I know, pal ๐
13.03.2026 19:06
๐ 0
๐ 0
๐ฌ 0
๐ 0
I know, pal ๐
When many students chose to answer the questions from your part in the exam - and answer them rather well :)
Agent running code-review while humans grading exams is not too bad, no? ;-)
Coming up with a post-hoc theory section when your results are weird:
How do people plan when a paper first submitted Oct 2024 is still in the review pipeline (2nd round revision) now 17 months later
Need to learn the secrets from those comfortably keeping up with "consulting" pace ~2 months new project cycle - design, data construction, analysis, writing - and be okay with uncertainty in fast probably suboptimal decision points ๐ค
In some fields, an acceptable high-resolution is 0.5ยฐ - quite a distance from โค30m ๐ค
When working on the theme of sustainable global agrifood systems, you really can't avoid addressing the issues of fertilizer imbalance and pollution ๐๐๐๐พโ ๏ธ๐
How much does it pay to publish an open access academic book? Read this thread for my story and ๐ฒ๐ฐ amounts. ๐งต
www.routledge.com/9781032908724
The ability to easily generate text-based computational illustration for non-geo collaborators is a great boost to comms ;-)
My two posts on AI in academia provoked a lot of people this week, and some of it was unnecessary. I'm sorry for the tone and provocation. But I stand by the substance.
Here are all my 20 theses. I'd genuinely like to know which ones folks *empirically* disagree or agree with.
The superpower of the empirical folks is the ability to quickly intuit, probably visualize in their heads, which variations in the data remain to be exploited after controlling for various controls (time-variant and time-varying) and fixed effects (all the demeaning) - amazing!
Red moon tonight (?) and how cool is this:
/skills adjust to the shared standard for evidence when necessary
What is the average+/-sd age when people lock-in career/field-wise..
There is a documentary on Netflix about the planetary boundaries ๐บ
How we take many things for granted today....
Wonder if Opus 4.6 or GPT 5.3 could have come up with the "floating coordinate" solution back then ๐; need to fact-check, but great series!
Quantitative lines of work move through phases. Chase better predictions ลท = f(x) using statistical and machine learning. Then the focus shifts to collecting richer, cleaner (y, x) pairs. Then exploring the causal y = f(x). Then? In the meantime, the scope/data type/domain expands too!
Recently realizing probably not using AI on the phone enough when on the go - neat to be able to simply ask Gemini something like "Please check my recent emails, anything that requires my response, ignore ..." (or, help picking butters ;-))
and feels conflicted whether bringing in more flawed humans (i.e. lawyers, but in the end a product of the legal system and incentives) led to better or worse resolution, both parties considered
Get what the hype of Netflix's "Marriage Story" is about now. What a script, dialogue, acting! Feels so real-life like with its messiness when imperfect humans with all their flawed qualities interact. No "cleanly designated" heroine/villain to root for, really. But life goes on
"Used responsibly, AI can amplify the impact of our work, rather than its own flaws." - Ismahane Elouafi, CGIAR
โAI can transform agricultural research, as long as we remain realistic about its capabilities and grounded about its risks,โ writes @cgiar-emd.bsky.social in a new #ScienceExpertVoices article. https://scim.ag/4aDqIYu
TIL in Windows Terminal one can actually split panels (vertically: Alt + Shift + plus key; horizontally: Alt + Shift + minus key) and resize the panel (Alt + Shift + arrow keys)
