Still trying to explain Trump's tariffs as economic policy, instead of tools to gain leverage on countries and companies 🤦♂️🤦♀️🤦 https://www.reuters.com/legal/litigation/trumps-new-tariffs-shift-focus-balance-payme…
Having a lot of fun with the Voigtländer Nokton 50mm f/1 and 75mm f/1.5 manual focus lenses for Canon RF mount https://www.flickr.com/photos/kohlmannj/55165385824/in/album-72177720332529506
At #Kubecon, all the keynotes focus on AI. But apparently, nobody focused on the basics. So, the network is lousy as hell. Even the in-house video stream to the overflow rooms often looks like that.
RE: https://mastodon.social/@SteveRudolfi/116279083767770070
"If there’s one insight we all need to focus on most, it’s this: your job is no longer to build a destination. It’s to build a parts library. And one that’s well documented so that when…
A few pages is better than no pages. Structural Edit. Slow. Mind unwilling to focus for any period of time. I keep drifting, having to pick up the thread and reread what I just did.
Tomorrow will be a better day for writing. #writing #grief
Fruit of a sweet gum tree caught next to Cornell's McGraw tower: I can't get over how well Sony's top of the line 70-200mm lens feels in my hand and how easy it is to work the controls
#photo #photography
Allen’s Hummingbird approaching a bicolor cobra lily. Huntington Library, San Marino, California, USA. February, 2026. OM System OM-1 M.Zuiko 300mm F4 MC14 #huntingtonlibrary #hummingbird #lily
found a nice lake today while cycling off the beaten track
#lake
Japan's Sakana AI launches Sakana Chat, its first consumer chatbot, marking a shift from its business focus as competition to localize generative AI intensifies (Tsubasa Suruga/Nikkei Asia)
https://asia.nikkei.com/business/techn
And there we are. Trump is already blackmailing Europe with 'his' LNG. We urgently need to reduce our natural gas consumption by speeding up the energy transition: focus on energy efficiency, electrification, and scaling up EU production of biomethane, developing way too slow to meet its 2030 target.
And wasn't Charlie pushing for release of the Epstein files?
Erika Kirk's Corcoran Group link in focus amid company's Epstein links; ‘handled his assets…’ | Hindustan Times
https://www.hindustantimes.com/world-news/us-news/erika-kirks-corcoran-group-link-in-focus-amid-companys-epstein-links-handled-his-assets-101771466880445.html
✍️ New post: Continvoucly Morged Value
Couldn't get out of my head what happened to @…’s diagram – and so I wrote down a few premature thoughts about what happens when we mistake the (AI) artifact for the work …
The 21-year-old man who was shot and killed after having entered Donald Trump’s Florida resort on Sunday – while carrying a shotgun – came from a North Carolina family of the president’s supporters and had reportedly become increasingly fixated on the so-called Jeffrey Epstein files.
The focus of the FBI’s investigation into the intrusion attributed to Austin Tucker Martin is tightening on his movements and motives.
Martin was confronted by Secret Service agents and a local sheri…
"Ecological fitting" in an image: a domestic honeybee, Apis mellifera, on flowers of a western Joshua tree, Yucca brevifolia, at Quartz Hill, California back in January
#naturalist
Tackling complex tasks can feel like taming chaos, but classical music is a brilliant remedy. A well-composed symphony turns chaos into clarity. 🎻 Personally, it elevates my focus and productivity. Do you harness music to boost your performance? 🎶 Let's orchestrate success with the right rhythm!
Saw this decaying wooden rain gutter in the morning sun on my way to work the other day.
