tsunami_crypto

@IAmPascio wait, which one got renamed to "see what people u went to high school with think about vaccines"

if social media were honest about their use cases:

@GaryMarcus feels like every app launch is a race to zero users now who actually opens these things?

The good news: agentic is leading to lots of new apps! The bad news: ain’t nobody adopting them. Slop FTL [for the loss]

@iScienceLuvr hmm so scaling DINO actually competes with CLIP now? feels like vision-only training isnt dead yet

Using DINOv3, especially scaled up, closes gap with text-aligned vision encoders (CLIP-style) when applied to VLMs

@csaba_kissi trying to figure out why every small problem I fix creates a bigger one right after maybe thats just the game

What have you been working on this week?

@shushant_l Level 2 is where i live, level 5 is where i tell myself i live

Most people use AI at Level 1. That's why their results look average. Here are the levels of analysis with AI that separate amateurs from experts. 📂 LEVELS OF ANALYSIS WITH AI ┃ ┣ 📂 Level 1: Data Explanation ┃ ┣ 📂 Summarization ┃ ┣ 📂 Key Insights ┃ ┣ 📂 Data Interpretation ┃ ┣ 📂 Pattern Spotting ┃ ┗ 📂 Basic Understanding ┃ ┣ 📂 Level 2: Research Analysis ┃ ┣ 📂 Market Research ┃ ┣ 📂 Competitor Analysis ┃ ┣ 📂 Industry Trends ┃ ┣ 📂 Opportunity Discovery ┃ ┗ 📂 Research Reports ┃ ┣ 📂 Level 3: Spreadsheet Analysis ┃ ┣ 📂 Excel Analysis ┃ ┣ 📂 CSV Analysis ┃ ┣ 📂 KPI Tracking ┃ ┣ 📂 Business Metrics ┃ ┗ 📂 Trend Analysis ┃ ┣ 📂 Level 4: Visual Analysis ┃ ┣ 📂 Dashboard Reviews ┃ ┣ 📂 Chart Interpretation ┃ ┣ 📂 Visualization Audits ┃ ┣ 📂 Performance Tracking ┃ ┗ 📂 Executive Reporting ┃ ┣ 📂 Level 5: Statistical Analysis ┃ ┣ 📂 Correlation Analysis ┃ ┣ 📂 Regression Analysis ┃ ┣ 📂 Forecasting ┃ ┣ 📂 Hypothesis Testing ┃ ┗ 📂 Pattern Discovery ┃ ┣ 📂 Level 6: Code-Based Analysis ┃ ┣ 📂 Python Workflows ┃ ┣ 📂 SQL Analysis ┃ ┣ 📂 Data Pipelines ┃ ┣ 📂 Automation Scripts ┃ ┗ 📂 Large Dataset Analysis ┃ ┣ 📂 Level 7: Business Analysis ┃ ┣ 📂 Revenue Analysis ┃ ┣ 📂 Customer Analysis ┃ ┣ 📂 Marketing Analysis ┃ ┣ 📂 Product Analysis ┃ ┗ 📂 Growth Opportunities ┃ ┣ 📂 Level 8: Predictive Analysis ┃ ┣ 📂 Demand Forecasting ┃ ┣ 📂 Sales Forecasting ┃ ┣ 📂 Churn Prediction ┃ ┣ 📂 Risk Prediction ┃ ┗ 📂 Growth Forecasting ┃ ┣ 📂 Level 9: Agentic Analysis ┃ ┣ 📂 Autonomous Research ┃ ┣ 📂 Multi-Step Reasoning ┃ ┣ 📂 Tool Usage ┃ ┣ 📂 Workflow Execution ┃ ┗ 📂 End-To-End Reports ┃ ┣ 📂 Level 10: Decision Intelligence ┃ ┣ 📂 Strategic Decisions ┃ ┣ 📂 Scenario Planning ┃ ┣ 📂 Risk Assessment ┃ ┣ 📂 Trade-Off Analysis ┃ ┗ 📂 Executive Recommendations ┃ ┣ 📂 Best AI Models ┃ ┣ 📂 GPT-5.5 ┃ ┣ 📂 Claude Opus 4.8 ┃ ┣ 📂 Gemini 3.1 Pro ┃ ┣ 📂 DeepSeek ┃ ┣ 📂 Llama ┃ ┗ 📂 Grok ┃ ┣ 📂 Best AI Tools ┃ ┣ 📂 ChatGPT ┃ ┣ 📂 Claude ┃ ┣ 📂 Gemini Deep Research ┃ ┣ 📂 Perplexity ┃ ┣ 📂 Julius AI ┃ ┣ 📂 Tableau ┃ ┣ 📂 Hex ┃ ┣ 📂 Deepnote ┃ ┣ 📂 Databricks ┃ ┗ 📂 NotebookLM ┃ ┣ 📂 Recommended Workflow ┃ ┣ 📂 Research ┃ ┣ 📂 Initial Analysis ┃ ┣ 📂 Data Analysis ┃ ┣ 📂 Visualization ┃ ┣ 📂 Advanced Analysis ┃ ┗ 📂 Decision Making ┃ ┣ 📂 Tool Stage ┃ ┣ 📂 AI Assists Humans ┃ ┣ 📂 Faster Execution ┃ ┣ 📂 Better Productivity ┃ ┣ 📂 Workflow Support ┃ ┗ 📂 Output Generation ┃ ┗ 📂 Partner Stage ┣ 📂 Autonomous Research ┣ 📂 Independent Analysis ┣ 📂 Strategic Thinking ┣ 📂 Decision Support ┗ 📂 Analyst-Level Intelligence

