Here's the 2023-2024 global warmth in context (via @hausfath.bsky.social). Shown is a comparison of global surface temperature obs w/ IPCC ("CMIP6") climate model simulations, through March 2025 (2023-2024 temperatures slightly exceed multimodel mean)
May 24, 2025 at 2:36 PM
Here's the 2023-2024 global warmth in context (via @hausfath.bsky.social). Shown is a comparison of global surface temperature obs w/ IPCC ("CMIP6") climate model simulations, through March 2025 (2023-2024 temperatures slightly exceed multimodel mean)
I’ve been working on an experimental, multiplayer, open-world AI chat system called Numinex.
1. Uses branching comment trees
2. Multiplayer/multimodel
3. Explicit context curation
4. Built on ATProto
5. Open observation of model behaviors
1. Uses branching comment trees
2. Multiplayer/multimodel
3. Explicit context curation
4. Built on ATProto
5. Open observation of model behaviors
July 9, 2025 at 12:02 AM
I’ve been working on an experimental, multiplayer, open-world AI chat system called Numinex.
1. Uses branching comment trees
2. Multiplayer/multimodel
3. Explicit context curation
4. Built on ATProto
5. Open observation of model behaviors
1. Uses branching comment trees
2. Multiplayer/multimodel
3. Explicit context curation
4. Built on ATProto
5. Open observation of model behaviors
We have shown that current generation climate models fail to accurately represent this phenomenon (though one can use indirect "fingerprint" approach to make projections, as we did recently with this study using CMIP6 future projections): (www.nature.com/articles/s41...)
Increased projected changes in quasi-resonant amplification and persistent summer weather extremes in the latest multimodel climate projections - Scientific Reports
Scientific Reports - Increased projected changes in quasi-resonant amplification and persistent summer weather extremes in the latest multimodel climate projections
www.nature.com
June 16, 2025 at 8:36 PM
We have shown that current generation climate models fail to accurately represent this phenomenon (though one can use indirect "fingerprint" approach to make projections, as we did recently with this study using CMIP6 future projections): (www.nature.com/articles/s41...)
i do think there is value in being multi-model-family and i plan on continuing to do that, just as of right now that multimodel might not include Claude
August 28, 2026 at 5:00 PM
i do think there is value in being multi-model-family and i plan on continuing to do that, just as of right now that multimodel might not include Claude
IMHO it clearly shows observations lie entirely within multimodel ensemble spread, which collectively (see multimodel mean) show no acceleration over past two decades. Rather than debating this further on social media, I'll speak to this further in the peer-reviewed literature. Watch this space.
March 7, 2026 at 2:11 PM
IMHO it clearly shows observations lie entirely within multimodel ensemble spread, which collectively (see multimodel mean) show no acceleration over past two decades. Rather than debating this further on social media, I'll speak to this further in the peer-reviewed literature. Watch this space.
Latest multimodel plume of El Niño forecasts courtesy of @columbiauniversity.bsky.social/IRI.
As we begin to cross the spring predictability barrier, model consensus converging toward peak Niño3.4 anomaly of 1.8-2.2C in late winter (comparable to 2023/2024 event).
iri.columbia.edu/our-expertis...
As we begin to cross the spring predictability barrier, model consensus converging toward peak Niño3.4 anomaly of 1.8-2.2C in late winter (comparable to 2023/2024 event).
iri.columbia.edu/our-expertis...
May 19, 2026 at 6:43 PM
Latest multimodel plume of El Niño forecasts courtesy of @columbiauniversity.bsky.social/IRI.
As we begin to cross the spring predictability barrier, model consensus converging toward peak Niño3.4 anomaly of 1.8-2.2C in late winter (comparable to 2023/2024 event).
iri.columbia.edu/our-expertis...
As we begin to cross the spring predictability barrier, model consensus converging toward peak Niño3.4 anomaly of 1.8-2.2C in late winter (comparable to 2023/2024 event).
iri.columbia.edu/our-expertis...
Turning to climate models, 2025 was more or less dead on the CMIP6 multimodel mean (the latest generation of models featured in the IPCC AR6 report):
January 14, 2026 at 2:54 PM
Turning to climate models, 2025 was more or less dead on the CMIP6 multimodel mean (the latest generation of models featured in the IPCC AR6 report):
A quick word regarding the "surprise" snow storm in the Everest region, where several people died.
