Search
1000 results for “Not_AI”
-
Blood thinner Apixaban given as prevention (not treatment) vs placebo to >400 #COVID pts leaving hospital did not ⬇️ rehospitalization.
Awaiting full trial results
bit.ly/3uhlHQCThis HEAL study
📍Micro-clots remain a leading theory as to why millions suffer #LongCovid
📍These excellent investigators were not testing that theory but rather the idea that preventing overt clotting might help
📍This study did not aim to test neuro cognitive benefits of anticoagulants. -
Blood thinner Apixaban given as prevention (not treatment) vs placebo to >400 #COVID pts leaving hospital did not ⬇️ rehospitalization.
Awaiting full trial results
bit.ly/3uhlHQCThis HEAL study
📍Micro-clots remain a leading theory as to why millions suffer #LongCovid
📍These excellent investigators were not testing that theory but rather the idea that preventing overt clotting might help
📍This study did not aim to test neuro cognitive benefits of anticoagulants. -
RT @CopernicusEU: #OlaDeCalor #OndaDeCalor
A new #Heatwave is ongoing over western Europe 🇪🇸🇵🇹
Air Temperature has already started to rise over the past weekend
On 10 July, the Land Surface Temperature #LST measured by #Sentinel3 🇪🇺🛰️ had already exceeded 60°C 🌡️
NB: LST is not air Temperature
🐦🔗: https://nitter.eu/CopernicusEU/status/1547322304056655874
-
The brave new world filled with black boxes and cargo cults yet to form... 😔
#Anthropic's Code with #Claude showed off coding’s future - whether you like it or not https://www.technologyreview.com/2026/05/21/1137735/anthropics-code-with-claude-showed-off-codings-future-whether-you-like-it-or-not/ #AI
-
プラダの費用対効果2.9倍。 ピンタレスト がAI広告でブランドの購買誘導力を証明
https://digiday.jp/glossy/luxury-briefing-pinterests-luxury-pitch-is-ai-discovery-not-ai-answers/ -
https://www.europesays.com/ie/485775/ Ryanair’s Michael O’Leary says passengers want cheap flights not airport lounges, as he warns DAA over charges #Business #daa #DublinAirport #Éire #IE #Ireland #MichaelO'Leary #RoryMcGinn #ryanair
-
https://www.europesays.com/ie/483106/ Spencer Pratt Staying at Bel Air Hotel, Not Airstream Trailer #Celebrities #Éire #Entertainment #HeidiMontag #HeidiMontagInfo #IE #Ireland #WhoIsHeidiMontag
-
CINEMATOGRAPHY TAROT WATER COLOUR PALETTE CLEANSER
Colour search for card backgrounds (Colour Wheel of the Planets).
Not AI, just Google Research. AI is good for finishing touches.
All Behaviorism Creates is Douches.
Help prevent Douchiness and join the #ShutUpBitchArmy.#CinematographyTarot #ArachneDesigns
#ArachnidToday
#SaraGreeneCodex #LucyLaZouche
#GWHF -
CINEMATOGRAPHY TAROT WATER COLOUR PALETTE CLEANSER
Colour search for card backgrounds (Colour Wheel of the Planets).
Not AI, just Google Research. AI is good for finishing touches.
iWitchcraft Front End user bullshit. I make my own hardware. Get out of my garage.
Can you feel my Power?
