As pessoas que nunca apoiaram os artesãos e artistas locais agora estão muito indignadas com a inteligência artificial.

Screego screen sharing for developers #screen #share #tools #web #webapp

The Digital Omnibus proposes to deregulate privacy & AI in Europe. It is a terrible idea. I'm one of the signatories of an open letter to the Norwegian government explaining why we need to keep GDPR, the AI Act and other tech regulation in place. Especially now. www.forbrukerradet.no/siste-nytt/d...

Brands in 2008: We want to be on social media and be more authentic! Brands in 2025: We want to use AI to make better ads for social media!

Double-edged sword, @gregory-ms.com got new training data and became more accurate in finding papers with hope for people with #MultipleSclerosis. Visits to the website are down because those papers don't show up as often. #MSchat

Sobre a fusão FCT/ANI. Há coisas com as quais concordo mais, outras menos E há uma com a qual não concordo NADA: a dupla tutela com a economia Se tb acham q este é um enorme erro, leiam, assinem e partilhem a carta aberta: www.publico.pt/2025/11/15/c... Para assinar: docs.google.com/forms/d/e/1F...

Human-aligned AI models prove more robust and reliable LLMs perform better when we give them a visual representation on how humans classify things. #ai #alignet #google #llm #open-science #open-source #to-read

Lucid Voice - Offline Voice to Text for Mac Offline voice transcription for macOS. Turn speech into perfect text 3x faster with privacy-first AI. No internet required, no subscriptions. #ai #apple #audio #macos #tools #transcribe

Pronto, agora é que ninguém lê Os Maias.

Cada vez gosto mais do LinkedIn pelas razões erradas. A normalização do discurso vazio é a suposta sabedoria quotidiana. Se não os podes vencer, diverte-te!

"Do no evil" used to be their motto. www.404media.co/google-has-c...

When I submitted a paper to AI & Society the email confirmation sent me to a website that told me my language quality score was only 3/10 but RUBRIQ could fix it. I study this stuff, so uploaded my paper to RUBRIQ, which proceeded to introduce scientific fabrication and change scientific terms.

Entrance The art of interacting incorrectly #art #beautifulweb #creativity #horses

Web summit is coming to Lisbon and I just want it to go away.

Mas afinal quem foi a pessoa que ontem não dormiu pq esteve a contar os 30mil raios que nos atingiram?

Remember, remember... Meet us on Wednesday, November 5th at Cerveteca Lisboa for warm smiles and cold drinks! #twittlis begins at 18h and ends when the last person leaves. Join us for a drink and good banter! 😁 www.openstreetmap.org/node/1004078... #lisbon #portugal #meetup

I have been using Claude more often now, and there's quite a difference in culture between it and ChatGPT. Claude feels friendlier and more transparent, and it guides me to concrete tasks. ChatGPT feels like a white box of mystery.

As part of my role in the #ai for higher #education project, I wrote up a tutorial with how I am currently using AI tools to prepare classes. ai-hed.eu/post/ai-acad... It is also my response to those who are pushing back the adoption of AI tools.

Terno não deve ser fato. Não há nada de terno em usar um fato, não há ternura. E fato não devia ser facto porque fica sem afecto ao qual está afecto.

This is a good read if you are working with #AI #Coding Agents. workingbruno.com/notes/smart-... by @[email protected]

When you’re sick and your friends bring you coffee ❤️ www.instagram.com/howaboutcoff...

Estou no ponto em que tenho de dizer às pessoas quando uma app NÃO É inteligência artificial!

New data on the corporate ROI from generative AI from a large-scale tracking survey by my colleagues at Wharton. They found that 75% already have a positive return on investment from AI, less than 5% negative. Also 46% of businesses leaders use AI daily. knowledge.wharton.upenn.edu/special-repo...

Chatbots on websites are becoming as annoying as ads. Answering questions about the site is a pain, customer support is hard to reach because someone fired half the people. Then they incarnate clippy and pop up “I see you are looking at our products, can I help you make a purchase?”. NO!

