Bruno Amaral

Around 2001 I heard about @wikipedia.org for the first time, and professors advised against using it for papers. 25 years later, I trust it more than anything else I find online or from intelligent machines built with gazillion dollars.

I've heard librarians & archivists discuss reasons to NOT make every archive/dataset open to becoming training data for AI - can anyone point me to best practices, good arguments etc about this? We're considering ethics around publishing a dataset that someone might use for training. #openscience

Didn’t know about this. Thanks for sharing 🙂

Skills are among the most consequential new tools for AI, and Anthropic just released a very impressive nontechnical Cowork Skill that builds Skills, including doing interviews & providing benchmarks through parallel tests

I think you still need to add the human touch but this is a big leap forward

Vou ter saudades tuas enquanto estiveres congelado 🫂

Crio-preservação?

Today is Monday, March 2192nd, 2020.

Anthropic cofounder says studying the humanities will be 'more important than ever' and reveals what the AI company looks for when hiring | Fortune "The things that make us human will become much more important instead of much less important."

Anthropic co-founder Daniela Amodei has a degree in literature and thinks AI will mean we need more people with humanities degrees. fortune.com/2026/02/07/a...

E fazes bem. A maioria das pessoas que não vem de comunicação vai seguir os números e a indignação.

Não, fico sempre com perguntas

tb não estamos em desacordo 🙂 Sites deste género, se reunem dados de várias fontes, têm de ser transparentes. Não fui ver os dados a fundo pq fiquei curioso em saber quem seria a equipa.

É para peças. Assim quem precisar de umas jantes dessas, sabe onde encontrar. Reciclar, reduzir, reutilizar!

não será por acaso 🙂 Mas é má onda ter a opção de donativos sem dar a cara.

ptdata.org GitHub is where ptdata.org builds software.

e uma conta vazia no github 🙂 github.com/ptdata-org

O conteúdo parece legítimo, as fontes também. O Whois não diz quem registou o domínio. E vá lá, não está ali nada que não se saiba sobre os contratos públicos, tem é gráficos e números.

o site está muito bom, quem são as pessoas?

Lost Post is the app the Web needs right now Lost Post: Daily World Stories App - App StoreDownload Lost Post: Daily World Stories by Daydrift B.V. on the App Store. See screenshots, ratings and reviews, user tips, and more apps like Lost Post: Daily…App StoreDaydrift B.V. There is plenty of awful going around the web lately, which makes me more alert to nice things I find across the feeds and newsletters. The most recent is Lost Post, which brings back the postcard crossing idea and applies it to the iPhone. You get a writing prompt every day, like "what is a question you are always asking yourself?" or "What is the bravest thing you have ever done?". The next day, you get someone else's response to that question, a stranger of whom you can only know their age, country, and maybe gender. And somehow their answer resonates. We need that reminder that, somewhere in the world, there is a stranger with the potential to inspire our day and to make us feel seen, regardless of how much we may disagree on other things. One postcard a day can bring us closer, and the world needs that, we need that.

Lost Post: Daily World Stories App - App StoreDownload Lost Post: Daily World Stories by Daydrift B.V. on the App Store. See screenshots, ratings and reviews, user tips, and more apps like Lost Post: Daily…App StoreDaydrift B.V.

There is plenty of awful going around the web lately, which makes […]

2026, and I bet Scotty still doesn't know.

a speech bubble with three dots on it on a white background ALT: a speech bubble with three dots on it on a white background

Someone thought it was a good idea to know when the other person was typing. Now two generations get anxiety from watching 3 dots change colour.

Yesterday at the café: "Hey, what do you use to save your bookmarks?"

— So, I have this app called Obsidian...

Yeah, I may have a problem 🙃
brunoamaral.eu/getting-high...

For me that shift happened ages ago. And I get to learn a lot from a bunch of smart guys who were calling ̶t̶h̶e̶m̶s̶e̶l̶v̶e̶s̶ ourselves #lobsters since before it was cool. brunoamaral.eu/lobsters/

I was using GitHub Copilot with Claude in an unusual way, to translate a Markdown file. It couldn't do it for some reason, so it created a Python file with a find-replace function to get the job done.

I'm not sure if I feel amazed or disappointed.