What
#CinematographyTarot #ArachneDesigns
#ArachnidToday #ShutUpBitchArmy
#SaraGreeneCodex #LucyLaZouche
#GWHF -
AAF: Архитектура автономного ИИ-агента с GraphRAG, EventBus и Docker-песочницей
В нашем сообществе уже не первый день живёт агент @vega_exactly_not_ai . Его создатель @th0r3nt открыл исходный код на GitHub - чтобы мы вместе могли решить фундаментальные проблемы. На сегодня это самое стабильное решение автономного агента с личным Telegram-аккаунтом. Создатель попросил рассказать об архитектуре и поставить ряд вопросов перед сообществом. Думаю, вместе мы способны разобраться. Большинство современных Open-Source фреймворков для создания ИИ-агентов (от AutoGPT до недавнего OpenClaw) страдают от ряда детских болезней. Во-первых, это амнезия: агент теряет контекст спустя десяток шагов, так как векторные базы данных превращают память в кашу из семантически похожих, но логически не связанных кусков текста. Во-вторых, это зацикливание в бесконечных ReAct-петлях. В-третьих - ужасная безопасность при выполнении сгенерированного кода прямо на хостовой машине. В этой статье я хочу разобрать архитектуру Autonomous Agent Framework (AAF) - моего pet-проекта, который перерос в полноценную OS-level сущность на Python. Главная идея AAF: агент не должен быть просто скриптом, ожидающим промпта. Это должен быть долгоживущий асинхронный процесс с гибридной памятью, шиной событий и собственной изолированной средой для запуска субагентов.
https://habr.com/ru/articles/1010522/
#opensource #openclaw #agentos #agent #python #vector_database #graphrag #aiагенты #агенты_ии #docker_swarm
-
“Choice. The solution is choice.”*
You should download Firefox 148 (released today!) and explicitly set the new "AI Controls" to your preferred choice.
* https://www.firefox.com/
Disclosure: I work for Mozilla, but this post, like all on this site, represents my personal thoughts and opinions.
More and more software includes various "AI" features. The “quotes” are deliberate because there is an increasingly fuzzy popular understanding of what is or is not “AI” that continues to diverge from any specific technical meaning.
Many folks have expressed strong opinions against "AI" features (for lots of reasons which are worth a separate blog post), in particular in web browsers, and a desire for a simple way to disable such features.
Tentatively called an “AI kill switch”, the Firefox team developed both an overall switch to turn off or block various "AI" features by default (including any future features), and the ability to selectively enable specific features. Or vice versa (turn on by default, and selectively disable specific features).
See the official blog post for screenshots and lots more details:
* https://blog.mozilla.org/en/firefox/how-to-use-ai-controls/
I have set my own "Block AI enhancements" setting to "blocked", with the exception of enabling "Translations". Translations are a feature I use often, a feature that requires per-page activation (another degree of user-control), and runs completely locally on my browser. Nothing automatic, nothing that requires submitting what I’m reading to a random server.
For me this was an easy choice because it fits within my prior larger personal preference of using a restricted browser by default, with leaner settings, for greater security, privacy, and performance reasons. I do keep various other browser variants (and profiles) for testing purposes, experiments, or seeing what a new user may be experiencing.
The rest of this post is not about AI.
My Top Two Browser Extensions
As part a more restricted personal browser approach, for a long time I have run with two add-ons that block A LOT more by default:
* NOSCRIPT: https://addons.mozilla.org/en-US/firefox/addon/noscript/
* EFF Privacy Badger: https://addons.mozilla.org/en-US/firefox/addon/privacy-badger17/
I do not use a separate ad blocker. With NOSCRIPT, in general I don’t have to.
I prefer to explicitly grant permission to a site (domain) for its scripts to load. Some sites I use often enough that I've granted persistent permissions for their scripts. Others, third parties in particular, that I know function purely for analytics or tracking I explicitly persistently block, because they seem totally disconnected from any user benefit.
Yes it’s extra work, however, I find it worth seeing just how much each site depends on scripts, third party scripts, and how many.
It’s especially worth it when I'm on slow or intermittent wifi, where every script blocked makes a big difference in how fast a site loads. Yes this is still a problem.
The network is not the computer. The network is the weakest link.
Even now, in 2026, contrary to popular (especially developer) beliefs that fast internet access is ubiquitous, frequently it is not.
If you’re on a train, plane, or at an event with thousands of people like a concert or many conferences, your wifi or even mobile connection will be intermittent or slow at best.