TLDR; The PSF has made the decision to put our community and our shared diversity, equity, and inclusion values ahead of seeking $1.5M in new revenue. Please read and share. pyfound.blogspot.com/2025/10/NSF-... 🧵

Estamos todos a contar que os computadores fiquem mais fáceis de usar com a IA. Também dizíamos que com a chegada da internet as pessoas iam ficar mais informadas.

I have survived an email that could have been a message on Slack

I don't think teachers and trainers have updated their view of prompting enough. Bigger models are better at figuring out intent, making prompt formulas less important. Reasoners eliminate the value of chain-of-thought prompting, etc Context & communicating goals are now key to getting good results

The AI bubble is popping, ROI is missing, and the Darwin Awards are comedy gold. I’m here for this valley of despair, because real work is happening: time-saving apps, automation that works. gregory-ms.com uses AI to analyze research—real inference and impact. Drop the hype,help people. #ai #science

I just love people who all of sudden enter their creative flow and thank god had their mobile to share on TikTok and Instagram!

O inem não vai mudar de nome. Fonte: a pagina do governo no instagram www.instagram.com/p/DQRD3CiAGU...

I just wanted to read some song lyrics on a website and then I read this: "We and our 980 partners process your personal data...". I don't want to shame the body count, but wtf??

Steven Tyler singing Cry Me a River sounds better than I expected. www.youtube.com/watch?v=S5JV...

It’s a weird world where the fakes are deep and the influencers are shallow.

Wrote up my first impressions of ChatGPT Atlas, OpenAI's new browser - I remain unconvinced by the entire category of "browser agents", the security and privacy challenges still feel insurmountable to me simonwillison.net/2025/Oct/21/...

Os influencers são importantes, as marcas têm tão pouca personalidade que precisam de a alugar a quem tem até demais.

On an individual level, I agree with @ginatrapani.org. On a societal level I expect a lot from hyped ai companies. I don’t want another fancy bot, I want to see AI cure disease and be a co-teacher, a thoughtful cricket on my shoulder that helps me understand. notetoself.studio/post/measure...

I wrote an updated guide on which AIs to use right now, & some tips on how to use them (and how to avoid falling into some common traps) A lot has changed since I last wrote a guide like this in the spring, and AI has gotten much more useful as a result. open.substack.com/pub/oneusefu...

I wish my handwriting wasn’t so cryptic. It would be fun to write blog posts by hand, even if I had to run it through a recognition software.

Ninguém gosta de censura, mas quando é o algoritmo das redes sociais a mostrar coisas bonitas e a ligar-nos à Matrix dos criadores de conteúdo, está tudo bem.

I hope as we move past the first wave of AI criticism ("it doesn't work, all hype") we get a new wave of AI criticism rooted in the acknowledgement that, yes, these systems are very powerful & quite useful and focusing a deep exploration of when AI uses are uplifting and when they are detrimental.

Prompting for video is still something I am learning, but the results are already pretty cute. 😁 @gregory-ms.com

No child should have to go through this.

Quick tip for most LLMs, ask them to respond with a presentation that uses #revealJS for the slides and you'll get something good enough for your next talk. #AI #prompt (Can we bring back the OG #lifehacker hashtag??)

GitHub - anthropics/prompt-eng-interactive-tutorial: Anthropic's Interactive Prompt Engineering Tutorial Anthropic's Interactive Prompt Engineering Tutorial - anthropics/prompt-eng-interactive-tutorial #anthropic #claude #education #github #learn #prompt #to-read #tutorial

Been browsing the #obsidian sub Reddit lately and it’s funny that everyone shares their own system and dashboards, (me included) but no one is talking about their accomplishments with the tool.

Was reading this post by @jilltxt.bsky.social. In a nutshell, there is an intermediary step where an #LLM may use the English language. For me, the regular Portuguese Joe-user, I am wondering if prompting in English gives us better results. #AI jilltxt.net/do-language-...

Claude Code tip: prompt "use sub-agents" and it will do exactly that, firing up multiple parallel sub-agents to achieve a task, each with a fresh context "Write markdown documentation for the context passed to each of the templates in this project. Use sub-agents." simonwillison.net/2025/Oct/11/...