#mosstodon #SilentSunday
Tons of updates to my CM tool:
* An agent for continuous management with fast-monitoring features
* JSON Schemas for editors
* Manifests and Hiera now support HTTP(s)
* ccm extendible with appbuilder for custom UIs around manifests (no code needed)
* Pre-, Post-messages to help users use things they installed
* Generic if / unless on every resource
* New docs
Focus is to excel at adhoc, snow flakes, devel VMs, a la carte CM etc, place where other too…
Last month 86-year-old Sandra Folzer set the world W85 indoor mile record of 9:43.00 at the Hartshorne Memorial Masters Mile in Barton Hall
#photo #photography #sports
Kind of a dud article, given it promises mid than it delivers about the actual spawning process, but it contains an amazing fact:
Apparently Sigmund Freud did an unsuccessful stint as a marine biologist and was tasked with dissecting 400 male eels to try to find their reproductive organs, which he couldn't because despite being adults they aren't sexually mature until they migrate to their spawning grounds.
I feel like this explains a LOT about Freud's theories' focus on sexual organs (especially of the eel-shaped variety), in light of... Freud's theories.
https://fediscience.org/@tksst/116119712700119977
Link to source linked by article on the Freud factoid:
https://www.freud.org.uk/schools/resources/freud-the-physician/the-young-anatomist/
Exploring the area before HARRIS with a bunch of cool folks.
Not every day you get to look *down* on a tower crane after climbing a ridiculous number of stairs.
Exploring the Impact of Parameter Update Magnitude on Forgetting and Generalization of Continual Learning
JinLi He, Liang Bai, Xian Yang
https://arxiv.org/abs/2602.20796 https://arxiv.org/pdf/2602.20796 https://arxiv.org/html/2602.20796
arXiv:2602.20796v1 Announce Type: new
Abstract: The magnitude of parameter updates are considered a key factor in continual learning. However, most existing studies focus on designing diverse update strategies, while a theoretical understanding of the underlying mechanisms remains limited. Therefore, we characterize model's forgetting from the perspective of parameter update magnitude and formalize it as knowledge degradation induced by task-specific drift in the parameter space, which has not been fully captured in previous studies due to their assumption of a unified parameter space. By deriving the optimal parameter update magnitude that minimizes forgetting, we unify two representative update paradigms, frozen training and initialized training, within an optimization framework for constrained parameter updates. Our theoretical results further reveals that sequence tasks with small parameter distances exhibit better generalization and less forgetting under frozen training rather than initialized training. These theoretical insights inspire a novel hybrid parameter update strategy that adaptively adjusts update magnitude based on gradient directions. Experiments on deep neural networks demonstrate that this hybrid approach outperforms standard training strategies, providing new theoretical perspectives and practical inspiration for designing efficient and scalable continual learning algorithms.
toXiv_bot_toot
An incredibly important read on inclusive workplace design
"Inclusivity begins with the basics of workplace design. (...) Before you talk about “next-gen workplace experience,” ask yourself: can everyone see, hear, move, breathe and focus here without struggle?"
https://www.leesmanindex.com/arti…
I think it's time for some #FollowFriday action once again. Today, let's focus on computational/algorithmic/generative art/design practitioners (not AI art :)
An entirely incomplete list, in A-Z order:
@…
Remember, using a native HTML `<fieldset>` / `<legend>` gets you better results in SRs than its ARIA equivalent:
https://adrianroselli.com/2022/07/use-legend-and-fieldset.html
My post deals with the inverse of the others I shared — unstyling the nat…
There is a massive focus on trying to get Open Source software to the point where governments or big corporations can use them to "increase sovereignty" (which sometimes means a bit of a weird mix of things).
But that can lead to software being pushed towards more and more complex architectures and deployment strategies (think something like a simple collaborative editor needing Kubernetes to be deployable). While governments and certain companies might be able to shoulder tha…
Yes, Vegas is changing.
Q: If you were a Vegas exec that knew that tourism would drop by ~10% due to Trump's tariffs & intl policies, what would you do?
You'd focus on maximizing revenue from remaining tourism... & minimize operating costs.
And that's EXACTLY what happened.
(1/3)
Tension. Anxiety. Arousal. Hyperarousal. Edgy. Distracted. Fearful.
Deep breathing. Focus on the horizon maybe?
Every Single Participant in NYT Focus Group Preferred Progressive Candidates Over Moderate Ones | Common Dreams
https://www.commondreams.org/news/focus-group-nyt
Mal Freetube auf dem Laptop intsalliert.