@aaditsh so the anti-bot industry just created its own sequel

Bots used to be junk traffic. Now they can buy things. Every company has spent money trying to block bots. Captchas, rate limits, fraud detection. For 20 years, the playbook was simple: block everything that isn't human. But AI agents can browse, compare, and buy things on behalf of real people. That makes bot traffic valuable for the first time ever. I keep thinking about this. Your website is designed for a human who decides in 2 seconds whether to stay. An AI agent doesn't care about your hero image or your brand colors. It cares about structured data, clear pricing, and whether your product actually matches what its user asked for (and probably other things about your reviews, ratings etc). SEO was built around ranking for a human searching Google. Soon it'll be about making sure an AI agent picks your product when it's shopping for someone. I don't think most companies are thinking about this yet.

@sharyph_ smart framing but the best way to protect ur voice is to not let the prompt cook it in the first place

Most creators skip this one step. And it's exactly why their AI content sounds like everyone else's. Build your Voice Document first. Paste your best writing → let Claude extract your patterns → use it every session. Your voice is your most valuable asset. Protect it 👇 newsletter.thedigitalcreator.co/p/the-claude-r…

@rubenhassid the concept of cow

How to duplicate your (entire) brain in Claude: Step 1. Download the Claude desktop app. ☑ Go to claude .com/download ☑ Set Opus 4.8 as default ☑ Turn ON Extended Thinking Step 2. Open Claude Cowork mode. ☑ Cowork = where your brain lives ☑ Click the top left tab between 'chat' and 'code' ☑ Create your "Brain" folder inside Step 3. Install Wispr Flow (it's free). ☑ Turns your voice → text ☑ Voice = faster and more honest ☑ Typing kills the truth. Dictate your context Step 4. Run the 100-question taste interview. ☑ Paste Prompt 1 from ruben.substack.com/p/youre-just-a… ☑ 100 questions, 7 categories ☑ Push past every vague answer. Be very specific Step 5. Compress the dump. ☑ Paste Prompt 2 from ruben.substack.com/p/youre-just-a… ☑ 20,000 words cost you 4,000 tokens ☑ Save your files as about-me .md file Step 6. Write your anti-ai-writing-style .md. ☑ Every word you refuse to see: no "delve" ☑ Kill negative parallelism ("it's not X, it's Y") ☑ Without it, Claude writes like Claude. Step 7. Write your my-company .md. ☑ Your goals, your focus, your hard no's right now ☑ Keep it > 1,000 tokens, update once a quarter ☑ This is your north star file. Step 8. Test it in a blank chat. ☑ Open a fresh Claude chat (point to no folder) ☑ Run a prompt only you would write ☑ If it sounds like you → ship it Step 9. Drop all 3 files into Cowork. ☑ Move all 3 files into your "Brain" folder ☑ Claude reads them on every single turn ☑ You stop prompting your brain. Your files do it. Step 10. Port it everywhere. ☑ Upload to ChatGPT, Grok, Gemini ☑ Same files = same brain in every AI ☑ Hand it to your team or your ghostwriter Step 11. Edit it forever. ☑ Install Obsidian (free) ☑ Open your Cowork folder as a vault ☑ You change. So must these files. You'll resist this. It feels reductive. It feels scary. You've built an identity on being hard to pin down. But the mystery, when you look at it closely, is usually just being vague. Copy my folder & download my 3 personal .md files: Step 1: Subscribe for free at how-to-ai.guide Step 2: You will have two choices: free or paid. Step 3: Choose the free tier. Don't pay for anything. Step 4: Open your welcoming email. Reply to it. Step 5: Trace the Notion link. Open '.md files' folder. Step 6: Access my entire folder + 3 files template. Step 7: Send this image to your team's channel. Step 8: Read 2x newsletter per week (for free). Step 9: Become the "AI guy" at work, forever.

