2026

Today in useless data: I don't know why I was curious about my labs "popular @evosep.bsky.social locations are" But here you go . I apparently need to use S5 more

6103

Brett Phinney @ucdproteomics.bsky.social · 02/07/2026

That’s some serious money you’re wagering 😉

010

Brett Phinney @ucdproteomics.bsky.social · 30/06/2026

I wonder what type of proteomics data this is ? Olink ? Anyone know ?

300

Brett Phinney @ucdproteomics.bsky.social · 10/06/2026

No, she works on plant evolutionary genetics.

100

Brett Phinney @ucdproteomics.bsky.social · 10/06/2026

My daughter (now a 2nd year grad student at Irvine) found this yesterday. She made it in 1st grade for a school project. Stick around until the last page you won't be sorry 🤣😂

7211

Brett Phinney @ucdproteomics.bsky.social · 10/06/2026

They are peptidomics/paleoproteomcis samples. DDA is still a bit better for those I think...

010

Brett Phinney @ucdproteomics.bsky.social · 10/06/2026

I just used Sage to search 120 Astral DDA files in 12 minutes my cluster. I honestly don't understand what Thermo's Ardia thingy does or why it's even needed

670

Brett Phinney @ucdproteomics.bsky.social · 18/05/2026

It's that time of year again for are Annual proteomics chart course! Going on almost 20 years! Honestly it's a good class for those wanting to add proteomics to your research. And Proteomics really is the best Omics if I don't say so myself #proteomics #massspec #proteome #massspectrometry

2163

Brett Phinney @ucdproteomics.bsky.social · 14/05/2026

Both I think . Although the lumos can still churn out some great data , it just takes a lot longer

110

Brett Phinney @ucdproteomics.bsky.social · 14/05/2026

I seem to have acquired and kept over 233K mass spec files over there years. Here is a breakdown

1110

Brett Phinney @ucdproteomics.bsky.social · 13/05/2026

See I Knew I would have some use for keeping every qc file I ever created 😁 I'll have more in a day or two

030

Brett Phinney @ucdproteomics.bsky.social · 11/05/2026

Well, I apparently saved over 32K BSA QC files from my LTQ era. Now I need to figure out what to do with them lol

350

Brett Phinney @ucdproteomics.bsky.social · 11/05/2026

I really like having a computer cluster. I have been using it a crazy art more since I have switched to claude code. Currently - 2048 cores running running (across 259 jobs)- 283 jobs pending - 1 GPU-A100 job pending

130

Brett Phinney @ucdproteomics.bsky.social · 11/05/2026

I haven't been using it lately, but there are no other options, so I might again :) If only to record what I have been doing

010

Brett Phinney @ucdproteomics.bsky.social · 25/03/2026

Claude Code and I whipped up some software for running DIA-NN and LIMPA. I kinda like it so far. Has a lot of features I've always wanted and now I can make it myself !!, well with Claude Codes help ;) youtu.be/lXxqDpCHXOA

youtu.be

De-LIMP Proteomics Analysis Software Part1

YouTube video by Brett Phinney

161