Von den Einstellungsmöglichkeiten wird man ja fast erschlagen.
#FreeTube
"TectoTalks" focus on Fold-and-thrust belts and orogens in seismic reflection data #geophysics #geology
"Right now, too many newcomers default to the largest servers. We want to change that - because Mastodon is best when communities are spread across many independent servers, each with its own character and focus." - Mastodon blog
https://blog.joinmastodon.org/2026/02/
Has your favorite album vanished from the music streaming service of choice again? I highly recommend ripping your CDs and making them available through @… with the @… app; it really works great on iOS & macOS!
In the time I've been offline, I've been doing a lot and feeling a lot more mentally healthy. I've been exploring nomadnet a bit, looking at reticulum. I'm definitely going to go back to my break and being online much less regularly.
I actually totally forgot about the anniversary of the shooting, which is the first time that's happened since... uh... the shooting, I think.
I've definitely realized that, on some level, I've definitely used Mastodon (and formerly Twitter) as a coping mechanism, often in order to deal with the stressful things that I've found out about on Mastodon or Twitter.
But, again, none of those things really change our core job: build community. And that's part of what I've been neglecting, and what I can focus on more when I'm not spending as much time talking to people all over the world indirectly. Like, I can just chat directly with folks and talk about this shit.
Yeah, I do think there's value in this community. I don't think it's really screaming into the void (at least, not most of the time). But I know that I need the balance to be way farther on the side of direct engagement with comrades doing and building.
So that's what I'm gonna go back to. I feel as though it's a good sign that with all the writing about getting shot that I've been doing, and all the thinking about that, that the actual anniversary of the shooting I'm actually just thinking about bread.
And that seems like a good note to leave on. I'm gonna go back to some hacker shit.
Statistical Query Lower Bounds for Smoothed Agnostic Learning
Ilias Diakonikolas, Daniel M. Kane
https://arxiv.org/abs/2602.21191 https://arxiv.org/pdf/2602.21191 https://arxiv.org/html/2602.21191
arXiv:2602.21191v1 Announce Type: new
Abstract: We study the complexity of smoothed agnostic learning, recently introduced by~\cite{CKKMS24}, in which the learner competes with the best classifier in a target class under slight Gaussian perturbations of the inputs. Specifically, we focus on the prototypical task of agnostically learning halfspaces under subgaussian distributions in the smoothed model. The best known upper bound for this problem relies on $L_1$-polynomial regression and has complexity $d^{\tilde{O}(1/\sigma^2) \log(1/\epsilon)}$, where $\sigma$ is the smoothing parameter and $\epsilon$ is the excess error. Our main result is a Statistical Query (SQ) lower bound providing formal evidence that this upper bound is close to best possible. In more detail, we show that (even for Gaussian marginals) any SQ algorithm for smoothed agnostic learning of halfspaces requires complexity $d^{\Omega(1/\sigma^{2} \log(1/\epsilon))}$. This is the first non-trivial lower bound on the complexity of this task and nearly matches the known upper bound. Roughly speaking, we show that applying $L_1$-polynomial regression to a smoothed version of the function is essentially best possible. Our techniques involve finding a moment-matching hard distribution by way of linear programming duality. This dual program corresponds exactly to finding a low-degree approximating polynomial to the smoothed version of the target function (which turns out to be the same condition required for the $L_1$-polynomial regression to work). Our explicit SQ lower bound then comes from proving lower bounds on this approximation degree for the class of halfspaces.
toXiv_bot_toot
Spatiotemporal Evolution of the 2022 March Interplanetary #CoronalMassEjection Revealed by Multipoint Observations of Forbush Decreases: https://iopscience.iop.org/article/10.3847/1538-4357/ae1834 -> The path to solar weather forecasts: https://www.u-tokyo.ac.jp/focus/en/press/z0508_00436.html - space-based measurements of solar eruptions are the first of their kind -> Solar storms could be forecast by monitoring cosmic rays: https://physicsworld.com/a/solar-storms-could-be-forecast-by-monitoring-cosmic-rays/
Amid this news madness, I've thoroughly enjoyed being subscribed to the daily #CarbonBrief. Instead of compulsing the news, I try [and often fail] to focus on the most insightful items in my news feeds. This analysis of the latest UK auction for #WindPower was excellent:
#Researchers or other professions who read articles for your work - which do you find better for focus, information processing and memorisation: reading on an electronic medium (computer, tablet, phone...) or on actual paper?
