![Comic. [Gravestone for Joseph Paul Overton with list of years underneath. 1960-2003 is in black at the top and crossed out in red. Underneath in red is 1965-2011, crossed out, then 1973-2018, crossed out, then 1982-]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:cz73r7iyiqn26upot4jtjdhk/bafkreiaut6sdohepzijosvhba55e5ojf35fdla4aejyushtqqr75am54cm)
Comic. [Gravestone for Joseph Paul Overton with list of years underneath. 1960-2003 is in black at the top and crossed out in red. Underneath in red is 1965-2011, crossed out, then 1973-2018, crossed out, then 1982-]
Overton
xkcd.com/3230/
AI summaries for Google search are particularly insidious because pre-AI snippets came with the website details, and I could do a quick scan to find the most trustworthy website to pick an answer from. Now not only are the search results tanked, but also impossible to reliably do a "quick" search.
A single line in this piece just made my day:
“The electricity system operator is understood to be preparing to run the grid without any gas for short periods as soon as this summer, in a first for the UK energy system.”
Renewables *work*. And progress matters🎉
www.theguardian.com/environment/...
Howarth and Jacobson respond to Zeke's post. They ignore his substantive argument about the merits of GWP-20, focusing instead on the importance of near-term climate action. But the case for GWP-20 shouldn't rest whether we like the policy outcomes it produces.
www.howarthlab.org/docs/GWP20_i...
Terminal text from a Claude Code session: Countries in ALL SSPs: 180 (PSE, SYR missing from SSP2) ↓ Filter: has temperature Valid countries: 180 Both historical and future outputs now have exactly 180 countries. Output The earliest_data_year column is preserved in the historical output so you can still see which countries have actual 1980 data vs extrapolated from 1990/1995: ABW: earliest_data_year=1990 AGO: earliest_data_year=1990 ALB: earliest_data_year=1980 ... ✻ Worked for 4m 39s ❯ commit these changes 6 files +363 -242
Coding with LLMs like Claude Code is fun and fast, but for scientific codes one has to be super careful that the LLM did not mis-interpret your intention which is always stated with some ambiguity.
(1/2)
question is not whether the feed is AI-generated, but whether it does what you want it to do. Using AI for feed curation (or coding) is not the same as using it for writing; the latter is justifiably more controversial. 4/4
used to curate feeds since the beginning of feed time. It's only recently that these techniques are being referred to as AI, as they have become very sophisticated. An AI technique may be able to better cater to your viewing preferences than traditional opaque statistics, if done properly. The 3/4
people I follow post close to 100 posts/day. If you follow a lot of people, some sort of statistics will have to be used to cull posts in your following feed to a manageable amount (or to select limited posts from the full firehose). For this reason, opaque (or proprietary) statistics has been 2/4
As a quantitative scientist who is developing a transparent simple statistical feed (called Skylimit) that is deliberately non-AI, I would like to defend/explain the place of so-called AI feeds in the ATmosphere. I consider AI to be just very sophisticated and opaque statistics. Some of the 1/4
New: The April update to the ECMWF seasonal forecast shows an even greater likelihood of a strong El Niño later this year.
A few ensemble members go to +3.5 °C anomaly in the Niño 3.4 region!
This is excellent. I choose not to use AI for my writing, but Naomi's piece reminds me that I'm in no position then to make uninformed judgements about its potential.
Indeed, that's the issue. I use Wise and the like for large money transfers. But for sub-dollar micropayments, fractional bitcoins or stable coins are the only viable option. I've been looking into micropayments for Bsky. The technology exists but the economics is tough.
I recently added the card feature to a web page and tested that it works with Bluesky. It is indeed the Open Graph (OG) metadata tags in the html. Usually your blogging software generates the OG metadata.
Work with me!
Earth, Environmental and Planetary Sciences at Brown University invites applications for a Doherty Postdoctoral Fellowship in Ocean Sciences. This competitive postdoctoral fellowship will be awarded for a one-year period, + 2nd year possible. Please see apply.interfolio.com/183855!
Trump regime again tries to shutter all NOAA labs in its 2027 White House budget. open.substack.com/pub/balanced...
My undergraduate Climate class is making Instagram/TikTok-style videos to explain modes of climate variability to non-experts as part of their homework.
My graduate Statistics class is making statistics-related memes for extra credit.
It's a fun time of year for sure!
Expert observers say they've never seen the IPCC spinning its wheels like this before, just as the world crashes into the 1.5C threshold and climate impacts accelerate ...
🌍🌐 The DestinE AI Tooling Workshop strengthened the technical foundation of advancing machine learning (ML) within the EU initiative. Last month, experts from ECMWF and partner institutions gathered in the Netherlands for a 2,5 day hands-on workshop. Read more ➡️ destine.ecmwf.int/news/destine...
Feedback welcome, if you try it out
This is the app that does volume filters: https://blog.skylimit.dev
It probabilistically limits the number of posts for each follower.
It also does followee-specific keyword filters.
It would be easy to add timed mutes, if it's of interest.
The app's usable but still fragile at this time (I use it)
April 1st video from the @tamuatmo.bsky.social social media team, featuring @eniels13.bsky.social.
www.instagram.com/reel/DWm1bRk...
@tamuartsci.bsky.social
There's a lot of truth to what you say. But professionals like journalists and bloggers may be willing to classify a post if there is a guarantee that their followees will be shown that particular post, instead of showing a random less impactful post. Skylimit provides such a guarantee
That's the exact issue I'm trying to tackle with https://skylimit.dev There are two ways to do that. 1) Convince posters to classify their posts using hashtags (so that more followers will see relevant posts) and/or 2) Readers specify what keywords they want to see in in posts shown from a followee
New article in Knowable Magazine on marine carbon dioxide removal & ocean alkalinity enhancement approaches and pilots studies already underway 🌊
Can we fight climate change by storing CO2 in the ocean? @knowablemag.bsky.social knowablemagazine.org/content/arti...
Maybe Bluesky will get mad at me for saying this, but this is a bizarre headline because if you actually read the article it says the exact opposite.
Please consider reading the whole article before doom-sharing the irresponsible headline.
Irresponsible journalism. Often article authors don't have control over the headlines but this is egregiously bad!
I have the same quibble. Although there are studies suggesting that weather may become less predictable, that's not happening now.
For #NSF and #NIH watchers, Grant Witness now has interactive data on numbers of grants and total funding obligations, broken down by institute and directorate, new awards and non-competitive renewals.
The stranglehold on new awards is still a disaster.
grant-witness.us/funding_curv...
