What truly hinders progress in biology research? You may have your favorite – automation, lack of great ideas, regulatory hurdles – but here's one I believe outweighs everything else – research planning. 1/3

This will be my ref for why I switched from AI to wet lab. Boils down to the key thing that matters in life, as Jordan Peterson once insightfully pinpointed: avoiding pain. But there's no mystery why most don't choose this path: near-term rewards are much more lucrative.

Another YouTube Channel Idea: "E. coli Design." A whole series where you explain how to design DNA to change the behavior or appearance of E. coli cells. All you'd need is some equipment to transform cells, a microscope and camera, and some basic molecular biology equipment.

@Pioneer__Labs Full @erika_alden_d / @Pioneer__Labs post:

@Pioneer__Labs 5. All stocks are neatly labeled (with a printed label) and stored in a database that's always accessible while down in the lab.

@Pioneer__Labs 4. Avoiding feeling overwhelmed: Every few weeks all incomplete projects are marked as "Deprioritized" and the team discusses what the true current priorities are.

@Pioneer__Labs 3. Quality checks: People are automatically reminded to fill out the "Measure of success" field upon the initiation of an experiment and "Conclusions" upon a completion.

@Pioneer__Labs 2. Structured notes: Each experiment is described in a single database entry with structured metadata, incl. status. Experiments are grouped into projects that describe higher-level picture.

@Pioneer__Labs 1. Review system: People get to review each others work upon completion to stay up to date and to make sure the notes make sense to others.

How to organize wet lab notes so that they are more than a pile of unintelligible scribbles? @Pioneer__Labs have a great process; below are several ideas that stood out for me:

Dear early career scientists! Why not join us in September in Lithuania’s beautiful capital city for the EMBO Young Scientists’ Forum (EYSF)? We will have have a full Night Science workshop & exciting research sessions. Abstract submission deadline: Aug 1 https://t.co/cbA8ZdGHIm

Full text:

I recently turned 40. On that occassion I sat down to reflect on a theme that's been on my mind for the past decade – Cain's journey. Commonly known for killing his brother, Cain is in fact a multifaceted and deeply human character who is worth meditating about.

When we think about the potential impact of human gene editing, it is hard to think big enough Every other technology alters the world to suit human preferences. Gene editing could alter human preferences If you change the basis of human needs and wants, all bets are off

You can automate anything but personal experiences: "A machine can write prose that appears to reflect on an experience, but the lived nature of that experience belongs solely to the human author." You won't earn more or become famous, but authentic life is the ultimate goal.

Brings so much more meaning to our efforts to optimize CRISPR-Cas systems. When things are constantly not working in the wet lab, it's great to be reminded that one should persist because ultimately it pays off big time. Definitely a highlight of the past few months for me!

Only if there are angels in your head will you      ever, possibly, see one. – Mary Oliver https://t.co/8nYBJ2kLRV

Encouragingly, OpenAI just published a paper doing just that for ML ( https://t.co/0sp7FnbYYU)! Their grading approach is great and possibly largely automatable. Plus, unlike them, I don't think we even need to run exp's – one can judge protocols just by looking at them. 6/6

My prediction – LLMs will struggle. But not because they don't know enough biology. They do. But they are not good (yet) at putting their knowledge together into a sequence of actions. So this benchmark would establish a new frontier to tackle. 5/6

Level 1: Given a paper, a model needs to provide a detailed protocol how each figure could be obtained, incl. concentrations and volumes. Level 2: Now LLMs are given only the main findings and need to figure out various controls that usually make up most of the figures. 4/6

Yet biology papers notoriously skimp on their methods. With a lot of effort, a trained biologist might be able to recap the real protocol and replicate published experiments. What about LLMs? 3/6