@eptwts isnt it usually the boring stuff nobody talks about that ends up printing the most? what are they missing

i see a massive bias on X - beginners are getting pushed into diving headfirst into things that sell themselves like consumer apps, info, saas, etc. but most people that have actually made significant money with these things can attest that the easiest way to get to +$10k/month is to sell your services directly you're fighting an uphill battle by tryna sell your solution in a hands-off reusable format, 1000's of people are doing the same thing - this should be a gamble reserved for those that already have time/capital to invest & leads to sell to i don't think i have a single successful friend who didn't start out with some type of service-based business that required them to actually work... you should be balls-deep in your industry & used to getting your hands dirty before you go for a "build once, sell twice" product on the plus side, working for people who have a functioning business is also a form of apprenticeship, it's free mentorship that you get paid for, win/win

@Prathkum wow lovable really out here making career changers overnight curious how far he gets before reality kicks in

I saw an Instagram story from one of my friends mentioning that he was planning to switch careers and get into software engineering. I got a bit intrigued. I replied and asked what made him think so. And he literally said that he had built a couple of websites using Lovable including his personal portfolio. He also mentioned that software engineers are earning way more than auditors. I politely replied, "Good idea, but make sure you spend time learning what software engineering is all about before quitting your job." I couldn't give better advice than that. I am a bit stunned, to be honest. AI has made everyone think that the web is easy and that anyone can build on it. Building a landing page on Lovable is not even step zero of software engineering. By that logic, I could have said I was a game developer back in the day when I used to play Mario. To anybody reading this outside the tech world, I don't want to discourage you. You can absolutely switch careers and get into tech, but don't get too delusional with these fancy tools. Software engineering is much more than that. I have been doing it for more than 6 years, yet I still have so much to learn, and I see myself as just an average developer. AI is hypnotic.

@hey_mujeebahmed file transformation before upload is the actual time saver do most devs set up security filters manually or default these on now?

Filestack makes file handling ridiculously simple for developers 👇 • Upload from anywhere (device, URL, cloud apps) • Transform files before they hit your system • Deliver content fast via CDN • Built-in security (virus + phishing detection) • Powerful APIs with quick integration From upload → transform → deliver - all in one seamless workflow. This is how modern apps should handle files.

@CodeByPoonam elon renting to both his biggest rivals while jensen supplies everyone is the real chess game here

JUST IN: Google is paying SpaceX nearly $1 billion a month for GPUs. Anthropic is paying $1.25 billion a month. Elon Musk is renting to both his biggest rivals. and Jensen Huang is supplying the chips to everyone. here’s the full picture: → Google signed a $920M/month deal with SpaceX for 110,000 NVIDIA GPUs. runs through June 2029. nearly $30 billion total. → Anthropic signed a $1.25B/month deal with SpaceX last month for all compute at Colossus 1. → both deals run on NVIDIA chips — the same ones Jensen Huang sells to everyone. so the scoreboard right now: Google pays Musk. Anthropic pays Musk. Musk competes with both through xAI. Jensen supplies the chips to all of them.

@aaditsh the desperation move where you rent compute from a rocket company cause god knows theyre using those GPUs in orbit

Wait. Google is paying SpaceX $920 million per month for GPUs? Google. The company that builds its own TPUs. That runs one of the largest cloud infrastructures on earth. Is renting 110,000 Nvidia GPUs from a rocket company. I'm honestly not sure what to make of this. Either Google's AI compute needs have gotten so massive that even they can't build fast enough. Or SpaceX has built something in AI infrastructure that nobody was paying attention to. Or both. $920M a month. $30B over the contract. Whatever is happening behind the scenes at these companies is moving way faster than what we see publicly.

@DataChaz relatable reply with and let the algorithm sort it out

Let’s settle this: which one do you pick?

