Recall

Read it later app graveyard? How to import 1,000 saved links into your second brain and actually use them. Watch how Nicole brought her dead links back to life in Recall. She chatted with all of them at once, compared them against the latest research on the web, spotted which ones were out of date, and found connections she'd missed: youtube.com/watch?v=BiVnrMow… The Japanese have a word for buying books and letting them pile up unread. Tsundoku (積ん読). Most of us have the digital version. Years of articles saved in Readwise, Pocket, Raindrop, Instapaper or a bookmarks folder, all waiting for a "later" that never comes. We just launched bulk URL imports, so you can bring up to 1,000 links into Recall at once. But moving links from one app to another isn't the point. The point is finally being able to use them. In Recall, that means three things: 1. You can talk to all of them As soon as the import finishes, every link can be chatted with. You don't have to tag, sort or read anything first. Ask a question and get an answer pulled from everything you've saved, with sources, so you can see which links it came from. 2. You can compare them to what's on the web Saved links get old. Research moves on, advice changes, and the article you saved three years ago might be wrong now. Your pile stops being a time capsule and becomes something you can keep up to date. 3. They get organized for you Once something is in Recall, it's connected to related content you already have. You don't have to build a folder system and hope you remember it later. How to import - Get your links into a CSV file (basically a list of links) - Open Recall and drop the file into "upload or drop anywhere" - Recall pulls them all in and skips anything you already have That's all it takes. A pile you felt guilty about turns into something you can ask questions, check against the web, and find your way around. What would you ask your saved links first? #readitlater #secondbrain

Capital, losses and supply chains. Three takeaways from Bloomberg Tech’s October 27 episode: 1. Meta is boosting its planned equity offering to almost $98 billion as it raises its bet on AI. The overall offering includes several parts, so the headline total should not be read as cash already raised. 2. Anthropic’s IPO filing reveals rapid growth alongside heavy spending and losses. The episode examines how investors are weighing that financial picture as the company prepares to go public. 3. Anduril is teaming up with Qualcomm on a laptop with a US supply chain. Palmer Luckey discusses the manufacturing challenge and the prototype plans, rather than a product already shipping at scale. Share the full episode with Recall for a quick summary, then go deeper where it matters. Source: Bloomberg Technology, October 27, 2026 youtube.com/watch?v=-F4jZfVn… #TechNews #summarizer

Google’s Gemini leader Josh Woodward looks for an “unlearning rate”: how quickly someone changes course when evidence challenges their approach. Test one work habit. Save the interview in Recall and revisit what supports or challenges it. #Learning #SecondBrain

Chip stocks can fall while AI demand rises. Demand, financing costs and expected returns answer different questions. Save the Reuters, ASML and TSMC sources in Recall, then compare the evidence. Sources dated Oct 14–15, 2026. #AI #ChipStocks

Three AI models. The same prompt. Three very different horses. Fireship asked Gemini 4 Argon, GPT-6 Astra and Claude Opus 5.5 to build a realistic interactive 3D horse. The results make a useful point about choosing a model: a benchmark score only tells you about the task being measured. In the comparisons shown in the video: • Gemini leads the long-video-understanding test. • Claude Opus beats Gemini on tasks in a computer terminal. • GPT Astra beats Gemini on the FrontierSWE v2 software-engineering test. Then the horse demo tells a different story. Claude and GPT produce more convincing shapes, while Gemini's result looks more like a blocky game character. One example gives you something to investigate, not a permanent ranking. Start with what you need the model to do. Run the same realistic task across the models you're considering, then compare output quality, the fixes needed and how easily you can use the result. Save the video in your second brain in Recall. When a real task comes up, ask: “Based on this comparison, which model should I try for my task, and why?” Return to the source, check the evidence and build your own recommendation for when to use each model. Source: Fireship, “Did Google just dethrone OpenAI and Anthropic?” (20 October 2026) youtube.com/watch?v=b2tjnxX8… #AIModels #SecondBrain