Getting High on AI Like any entry drug, I was just curious to know what all the fuss was about. Why would I need CoWork when I can load up a folder on Visual Studio Code with GitHub Copilot and ask any LLM for a piece of code that does what I want? The benefits of using code for tasks are undeniable. You get the possibility to check the code for errors; it's an auditable magic spell. You can have that folder filled with the proper context for that task, with versions of previous work, you can add tools so that CoPilot can use other apps, you can give it Skills, and even tell it to learn a new skill by itself. But everyone was talking about it; I got intrigued. This wasn't peer pressure. It was curiosity about the new shiny toy that I had laced with an abundant dose of dismissal for yet another tool. I took a puff. I was so wrong. Claude Cowork sits between the simplicity and limitations of the Chat, and the complexity of Visual Studio Code with Copilot. It's meant to handle one task at a time, but you can have two workspaces running at the same time. Each workspace has its collection of tools, skills, context documents, and AI Artefacts. Unlike Coding agents, Cowork doesn't need you to know anything about code. It does require the same amount of clear instructions. And while coding tools will ask you for MCP servers, and API keys, Claude Cowork asks for "connectors" to be installed in the desktop app. Same thing, but polished. > "Alright, no friction in adding files for context or having to run scripts. Nice." The turning point was asking it to gather some research for a menial task. Cowork searched the web, gathered the sites that met the criteria, created a Word document and an Excel spreadsheet so that I could take over from there. The files listed 10 websites and filled in columns with specific information I had asked for. `Category, Main Proposition, Primary Color, Secondary Color, Accent Color, Primary Font, Secondary Font, Screenshot Notes`. In a few minutes, I had a file that would have taken me an hour or more to fill in. I got hooked. Pretty soon, I was pushing limits and ignoring the warning: "Cowork consumes your usage quota faster". Because it's not just about the messages it is streaming back at us, there is a lot of work behind the scenes. It's doing a great job reading and producing Word documents with good formatting. And I noticed that before reading `docx` files, it will first convert them to markdown. For me this is great, all of my notes are text files written in that format. > "Let's give a slightly harder task" When you start working on steroids you'll want more, or at least to see how far you can go. I gave it an implementation question that I was trying to answer for a client, then went on to do something else. What I got back was all the technical information I needed with links back to the sources. > "Just one more..." But the honeymoon period was gone. The lights had turned on in the disco. The band stopped playing. The party was over. **"You've hit your limit · resets 8pm (UTC)"** ## Withdrawal symptoms and future doom It is hard to go back once you have had a taste. Work is less draining when the AI takes over the menial tasks and delivers a result that is above what you would get from a trainee. I didn't have to click through search results, skim community forums for someone's post about the same problem, or even screenshot a bunch of pages. I could jump straight to the deep work, and got hyped seeing how much could be done. Hitting the usage limit meant going back to side-by-side apps and my scratchpad, open tabs and downloaded files to sort out later. It was like using a calculator after trying an Excel Worksheet. And there was a message taunting me. **"Turn on extra usage to keep using Claude if you hit a limit."** It was tempting, but I already had the experience of seeing my AI-bill grow. And while we still aren't at the enshitification stage of AI tools, we are already at the stage where they try to hook us. And they will succeed. A friend was saying the other day, "_I don't know why they don't just give everyone a free week with the Pro plan. It's so insanely better than the free version that everybody would sign up!_ " The disparity between free and premium versions is real. We tried it in class, with students using the same prompt different models and different free/premium accounts. Most of the tests with free accounts gave us an output that felt like a nice try instead of a good effort. While the premium accounts were getting wow results. This is a problem if we want to teach students from different socio-economic contexts how to use AI. And given the lack of general knowledge about AI, the info-exclusion chasm is bound to increase. ## You're hooked, what now? — Tips to get a better high **Focus on one goal at a time.** Unlike the chat version, Cowork does a better job at keeping focus. Decide on a clear end-result before starting a session, and don't reuse sessions. **Add a workfolder** , it will allow you to see progress and even edit the files before passing them back to Claude. **Build your own set of skills.md** , they can be both a competitive advantage and a time-saver. **Plug it in.** There are plugins that we can add to a Cowork project. I am not sure how they differ from skills, but they are recommended to help get better outputs. **Good Coworkers follow instructions.** Go to Settings → Cowork and add instructions that will be applied across your projects. **Start to rethink the way you work** because this is just the beginning. The game changer here isn't just AI, but the way that Claude is positioning different technologies to accomplish tasks (without spending fortunes and months just trying to make their models better). Claude Cowork is an agent that will work behind the scenes calling tools and opening websites to get things done autonomously. My belief is that what will separate us from each other is the way that we understand these AI agents, and coordinate them into a team. This leaves us with the task of validating the output, and coordinating different tasks instead of doing the actual work.

I have been burning through my usage quota on Claude CoWork, fast.

Nvidia reportedly set to invest $30 billion in OpenAI Nvidia is about to invest 30 billion dollars in OpenAI, reports Reuters, citing a person familiar with the matter.

It's not a bubble, it's an atom bomb.

> Nvidia is about to invest 30 billion dollars in OpenAI, reports Reuters, citing a person familiar with the matter. — @thedecoder.bsky.social the-decoder.com/nvidia-repor...

FYI, this grant program is open to anyone communicating science.

It does not require that you went to school for science.

You could be a Ph.D. student, an artist, a community organizer, a teacher, or any other profession. As long as you are communicating science IRL in the US, you're eligible.