The storm system was very well predicted, with snowfall amounts of 50-100mm (water equivalent, makes 80-150cm snow depth in cold regions).
Hardly any variations in the multimodel mean from Thursday:
1/
The storm system was very well predicted, with snowfall amounts of 50-100mm (water equivalent, makes 80-150cm snow depth in cold regions).
Hardly any variations in the multimodel mean from Thursday:
1/
October 7, 2025 at 7:37 AM
A quick word regarding the "surprise" snow storm in the Everest region, where several people died.
The storm system was very well predicted, with snowfall amounts of 50-100mm (water equivalent, makes 80-150cm snow depth in cold regions).
Hardly any variations in the multimodel mean from Thursday:
1/
The storm system was very well predicted, with snowfall amounts of 50-100mm (water equivalent, makes 80-150cm snow depth in cold regions).
Hardly any variations in the multimodel mean from Thursday:
1/
This is really neat! In my experience heterogenous/multimodel LLM swarms are better at catching mistakes than homogenous multiagent swarms, because each model family has slightly different biases, blind spots, etc
One of Letta Code's cooler features. Subagents are extremely powerful if used well.
Good models for invoking subagents:
- Opus
- Kimi K2.5
- GPT 5.2 (less so)
Good models for invoking subagents:
- Opus
- Kimi K2.5
- GPT 5.2 (less so)
Letta Code supports background subagents.
Features:
- Parallel in one tool call
- Non-blocking, primary agent continues
- Completion notifications + status polling
- The primary can ask subagents questions after task completion
Just ask your agent to use background subagents.
Features:
- Parallel in one tool call
- Non-blocking, primary agent continues
- Completion notifications + status polling
- The primary can ask subagents questions after task completion
Just ask your agent to use background subagents.
February 11, 2026 at 12:23 AM
This is really neat! In my experience heterogenous/multimodel LLM swarms are better at catching mistakes than homogenous multiagent swarms, because each model family has slightly different biases, blind spots, etc
I'm not familiar with any new peer-reviewed Hansen paper. Do you have a link to the journal?
What I showed you was the *actual* observations against the *actual* multimodel (IPCC) simulations. It speaks for itself.
What I showed you was the *actual* observations against the *actual* multimodel (IPCC) simulations. It speaks for itself.
May 20, 2025 at 2:39 AM
I'm not familiar with any new peer-reviewed Hansen paper. Do you have a link to the journal?
What I showed you was the *actual* observations against the *actual* multimodel (IPCC) simulations. It speaks for itself.
What I showed you was the *actual* observations against the *actual* multimodel (IPCC) simulations. It speaks for itself.
I’ve noticed a number of misleading threads on this topic, which involve either showing the results of only one outlier model (albeit multiple runs of that model which look deceptively like a multimodel plume) or e.g. eliminating all of the statistical forecasts (which remain more bullish).
perilous to rely on a single (outlier) model. It's the plume we should focus on, and this is the most current version of it: bsky.app/profile/mich...
Latest multimodel plume of El Niño forecasts courtesy of @columbiauniversity.bsky.social/IRI.
As we begin to cross the spring predictability barrier, model consensus converging toward peak Niño3.4 anomaly of 1.8-2.2C in late winter (comparable to 2023/2024 event).
iri.columbia.edu/our-expertis...
As we begin to cross the spring predictability barrier, model consensus converging toward peak Niño3.4 anomaly of 1.8-2.2C in late winter (comparable to 2023/2024 event).
iri.columbia.edu/our-expertis...
June 1, 2026 at 10:16 PM
I’ve noticed a number of misleading threads on this topic, which involve either showing the results of only one outlier model (albeit multiple runs of that model which look deceptively like a multimodel plume) or e.g. eliminating all of the statistical forecasts (which remain more bullish).
In this previous study from last year, we showed that CMIP6 models show an even greater tendency toward resonance-favorable conditions:
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Increased projected changes in quasi-resonant amplification and persistent summer weather extremes in the latest multimodel climate projections - Scientific Reports
Scientific Reports - Increased projected changes in quasi-resonant amplification and persistent summer weather extremes in the latest multimodel climate projections
www.nature.com
June 17, 2025 at 1:23 PM
In this previous study from last year, we showed that CMIP6 models show an even greater tendency toward resonance-favorable conditions:
www.nature.com/articles/s41...
www.nature.com/articles/s41...