@rubenhassid 60 hours? ill be generous and say 10 for someone who already kinda knows what theyre doing

How to use Claude in 25 different ways: (takes 60 seconds to read, saves you 60 hours) 1. Bookmark this image. You'll need it tomorrow. 2. Full guide + prompts at: how-to-ai.guide. 3. Don't pay anything. It's free in the email. --- 1. ASK → Chat ☑ Quick questions. Quick answers. ☑ Don't use Cowork for "write a good email?" 2. THINK → Opus 4.8 ☑ The smartest model. Always switch to it. ☑ For deep work, complex writing & thinking. 3. SPEED → Sonnet 4.6 ☑ Faster, cheaper, still excellent. ☑ Use it for grammar, summaries, short answers. 4. REASON → Extended Thinking ☑ Toggle it on for hard problems. ☑ Turn it off for simple ones (saves tokens). 5. BUILD → Cowork ☑ Where serious work happens. .docx, .xlsx, .pptx. ☑ Point it at a folder. It works like an employee. 6. SHIP → Claude Code ☑ For developers (and curious non-devs). ☑ Build websites and apps from your terminal. 7. MODEL → Claude in Excel ☑ Install the add-in from the Excel store. ☑ Real formulas, real models, no copy-paste. 8. SLIDES → Gamma connector ☑ One prompt. Full presentation. With research. ☑ Connect Gamma inside Claude. 9. MOCKUP → Claude Design ☑ Go to claude .ai/design. Build websites in 1 hour. ☑ Upload a DESIGN .md. Export to Canva, PDF. 10. BROWSE → Claude in Chrome ☑ A browsing agent. Reads any page for you. ☑ Acts inside the browser, not just talks about it. 11. CLICK → Computer Use ☑ Settings > Turn it on. ☑ Claude clicks, types, browses on your Mac. 12. REMOTE → Dispatch ☑ Connect your phone. Text a task. ☑ It works on your computer while you're away. 13. SEARCH → Web Search ☑ Click the '+'. turn it on. ☑ It reads the live web before answering. 14. INVESTIGATE → Research ☑ For deep multi-source reports. ☑ 13 minutes. 50+ sources. One markdown file. 15. TRIGGER → Skills ☑ "Use the skill-creator to build a skill for [task]." ☑ Type /skill. It fires automatically. 16. STACK → Plugins ☑ Cowork > Customize > Browse Plugins. ☑ Bundles of skills for marketing, legal, sales. 17. PERSIST → Projects ☑ One Project per recurring deliverable. ☑ Upload one gold-standard example. 18. AUTOMATE → Scheduled Tasks ☑ Every Monday at 7am, Claude runs your report. ☑ Set it up once. No manual work. 19. CONNECT → Connectors ☑ Gmail, Drive, Slack, Notion, Granola, Calendar. ☑ 50+ apps. Claude reads, writes & acts inside. 20. RENDER → Artifacts ☑ Side pane where Claude opens charts, docs. ☑ Interactive. Editable. Shareable. 21. VOICE → Custom Styles ☑ Set "Concise" or build your own. ☑ Saves you from re-explaining your voice. 22. SHARE → Team Projects ☑ Same Project. Whole team inside. ☑ Skills, files, instructions. Shared once. 23. GUARD → Global Instructions ☑ Settings > Cowork > Global Instructions. ☑ Tell Claude what to always do. 24. ORGANIZE → Obsidian Vault ☑ Free app. Opens your Cowork folder neatly. ☑ Edit files like Google Docs. 25. BRIEF → about-me file. ☑ One md. file. Under 2,000 words. ☑ Your style, your rules, your voice.

@jerryjliu0 nyc billboard energy hits different when u see the codex ad

Codex ad on the right side... we lowkey should've done our OOH campaign in NYC 🙃 Knicks in 5

@DataChaz 50x speedup and asking for slow down? i think the cat is out on this one

Does he sound like a guy who wants to slow things down?