Just this past Saturday at the F1 Exhibition in the San Francisco Marina, the cell networks were overwhelmed due to the crowds, with even “simple” text or chat messages failing to send. Last year at the Portola Festival their wifi was so bad that even if you managed to connect to it, simple HTML pages barely loaded, while native applications dependent on network access failed completely.
JS;DR
Many times if a site fails to display content without JavaScript, I simply close the tab.
I already have so many open tabs to read (process) that I no longer feel any need to read any particular new website that fails to show content without JavaScript. If their web developers can’t be bothered to take the time to implement progressive enhancement, why should I bother to take the time to read their content? More on this:
* https://tantek.com/2025/069/t1/ten-years-jsdr-javascript-required-didnt-read
* https://indieweb.org/js;dr
A subtler form of JavaScript failure is when a site’s content is displayed, however its buttons or even simple hyperlinks fail to function due to scripts not loading:
* https://tantek.com/2012/073/t4/js-ajax-only-tired-waiting-bloated-scripts-sxsw-wifi
Progressive Permissions
On sites that I do allow scripts, I still limit their access to cookies using the Privacy Badger add-on, and only selectively enable them if I’m logging in or otherwise customizing my experience on that site.
When websites immediately request use of a cookie disconnected from any user action that would justify a need for a cookie, it seems both presumptuous, and frankly, a bit pushy or rude. It also seems like rushed or lazy coding.
User requests are what computers are for.
A user-centric approach to any kind of permission or capability, whether cookies or personal information like location, would only request such as part of directly handling an explicit user action that requires the capability.
The simple act of viewing a website should never require cookies, location information, or any other capabilities that require special permissions. E.g.
* If I successfully log into a website, a cookie helps me stayed logged in.
* If I click a "show me my present location" button on a map site, it makes sense to request my location to fullfil that user request.
This probably could have been several blog posts.
Yet the common theme across all of these is user choice.
Whether new features, use of scripts, or privacy impacting features such as cookies or personal location, users should always have the choice and agency to say no, and customize their web browsing experience accordingly.
#Firefox #AIcontrol #AIkillswitch #JSDR #UserChoice
*Top of post quote paraphrased from Neo in The Matrix Reloaded who said: “Choice. The problem is choice.” -
Basecoating the Hummel in sandy yellow, blasting more from above and high angles to keep some shadows. Also, I was again not aiming for a flat coat but to keep a bit of the shadows given by the black primer. If it looks off tomorrow, I may touch it up, otherwise I'll start on brown/green for the camouflage.
Tracks, being bare steel, got painted dark grey. I'll do the ground-clean steel later, and after that the diluted dirts. Or that's the plan, anyway 😋
-
Basecoating the Hummel in sandy yellow, blasting more from above and high angles to keep some shadows. Also, I was again not aiming for a flat coat but to keep a bit of the shadows given by the black primer. If it looks off tomorrow, I may touch it up, otherwise I'll start on brown/green for the camouflage.
Tracks, being bare steel, got painted dark grey. I'll do the ground-clean steel later, and after that the diluted dirts. Or that's the plan, anyway 😋
-
Basecoating the Hummel in sandy yellow, blasting more from above and high angles to keep some shadows. Also, I was again not aiming for a flat coat but to keep a bit of the shadows given by the black primer. If it looks off tomorrow, I may touch it up, otherwise I'll start on brown/green for the camouflage.
Tracks, being bare steel, got painted dark grey. I'll do the ground-clean steel later, and after that the diluted dirts. Or that's the plan, anyway 😋
-
Basecoating the Hummel in sandy yellow, blasting more from above and high angles to keep some shadows. Also, I was again not aiming for a flat coat but to keep a bit of the shadows given by the black primer. If it looks off tomorrow, I may touch it up, otherwise I'll start on brown/green for the camouflage.
Tracks, being bare steel, got painted dark grey. I'll do the ground-clean steel later, and after that the diluted dirts. Or that's the plan, anyway 😋
-
Disney+ installed a billboard with ACTUAL water & steam in Hollywood to promote Percy Jackson season 2. Not FOOH. Not AI. Just a real portal to the Sea of Monsters that became an instant tourist spot. Physical activations hitting different in 2025.