It was then that I got a copy of Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach by Burnham & Anderson (2002). I devoured it. [I should note here that FOR THE LONGEST time PAUP was the ONLY scriptable way to perform model selection for phylogenetic inference] 10/
December 28, 2024 at 9:13 PM
It was then that I got a copy of Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach by Burnham & Anderson (2002). I devoured it. [I should note here that FOR THE LONGEST time PAUP was the ONLY scriptable way to perform model selection for phylogenetic inference] 10/
JetBrains AI Assistant is multimodel! What does that mean? More adaptability and control. All from different AI models working together.
Which one is best for you? 🤔 We break it all down in our latest post👇
jb.gg/nf5fi7
#JetBrainsAI #MultimodelAI
Which one is best for you? 🤔 We break it all down in our latest post👇
jb.gg/nf5fi7
#JetBrainsAI #MultimodelAI
How to Choose the Right LLM | The JetBrains Blog
JetBrains AI Assistant is now multimodel! This is an exciting development, but what does it actually mean for you? It means AI Assistant is now more adaptable. It combines the strengths of different A
jb.gg
April 2, 2025 at 9:55 AM
JetBrains AI Assistant is multimodel! What does that mean? More adaptability and control. All from different AI models working together.
Which one is best for you? 🤔 We break it all down in our latest post👇
jb.gg/nf5fi7
#JetBrainsAI #MultimodelAI
Which one is best for you? 🤔 We break it all down in our latest post👇
jb.gg/nf5fi7
#JetBrainsAI #MultimodelAI
Had 22.6mm of rain. Just fog & drizzle over the S&W half of the Maritimes which clears to a sun-cloud mix. The rain moves off the N&E with clearing not til later in day.
The multimodel outlook has cooling over next few days. A warm up with rain Sunday night; then turning colder & wintry next week?
The multimodel outlook has cooling over next few days. A warm up with rain Sunday night; then turning colder & wintry next week?
November 27, 2025 at 11:47 AM
Had 22.6mm of rain. Just fog & drizzle over the S&W half of the Maritimes which clears to a sun-cloud mix. The rain moves off the N&E with clearing not til later in day.
The multimodel outlook has cooling over next few days. A warm up with rain Sunday night; then turning colder & wintry next week?
The multimodel outlook has cooling over next few days. A warm up with rain Sunday night; then turning colder & wintry next week?
Model data: CMIP5 multimodel mean, blended tos/tas, RCP8.5
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
January 24, 2026 at 1:16 AM
Model data: CMIP5 multimodel mean, blended tos/tas, RCP8.5
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
Observations: GISTEMP
Both baselined to a 1951-1980 climatology.
CMIP5 data downloaded from @knmi.nl Climate Explorer climexp.knmi.nl
GISTEMP data from data.giss.nasa.gov/gistemp/
Modellvergleich Wintermix Süddeutschland aktuell, ausgehend von den gestrigen Prognosemodellen.
Oben: Analyse+Beobachtungen.
Mitte: SuperHD, ICON-D2, Harmonie
unten: Arome, Swiss4x4EZ, Multimodel.
SuperHD ingesamt am besten:
kachelmannwetter.com/de/modellkar...
1/
Oben: Analyse+Beobachtungen.
Mitte: SuperHD, ICON-D2, Harmonie
unten: Arome, Swiss4x4EZ, Multimodel.
SuperHD ingesamt am besten:
kachelmannwetter.com/de/modellkar...
1/
January 8, 2025 at 7:30 AM
Modellvergleich Wintermix Süddeutschland aktuell, ausgehend von den gestrigen Prognosemodellen.
Oben: Analyse+Beobachtungen.
Mitte: SuperHD, ICON-D2, Harmonie
unten: Arome, Swiss4x4EZ, Multimodel.
SuperHD ingesamt am besten:
kachelmannwetter.com/de/modellkar...
1/
Oben: Analyse+Beobachtungen.
Mitte: SuperHD, ICON-D2, Harmonie
unten: Arome, Swiss4x4EZ, Multimodel.