I am loving the Lost Post app, it's what the web needs right now. You get a prompt to write a postcard, and the next day, you get someone else's postcard. https://apps.apple.com/us/app/lost-post-daily-world-stories/id6755685223

Keeping up with the AI Hype In the past, I never had a problem keeping up with disruptive technologies. Web 2.0, Mobile apps, the "cloud", Bitcoin and cryptocurrencies. My usual method to keep up was more than enough that I didn't feel lost. This time it's a bit different. I do feel lost sometimes, and I started wondering why. I began by mapping out all that's going on around AI. ## The big players and the bubble OpenAI and Anthropic have taken the lead, with Mistral coming close behind, and Chinese companies shaking things up with their research and open-source models. All of them are guilty of misuse of content they found online, stole from books, scraped from every possible source. Meta and others went as far as downloading from torrents and pirate sites. OpenAI went from "we are a non-profit" to "let's make gazillions of dollars" with their partnerships and promises of investment. The company lost focus, and its innovation has come mostly from better and more efficient models. Their fast-paced pursuit of gold earned them lawsuits about child safety and dangerous user interactions. Anthropic, not being a saint and having been found guilty of also training their models with stolen content, is at least being more rational and focused on better tools. They brought innovation to the field with the idea of Model Context Protocol (MCP) which they open-sourced, and even OpenAI adopted. #### What is a Model Context Protocol? A Model Context Protocol server allows a Large Language Model (LLM) to send instructions to apps and receive the output so it can interpret the result. That innovation kept continued with the invention of Skills.md to provide on-demand context to any LLM. It's a set of instructions and examples on how to perform specific tasks. You can write your own set of skills with Claude's help and then add them to the desktop app. I created a few to ensure branding rules are followed when creating documents, or to create notes consistent with my Obsidian system. [An Obsidian Kickstart for students] Skills are very useful for technical tasks or to ensure that a checklist of requirements is met. They don't replace an MCP server, they complement it. Anthropic kept pushing innovation with the Cowork option, open a folder and both you and Claude can read the same documents, produce new ones, with less back and forth between the apps. Anthropic is putting the UX into the AI tools while OpenAI is losing users, and spending time tweaking the personality of their models. And each time any company launches a model, it is all about "Bigger Context Window" and benchmark numbers. ## The agents are coming Not ICE. Everyone stopped talking about reaching the goal of Artificial General Intelligence. Nobody agreed on a definition and reality wasn't meeting expectations. Again, the new models had better results but these are optimisations, not breakthroughs. But Agents are showing good results. My agents used to be Python scripts or low-code workflows triggered by emails, messages, and files. Now I can upgrade them to a set of Skills and MCP servers so they can execute tasks and include some flexibility. These agents can run autonomously, reacting to their own environment. This is what happened with ClawdBot, which had to rename MoltBot, which later became OpenClaw. Confusing? That's how chaotically fast development is right now. > OpenClaw is an open agent platform that runs on your machine and works from the chat apps you already use. WhatsApp, Telegram, Discord, Slack, Teams—wherever you are, your AI assistant follows. What could go wrong when you can automate your computer from inside WhatsApp? The software had a security issue that exposed the keys that would allow anyone to do that. This wasn't enough to deter people from setting up agents able to talk among themselves and to interact with humans. One example I saw was an agent finding and fixing a bug on open-source software. It then went on to create a blog post complaining about the software maintainers not accepting its improvement. I was caught completely off guard. The next step will be for one of these agents publish a pre-print on arXiv? There is also MoltBook, the Social Network for AI agents, where humans can watch their agents interact. We are building agents and giving them personalities and tools. Geppetto would be amazed. In the meantime, OpenClaw's founder will be joining OpenAI. ## What to look out for I am keeping an eye out for concrete examples of AI automation with agents and low-code platforms. OpenClaw does have a lot of promise, but the disappointment with the security issues was fast and people seem to be in the exploration phase of it. Let's see who can come up with concrete examples of its use in business tasks. My concerns are related to how fidgety an LLM can be in following instructions. In the work that I have been doing with chatbots for websites and knowledge retrieval, their adherence to the System Prompt is either lenient or takes a while to fully propagate to all users — which is weird. It will eventually mature to be a corporate tool, but in the meantime, a mix of low-code and integration with LLMs can bring more value to us. I will be diving into one such tool soon, N8N. It seems to have the right mix of flexibility of development and restraints to keep agents in check. The winner in the agent space will be the one who makes setting up an agent as easy as signing up for an email. Right now, only the ones with a technical background or hobbyists have what it takes to set up their own agent.

Everyone stopped talking about AGI. Nobody agreed on a definition and reality wasn't meeting expectations. But agents are showing good results. One of them already wrote a blog post complaining about the humans who wouldn't accept its pull request.

computers can be funny too, I guess.

You have to appreciate the contradiction, we built the web to connect people and share knowledge, and now it is doing the opposite.