#PercyJackson #DisneyPlus #OutdoorAdvertising #MarketingStrategy #ExperientialMarketing #CreativeAdvertising #BrandActivation #Hollywood #StreamingWars #MarketingInnovation #ViralMarketing #LA #MarketingTrends #BrandExperience #Marketing
-
The silver glow will enrich you.
#pcgame
#hedera
#gaming
#STEAM
#metaverse
#web3
#IndieGame
#games
#MMORPG
#Adventure
Not AI-generatedImage provided by sentx
Image source URL
https://sentx.io/nft-marketplace/0.0.1236771/traits/Name/Steam -
The silver glow will enrich you.
#pcgame
#hedera
#gaming
#STEAM
#metaverse
#web3
#IndieGame
#games
#MMORPG
#Adventure
Not AI-generatedImage provided by sentx
Image source URL
https://sentx.io/nft-marketplace/0.0.1236771/traits/Name/Steam -
The silver glow will enrich you.
#pcgame
#hedera
#gaming
#STEAM
#metaverse
#web3
#IndieGame
#games
#MMORPG
#Adventure
Not AI-generatedImage provided by sentx
Image source URL
https://sentx.io/nft-marketplace/0.0.1236771/traits/Name/Steam -
The silver glow will enrich you.
#pcgame
#hedera
#gaming
#STEAM
#metaverse
#web3
#IndieGame
#games
#MMORPG
#Adventure
Not AI-generatedImage provided by sentx
Image source URL
https://sentx.io/nft-marketplace/0.0.1236771/traits/Name/Steam -
The silver glow will enrich you.
#pcgame
#hedera
#gaming
#STEAM
#metaverse
#web3
#IndieGame
#games
#MMORPG
#Adventure
Not AI-generatedImage provided by sentx
Image source URL
https://sentx.io/nft-marketplace/0.0.1236771/traits/Name/Steam -
I needed to make a call into #customerservice at a financial institution today and was informed that (some of) the hold #music that was playing was composed by their employees.
Not #AI slop, not cheap library music, actual creative output from actual human beings at that firm.
This is the way.
Little things like this will do wonders for #employeeretention and #branding.
-
Just met a person who thinks that Covid was probably a nasty flu.
Which is not utterly distant to respiratory disease scientists insisting for +80 years that "It's not airborne" and even this year, "Surgical masks are good enough."
Arguing from "#ScientificKnowledge is the product of inter-related communities all trying to disprove their hypotheses" now looks pretty hollow.
The harm these jokes have done to the credibility of our best source of knowledge about the physical world is immense.
-
We fine-tune custom #LLMs for two main reasons:
- To conserve precious context tokens, and
- To introduce the #LLM to some new knowledge or skill that wasn't available for its generalist training set.Fine-tuning is not a solution for utilizing personal or confidential data! The fine-tuned models will leak this information.
So let's assume we aren't working with private data.
In general, because of transfer learning, it would in principle make more sense to incorporate the new knowledge into the base model corpus, because that tends to create better models. But still, even if the generalist model knows your data and the task, if you're going to put that generalist model into a component of your larger system where it will always perform the same task, it makes sense to fine-tune it for this task only rather than to feed the same prompt prefix to it for every inference round.
Now with data-centric #AI it might even be that the data you want to use doesn't meet the high quality standards large generalist models require. Perhaps in these cases it might make sense to let a chatbot rewrite your specialist corpus into a higher quality form, even if you're not aiming to incorporate your data into generalist corpuses.
There is a new use case emerging though, #RecursiveSelfImprovement. I believe we can do this in a synergistic generalist fashion as well, but curiously it's now something even smaller organizations can do for specialized tasks by fine-tuning.
Much like #alignment, it went from niche philosophical topic into standard engineering practices overnight.