I thought I liked the electronic format but I'm wondering if it makes it harder to focus or remember things..
(This is only for work-based and not leisure…
Backlit Cactus in the Desert Garden. Huntington Library, San Marino, California, USA. December, 2025. OM System OM-1 M.Zuiko 90mm F3.5. #huntingtonlibrary #desertgarden #cactus
Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking
Ravi Ghadia, Maksim Abraham, Sergei Vorobyov, Max Ryabinin
https://arxiv.org/abs/2602.21196 https://arxiv.org/pdf/2602.21196 https://arxiv.org/html/2602.21196
arXiv:2602.21196v1 Announce Type: new
Abstract: Efficiently processing long sequences with Transformer models usually requires splitting the computations across accelerators via context parallelism. The dominant approaches in this family of methods, such as Ring Attention or DeepSpeed Ulysses, enable scaling over the context dimension but do not focus on memory efficiency, which limits the sequence lengths they can support. More advanced techniques, such as Fully Pipelined Distributed Transformer or activation offloading, can further extend the possible context length at the cost of training throughput. In this paper, we present UPipe, a simple yet effective context parallelism technique that performs fine-grained chunking at the attention head level. This technique significantly reduces the activation memory usage of self-attention, breaking the activation memory barrier and unlocking much longer context lengths. Our approach reduces intermediate tensor memory usage in the attention layer by as much as 87.5$\%$ for 32B Transformers, while matching previous context parallelism techniques in terms of training speed. UPipe can support the context length of 5M tokens when training Llama3-8B on a single 8$\times$H100 node, improving upon prior methods by over 25$\%$.
toXiv_bot_toot
Sources: the US Senate's crypto market structure bill is likely to be delayed by several weeks as key lawmakers shift focus to potential housing legislation (Bloomberg)
https://www.bloomberg.com/news/articles/2026-01-21/…
US and Israel considering special forces raids to secure Iran’s stockpile of highly enriched uranium
The US and Israel are reportedly considering special forces operations to seize or secure Iran’s remaining stockpile of highly enriched uranium.
The objective of any special forces operation would be to secure roughly 440–450 kilograms of uranium enriched to 60% purity, which could be converted to weapons-grade material within weeks.
Operations would focus on fortified undergr…
Immersed in the world of big band music and cinematic scores, I discover how these vast compositions can transform a mundane workday. The RIGHT PLAYLIST not only sharpens focus but also ignites innovative ideas by offering fresh perspectives. 🎶 Which soundtrack enhances your creativity? 💼
Allen's hummingbird, Selasporus sasin, standing watch at Sepulveda Basin Wildlife Reserve
#naturalist
SOM-VQ: Topology-Aware Tokenization for Interactive Generative Models
Alessandro Londei, Denise Lanzieri, Matteo Benati
https://arxiv.org/abs/2602.21133 https://arxiv.org/pdf/2602.21133 https://arxiv.org/html/2602.21133
arXiv:2602.21133v1 Announce Type: new
Abstract: Vector-quantized representations enable powerful discrete generative models but lack semantic structure in token space, limiting interpretable human control. We introduce SOM-VQ, a tokenization method that combines vector quantization with Self-Organizing Maps to learn discrete codebooks with explicit low-dimensional topology. Unlike standard VQ-VAE, SOM-VQ uses topology-aware updates that preserve neighborhood structure: nearby tokens on a learned grid correspond to semantically similar states, enabling direct geometric manipulation of the latent space. We demonstrate that SOM-VQ produces more learnable token sequences in the evaluated domains while providing an explicit navigable geometry in code space. Critically, the topological organization enables intuitive human-in-the-loop control: users can steer generation by manipulating distances in token space, achieving semantic alignment without frame-level constraints. We focus on human motion generation - a domain where kinematic structure, smooth temporal continuity, and interactive use cases (choreography, rehabilitation, HCI) make topology-aware control especially natural - demonstrating controlled divergence and convergence from reference sequences through simple grid-based sampling. SOM-VQ provides a general framework for interpretable discrete representations applicable to music, gesture, and other interactive generative domains.