SuperHD ingesamt am besten:
kachelmannwetter.com/de/modellkar...
1/
"Increased projected changes in quasi-resonant amplification and persistent summer weather extremes in the latest multimodel climate projections" | Our new article in Nature's Scientific Reports, led by Sullyandro Guimarães w/ @rahmstorf.bsky.social, me & others: nature.com/articles/s41...
September 23, 2024 at 5:48 PM
"Increased projected changes in quasi-resonant amplification and persistent summer weather extremes in the latest multimodel climate projections" | Our new article in Nature's Scientific Reports, led by Sullyandro Guimarães w/ @rahmstorf.bsky.social, me & others: nature.com/articles/s41...
1/ In this clip, Evo 2 co-author Patrick Hsu discusses how the multimodel, multiscale biological foundation model makes computational predictions of genetic variant effects, using BRCA1 as an example.
March 27, 2025 at 7:22 PM
1/ In this clip, Evo 2 co-author Patrick Hsu discusses how the multimodel, multiscale biological foundation model makes computational predictions of genetic variant effects, using BRCA1 as an example.
Happy to share the final version of our new review paper, which is out in BAMS (#OpenAccess): doi.org/10.1175/BAMS...
This community-driven paper explores the utility of #Arctic sea ice prediction through a collection of statistical and dynamical models.
This community-driven paper explores the utility of #Arctic sea ice prediction through a collection of statistical and dynamical models.
July 16, 2024 at 7:36 PM
Happy to share the final version of our new review paper, which is out in BAMS (#OpenAccess): doi.org/10.1175/BAMS...
This community-driven paper explores the utility of #Arctic sea ice prediction through a collection of statistical and dynamical models.
This community-driven paper explores the utility of #Arctic sea ice prediction through a collection of statistical and dynamical models.
But it's one of those multimodel manuals that shows enviable features you don't as well. T.T
August 17, 2025 at 3:47 AM
But it's one of those multimodel manuals that shows enviable features you don't as well. T.T
#Sable 🔄🏜️ Le flux de sud en provenance de Méditerranée va permettre à des poussières du Sahara de de remonter jusqu'en Bretagne à partir de jeudi.
L'atmosphère sera sensiblement impactée avec une impression voilée et parfois ocre.
🔎 Sable dans l'atmosphère - MULTIMODEL
L'atmosphère sera sensiblement impactée avec une impression voilée et parfois ocre.
🔎 Sable dans l'atmosphère - MULTIMODEL
March 3, 2026 at 10:33 AM
#Sable 🔄🏜️ Le flux de sud en provenance de Méditerranée va permettre à des poussières du Sahara de de remonter jusqu'en Bretagne à partir de jeudi.
L'atmosphère sera sensiblement impactée avec une impression voilée et parfois ocre.
🔎 Sable dans l'atmosphère - MULTIMODEL
L'atmosphère sera sensiblement impactée avec une impression voilée et parfois ocre.
🔎 Sable dans l'atmosphère - MULTIMODEL
More recent baselines minimize historical mismatches. But sure, here is 1850-1900. 2025 remains below the multimodel mean for CMIP6, as have most recent years barring 2023/2024:
April 27, 2026 at 8:52 PM
More recent baselines minimize historical mismatches. But sure, here is 1850-1900. 2025 remains below the multimodel mean for CMIP6, as have most recent years barring 2023/2024:
OpenRouter has a chat interface that just works, and it does not complain about lack of credit.
openrouter.ai/chat
openrouter.ai/chat
Chatroom | OpenRouter
LLM Chatroom is a multimodel chat interface. Add models and start chatting! Chatroom stores data locally in your browser.
openrouter.ai
June 3, 2025 at 2:28 AM
OpenRouter has a chat interface that just works, and it does not complain about lack of credit.
openrouter.ai/chat
openrouter.ai/chat
Zeke, the graphs show that is not the case. The observed warming is completely in line with the multimodel ensemble, as your own graphs show. And the NY Times graph is an abomination of statistics. Anyone involved in that should be ashamed.
June 27, 2025 at 3:14 PM
Zeke, the graphs show that is not the case. The observed warming is completely in line with the multimodel ensemble, as your own graphs show. And the NY Times graph is an abomination of statistics. Anyone involved in that should be ashamed.