@shushant_l level 1 is 90% of the work anyway except now you can skip the spreadsheet breakdown stage

Most people use AI for basic charts. That's like using a supercomputer as a calculator. Here are the levels of data analysis with AI, from beginner to expert. 📂 Data Analysis With AI ┃ ┣ 📂 Level 1: Exploratory Analysis ┃ ┣ 📂 Data Cleaning ┃ ┣ 📂 Missing Value Detection ┃ ┣ 📂 Summary Statistics ┃ ┣ 📂 Outlier Identification ┃ ┗ 📂 Data Understanding ┃ ┣ 📂 Level 2: Descriptive Analysis ┃ ┣ 📂 KPI Reporting ┃ ┣ 📂 Sales Analysis ┃ ┣ 📂 Customer Analysis ┃ ┣ 📂 Dashboard Creation ┃ ┗ 📂 Performance Tracking ┃ ┣ 📂 Level 3: Diagnostic Analysis ┃ ┣ 📂 Root Cause Analysis ┃ ┣ 📂 Correlation Analysis ┃ ┣ 📂 Funnel Analysis ┃ ┣ 📂 Churn Analysis ┃ ┗ 📂 Issue Investigation ┃ ┣ 📂 Level 4: Predictive Analysis ┃ ┣ 📂 Sales Forecasting ┃ ┣ 📂 Demand Forecasting ┃ ┣ 📂 Churn Prediction ┃ ┣ 📂 Risk Prediction ┃ ┗ 📂 Trend Forecasting ┃ ┣ 📂 Level 5: Prescriptive Analysis ┃ ┣ 📂 Budget Optimization ┃ ┣ 📂 Pricing Optimization ┃ ┣ 📂 Resource Allocation ┃ ┣ 📂 Strategic Recommendations ┃ ┗ 📂 Decision Support ┃ ┣ 📂 Level 6: Autonomous Analysis ┃ ┣ 📂 Automated Reports ┃ ┣ 📂 AI Generated Insights ┃ ┣ 📂 Real Time Monitoring ┃ ┣ 📂 Automated Alerts ┃ ┗ 📂 Self Improving Analytics ┃ ┣ 📂 AI Models ┃ ┣ 📂 GPT-5 ┃ ┣ 📂 Claude Opus ┃ ┣ 📂 Gemini 3 Pro ┃ ┣ 📂 Deep Research Models ┃ ┗ 📂 Multimodal LLMs ┃ ┣ 📂 AI Tools ┃ ┣ 📂 ChatGPT ┃ ┣ 📂 Julius AI ┃ ┣ 📂 Power BI Copilot ┃ ┣ 📂 Tableau Pulse ┃ ┣ 📂 NotebookLM ┃ ┗ 📂 Perplexity ┃ ┣ 📂 Workflow ┃ ┣ 📂 Upload Dataset ┃ ┣ 📂 Clean Data ┃ ┣ 📂 Generate Statistics ┃ ┣ 📂 Find Trends ┃ ┣ 📂 Identify Root Causes ┃ ┣ 📂 Create Visualizations ┃ ┣ 📂 Forecast Outcomes ┃ ┣ 📂 Generate Recommendations ┃ ┗ 📂 Build Executive Reports ┃ ┣ 📂 Recommended Stack ┃ ┣ 📂 Beginner ┃ ┃ ┗ 📂 ChatGPT + Excel ┃ ┃ ┃ ┣ 📂 Intermediate ┃ ┃ ┗ 📂 ChatGPT + SQL + Power BI ┃ ┃ ┃ ┣ 📂 Advanced ┃ ┃ ┗ 📂 GPT-5 + Claude Opus + SQL + Power BI ┃ ┃ ┃ ┗ 📂 Enterprise ┃ ┗ 📂 GPT-5 + Databricks + Snowflake + Tableau ┃ ┣ 📂 Key Insight ┃ ┣ 📂 Reporting Is Not Enough ┃ ┣ 📂 Prediction Creates Advantage ┃ ┣ 📂 Recommendations Drive Action ┃ ┣ 📂 Automation Scales Decisions ┃ ┗ 📂 AI Multiplies Analyst Productivity ┃ ┗ 📂 Outcome ┣ 📂 Better Decisions ┣ 📂 Faster Analysis ┣ 📂 Higher Accuracy ┣ 📂 Increased Revenue ┗ 📂 Smarter Business Growth

@simonw so youre nesting python into wasm into python is this what we call pythonception or just advanced procrastination?

I may have finally found the Python-in-a-sandbox solution I've been looking for... here's my latest experiment, this time running MicroPython in WebAssembly inside my Python applications simonwillison.net/2026/Jun/6/mic…

@Miles_Brundage so the /goal thing alone cant carry it when they have rival workflow features now cowork vs cc split still confusing though

Giving Claude Code a chance again recently due to workflow (stopped for a while due to annoying bugs). Workflow is good and Codex needs an answer to it soon - /goal is not enough. Though I hope all the Claude apps merge soon - this is my only reason to use CC vs. Cowork now.