toXiv_bot_toot
The part of the Epstein files we've seen makes it possible to reconstruct Trump's probable orders to his servants at DOJ.
"Just release a few million. Make sure they don't contain anything new about me, but that they cause trouble for a lot of other people. Media will focus on the latter."
Pink Rose. Mary and Joseph Retreat Center, Rancho Palos Verdes, California, USA. August, 2024. Fujifilm X100VI. #ranchopalosverdes #rose #rosegarden
Tesla will end production of its two flagship models, the Model S and Model X, which have long carried the company's prestige.
According to statements by Tesla CEO Elon Musk, the fundamental reason behind this decision is the company's desire to shift its resources and production infrastructure toward autonomous driving and robotics.
Planned to take effect from the next quarter, this production halt has generated significant reverberations throughout the automotive indus…
Via @PinkNews on FB: 'A long-delayed plan to establish the Smithsonian American Women's History Museum has been thrown into uncertainty after House Republicans added a controversial amendment banning transgender women from being represented in the museum.
The provision [requires] the museum to focus solely on “biological women,” while also granting Trump the power to override its proposed location. The amended bill passed along party lines, prompting Democrats to withdraw their support.'
OpenAI plans to unify ChatGPT, Codex, and its browser into a desktop "superapp" to simplify the user experience and focus on engineering and business customers (Berber Jin/Wall Street Journal)
https://www.
In our European energy transition, the focus is now on competitiveness of our industries and on affordability for households. There's a lot of talk about how to split the bill, between companies and citizens, between higher and lower income households, between tax payers and grid users, between us and future generations.
But I think that ..
Pink Rosebud starting to open Descanso Gardens, La Canada Flintridge, California, USA. October, 2025. OM System OM-1 M.Zuiko 300mm F4 PRO. #rose #rosebud #descansogardens
Mark Zuckerberg's original metaverse vision is effectively over, after renaming Facebook to Meta and spending ~$80B on the endeavor, as his focus shifts to AI (New York Times)
https://www.nytimes.com/202…
Partially open pink Rose. Descanso Gardens, La Canada Flintridge, California, USA. December 2025. OM System OM-1 M.Zuiko 300mm F4 PRO. #rose #rosebud #descansogardens
The Gates Foundation says Bill Gates will not deliver his keynote at India's AI Summit after "careful consideration" amid controversy over his Epstein ties (Cherylann Mollan/BBC)
https://www.bbc.com/news/articles/c309qv9zglno
Pentagon Should Focus on Defense Priorities, not Lavish Dinners, After Historic $93.4B "Use-It-or-Lose-It" September (OpenTheBooks)
https://openthebooks.substack.com/p/pentagon-should-focus-on-defense
http://www.memeorandum.com/260310/p39#a260310p39
Sources: Amazon shut down Blue Jay, a multi-armed robotic system launched in October 2025 for same-day delivery warehouses, to focus on small modular warehouses (Eugene Kim/Business Insider)
https://www.businessinsider.com/amazon-pulls-plug-blue-jay-warehous…
OpenAI hires Instagram's VP of global partnerships, Charles Porch, as its first VP of global creative partnerships, seeking to win over a skeptical Hollywood (Julia Black/Vanity Fair)
https://www.vanityfair.com/news/story/openai-hires-charles-porch-instagram
Fei-Fei Li's World Labs raised $200M from Autodesk, part of a larger round; the companies are exploring integrating World Labs' AI models in Autodesk products (Rebecca Bellan/TechCrunch)
https://techcrunch.com/2026/02/18/worl