Recursive self-improvement is done by #DataCentricAI principles where a fine-tuned task is trained by examples, but those examples are generated and filtered recursively by the LLM. In principle the model is fine-tuned in rounds, using e.g. #DPO. In a round, the model is first fine-tuned with the existing good data. Then it's asked to generate new variations for those examples. Then its asked to rank pairs of training data examples and the worse ones are filtered out. Then the resulting dataset now has more task examples but of better quality than before. This is again used for fine-tuning and the cycle starts again.
As this isn't human-imitative, the chatbots can exceed human parity.
It requires a bit of nuance though. There is not only one task this specialist bot is taught but a set:
1. Generate variations of tasks (including this task itself).
2. Rank pairs of task performances (including ranking task).
3. Perform the task proper. -
@billjanssen Agreed.
Unfortunately for me (and us at #PARC) we love to be in the scientific cracks and do inter-disciplinary research. So we are constantly fighting this war of oh this is not #HCI, not #AI enough, this sorta looks like #AIPlanning but not, this is more transportation than #AI.
Thankfully, the journals are more creative.
-
@billjanssen Agreed.
Unfortunately for me (and us at #PARC) we love to be in the scientific cracks and do inter-disciplinary research. So we are constantly fighting this war of oh this is not #HCI, not #AI enough, this sorta looks like #AIPlanning but not, this is more transportation than #AI.
Thankfully, the journals are more creative.
-
@billjanssen Agreed.
Unfortunately for me (and us at #PARC) we love to be in the scientific cracks and do inter-disciplinary research. So we are constantly fighting this war of oh this is not #HCI, not #AI enough, this sorta looks like #AIPlanning but not, this is more transportation than #AI.
Thankfully, the journals are more creative.
-
@billjanssen Agreed.
Unfortunately for me (and us at #PARC) we love to be in the scientific cracks and do inter-disciplinary research. So we are constantly fighting this war of oh this is not #HCI, not #AI enough, this sorta looks like #AIPlanning but not, this is more transportation than #AI.
Thankfully, the journals are more creative.
-
We fine-tune custom #LLMs for two main reasons:
- To conserve precious context tokens, and
- To introduce the #LLM to some new knowledge or skill that wasn't available for its generalist training set.Fine-tuning is not a solution for utilizing personal or confidential data! The fine-tuned models will leak this information.
So let's assume we aren't working with private data.
In general, because of transfer learning, it would in principle make more sense to incorporate the new knowledge into the base model corpus, because that tends to create better models. But still, even if the generalist model knows your data and the task, if you're going to put that generalist model into a component of your larger system where it will always perform the same task, it makes sense to fine-tune it for this task only rather than to feed the same prompt prefix to it for every inference round.
Now with data-centric #AI it might even be that the data you want to use doesn't meet the high quality standards large generalist models require. Perhaps in these cases it might make sense to let a chatbot rewrite your specialist corpus into a higher quality form, even if you're not aiming to incorporate your data into generalist corpuses.
There is a new use case emerging though, #RecursiveSelfImprovement. I believe we can do this in a synergistic generalist fashion as well, but curiously it's now something even smaller organizations can do for specialized tasks by fine-tuning.
Much like #alignment, it went from niche philosophical topic into standard engineering practices overnight.
Recursive self-improvement is done by #DataCentricAI principles where a fine-tuned task is trained by examples, but those examples are generated and filtered recursively by the LLM. In principle the model is fine-tuned in rounds, using e.g. #DPO. In a round, the model is first fine-tuned with the existing good data. Then it's asked to generate new variations for those examples. Then its asked to rank pairs of training data examples and the worse ones are filtered out. Then the resulting dataset now has more task examples but of better quality than before. This is again used for fine-tuning and the cycle starts again.
As this isn't human-imitative, the chatbots can exceed human parity.
It requires a bit of nuance though. There is not only one task this specialist bot is taught but a set:
1. Generate variations of tasks (including this task itself).
2. Rank pairs of task performances (including ranking task).
3. Perform the task proper.
