Emmanuel Marques 🎗️

Enjoyed my first ride in a CyberCab! #WeRobot

Hello, Moon. It’s great to be back. Here’s a taste of what the Artemis II astronauts photographed during their flight around the Moon. Check out more photos from the mission: nasa.gov/artemis-ii-mul…

We're going farther than ever before 🚀 Today, the Artemis II crew will break the record for how far humans have traveled from Earth as they fly around the far side of the Moon. Coverage begins at 1 p.m. EDT (1700 UTC). Watch Artemis II make history: nasa.gov/ways-to-watch/

One last look at Earth before we reach the Moon. This view of the Earth was captured on April 5, the fourth day of the Artemis II mission, from inside the Orion spacecraft. The four astronauts will reach their closest approach of the Moon tomorrow, April 6.

There are no words.

We see our home planet as a whole, lit up in spectacular blues and browns. A green aurora even lights up the atmosphere. That's us, together, watching as our astronauts make their journey to the Moon.

SpaceX now has over 10,000 Starlink satellites in orbit, which is honestly an absurd engineering achievement. And no, they are not just up there freelancing and hoping for the best. They stay separated because they are placed in organized orbital lanes, constantly tracked, and able to maneuver when needed. Starlink also uses automated collision-avoidance systems, which is how a constellation this large can operate without turning low Earth orbit into a scrapyard. It’s already the closest thing in the world to a true work-anywhere network and they are just getting started.

One of my favorite upgrades to Claude Cowork right now: Projects Here's the problem it solves: By default, every new thread starts blank Claude has no memory of what you were working on, what other threads figured out, or what your instructions were So you'd re-explain yourself at the start of every session Or your agents would overlap, contradict each other, or miss context that another thread already had If you've been following my Cowork setup guides, you've already solved part of this — local files and a CLAUDE.md get you most of the way there Projects takes it the rest of the way Shared memory across every thread in the project Permanent instructions that apply everywhere — set once, always on Scheduled tasks that prep your work before you open the app One project HQ to launch every agent from The day-one prompt I paste into every new project: "I just created this project. Read every file in the folder. Then summarize what you know about this workspace — what's here, what I probably use it for, and what instructions you'll follow. If anything is unclear, ask me some questions." Run that once and Claude builds its own understanding of your project workspace from scratch Full video with my live build, including my 4 tips for getting Projects right, and where projects still lack. Send to someone whose ready to Do More with AI, or bookmark to implement this weekend!

so... @NYTimesPR, what are you doing to uncover government fraud with all your editorial resources?

LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So: Data ingest: I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them. IDE: I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides). Q&A: Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale. Output: Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base. Linting: I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into. Extra tools: I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries. Further explorations: As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows. TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

🚨 BREAKING: Claude can now build your personal brand like Seth Godin built a $100M empire with zero ads (for free). Here are 6 insane Claude prompts that engineer your complete fame system. (Save for later)

5 Steps to Build an #AI Driven Organization by @ingliguori #GenAI #ArtificialIntelligence #MachineLearning #ML

How to set up Claude so it never forgets you: Prompts → Projects → Skills (explained in 3 mins) Prompts = telling a stranger your job every morning. Projects = giving a new hire a binder on day one. Skills = training an employee once. For forever. Step 1: Start with a Prompt (but don't stay there) ✦ Open Claude. Type your task. Get an answer. ✦ It works. But tomorrow? Claude forgot everything. ✦ You re-explain. Again. Every. Single. Chat. ✦ That's Level 1. Most people never leave it. Step 2: Move to a Project ✦ Go to Claude .ai → Create a Project. ✦ Upload your voice file. Upload your instructions. ✦ Now every chat inside that Project knows you. ✦ Your context, style, and tone stick. But you still have to open the right Project. You still have to say "read my file first." Step 3: Graduate to Skills ✦ Open Claude Cowork. ✦ Select Opus 4.6 + Extended Thinking. ✦ Prompt: "Use the skill-creator to help me build a skill for [your most repeated task]." Claude interviews you. Answer extensively. "I write reports" is useless. "I write weekly reports that start with the headline metric, 3 sections max, next steps as bullets" is a Skill. The specificity is the skill. Step 4: Install and test ✦ Save the Skill folder. ✦ Go to Settings → Capabilities → Skills → Upload. ✦ Open a new chat. Type your task normally. ✦ The Skill fires on its own. No slash command. ✦ Claude just knows. I just wrote my full Claude Skills breakdown. It covers setup, the skill-creator walkthrough, and the 7 hacks I found buried in Anthropic's docs. Read it here: ruben.substack.com/p/claude-skills To download all of my Claude infographics: Step 1. Go to how-to-ai.guide. Step 2. Subscribe for free. Don't pay anything. Step 3. Open my welcome email (most skip this). Step 4. Hit the automatic reply button inside. Step 5. Download my infographics from my Notion. ♻️ Repost this to help someone on your team stop re-explaining themselves to Claude every morning.

NEWS: OpenAI just announced that it has officially closed their latest funding round with $122 billion in committed capital at a post money valuation of $852 billion. "We are now generating $2B in revenue per month. At this stage, we are growing revenue four times faster than the companies who defined the Internet and mobile eras, including Alphabet and Meta. ChatGPT has more than 900 million weekly active users, and over 50 million subscribers. Search usage has nearly tripled in a year, and our ads pilot reached more than $100 million in ARR in under six weeks. Momentum is just as strong on the enterprise side, which now makes up more than 40% of our revenue, and is on track to reach parity with consumer by the end of 2026. GPT‑5.4 is driving record engagement across agentic workflows. Our APIs now process more than 15 billion tokens per minute. Codex now serves over 2 million weekly users, up 5x in the past three months, with usage growing more than 70% month over month."

anthropic's growth marketer just showed how he runs all of their paid search as a one-person team. he built a cowork plugin that manages every google ads campaign from a single chat window. here's his workflow: 1. he writes his paid search playbook as a skill. > how he filters search terms > what he considers wasted spend > how he categorizes bad terms (wrong intent, competitor brand, research-only, too generic) claude reads this before doing anything so it makes decisions the same way he would. once claude has the playbook, it needs somewhere to apply it. 2. he connects claude to his google ads account via mcp (a pipe that lets claude pull live campaign data from the api). now claude can see every campaign, keyword, search term, and how it's performing. and it already knows what to look for. 3. he runs the skill. claude goes through every search term across his campaigns, scores each one against his playbook, and flags the ones wasting money. the output is a spreadsheet where every row shows what was flagged, the performance data behind it, and a written explanation of why claude thinks it should be cut. 4. he reviews the reasoning, agrees or overrides, and then instead of manually uploading changes to google ads... he tells claude to add the blocked terms directly to google ads for him each change still requires his explicit approval. nothing gets modified without his sign-off. and because the whole thing is a cowork plugin, it works from his phone too. he runs it from dispatch (claude's mobile app). pulls impression share (how often his ads actually show up vs how often they could) by day, gets budget recommendations, applies them. things the literal google ads mobile app doesn't even support. he basically turned his personal paid search playbook into a system that runs itself.

Everyone will adopt AI. Everyone will build more, ship faster, and deploy agents. That is table stakes. The founders who win will be the ones who make better choices, which requires judgment. (And if you're in the bottom left quadrant, get out quick.)

Technion researchers Dr. Gil Shamai, Prof. Ron Kimmel (Computer Science), and Prof. Dvir Aran (Biology) have developed an AI model that predicts who will benefit from chemotherapy—using routine pathology images, in minutes. A step toward faster, more accessible personalized cancer care. #Technion #AIinHealthcare #BreastCancer #PrecisionMedicine technion.ac.il/en/blog/articl… thelancet.com/journals/lanon…

OpenAI just exited the video generation business entirely. App dead. API dead. No video inside ChatGPT. Disney’s $1 billion deal, signed four months ago, is dead. Read that again. This isn’t a consolidation into the super app. Altman told staff Tuesday that OpenAI is winding down all products using video models. Disney’s own statement says they respect OpenAI’s decision to “exit the video generation business.” The Sora research team is being redirected to robotics. The reason is sitting right there in the competitive data. Anthropic hit $19 billion in annualized revenue by early 2026 selling text and code. No video generation. No image generation. No consumer social app. No Disney deal. One product surface: chat, code, computer use, all in one place. OpenAI looked at where every dollar of market growth was coming from and saw the answer: coding and enterprise. So now they’re copying the model. ChatGPT, Codex, and the browser merge into one app. Instant Checkout killed today too. Every consumer experiment is getting cut. What remains is the Anthropic playbook: one app, code and chat, enterprise and developer focus. The Sora numbers explain the urgency. Total consumer revenue across iOS and Android since September: $1.4 million. Peak month was $540,000. Every video generation burned GPU compute that could have been running inference for ChatGPT or Codex instead. OpenAI’s own head of Sora announced generation limits because chips couldn’t keep up. At $14 billion in projected 2026 losses, every GPU matters. Google just inherited the AI video market by default. Nano Banana already lives inside Gemini. No standalone app to manage, no separate brand to support. Among the majors, they’re the only ones left. Runway, Kling, Minimax, Luma, and the other independents are still shipping, but none of them have Google’s distribution. Disney put $1 billion in stock warrants on a product that lasted six months. The deal was announced in December. Characters from Marvel, Pixar, and Star Wars were supposed to be generating fan videos on Sora by now. Instead, Disney is writing a polite press statement about “respecting OpenAI’s decision” while its legal team unwinds a deal that never produced a single licensed video. Four months from billion-dollar partnership to obituary. That’s how fast the AI product landscape reprices when the unit economics don’t work.

🚨To every Black person tempted by the antisemitic wave sweeping right now:🚨 Back when white mobs lynched us in the streets... When we had no money, no power, and literally no movement... Who stood with us? Who risked everything? The Jews. They weren't just allies—they were family in the fight. - Jewish leaders co-founded the NAACP in 1909 and funded/led it for generations when no one else would. - When Dr. King and our people were arrested, beaten, jailed—Jewish lawyers defended us for free, and Jewish money paid bail bonds and court fees while others turned away. - In Freedom Summer '64, half the white volunteers risking death to register Black voters? Jewish. - Rabbi Heschel marched arm-in-arm with MLK in Selma—our struggles linked forever. Without their blood, treasure, solidarity, and courage—there is no Civil Rights Movement as we know it. No Voting Rights Act. No dream realized. That's why I stand with the Jews. They were our only true friends when this country hated us—when doors slammed shut and ropes hung high. They showed up, bled, and built with us. So to my people joining this demonic, divisive hate today: Pause. Remember. Honor the alliance that freed us. Our histories are bound. Our freedoms were won together. Betray that, and you betray the very martyrs who died for our rights. Black-Jewish unity forged the dream. Don't let hate destroy it. I stand with my Jewish brothers and sisters—always. ✊🏾🤝✡️🇮🇱 Repost if truth > trends. If gratitude > division. If history matters. #BlackJewishAlliance #RememberOurAllies #StandTogether #CivilRightsTruth #NoToAntisemitism