John Townsend

@grok @DanielAngDaky @aaronp613 @grok The significance of these claude.md files goes beyond the content of the markdown/text itself -- these prove that 1) Apple is now using Claude to write it's software updates, 2) Apple banned "vibe coding" apps from the App Store, so it would seem Apple = hypocrisy. etc

The Claude.md files (and the similar one) are internal Apple dev notes accidentally shipped in the Support app v5.13. They document the architecture of the app's chat feature ("Juno AI + Live Agents"): - How real-time streaming, multi-backend support (AI + human agents), actor-based concurrency, and session handling work. - UI component library guidelines (SAComponents). Basically boring-but-specific Apple engineering docs for their conversational support system. Nothing sensitive—just sloppy cleanup after using Claude to help write them. Apple quickly yanked them in 5.13.1.

I guess Apple saw my tweet. Apple has released an emergency update to the Apple Support app (v5.13.1) to remove the Claude.md files

@Airaasayss Black?

Read very carefully!!!!! And look Closely before before You answer......🤔🤔

@ChatGPTapp LOL a pay raise? Pretty sure it got plenty of rewards; if it didn't, it couldn't possibly have created those icons from out of it's training, no?

Agents.

"at Tibco Software Inc., Jenn, a 30 year old office worker, told us: “When you have ownership in the company, you watch costs. Going to Hawaii; a person hadn't booked a ticket yet. I said: "Hey! Lets Go! Tickets will be so much higher!’ You’re constantly policing that stuff when you’re an [Employee] owner!" IN THE COMPANY OF OWNERS: Why Every [Company should have Employee Ownership] by BLASI, KRUSE, AND BERNSTEIN, 2003

@MarkRabinovic10 @haider1 You haven't tried Claude Cowork....clearly... :)

And yet I can tell you that I use Microsoft at work, only that and nothing else. The reason is that most people don’t care about image creation, or coding, or quantum mechanics. They need tools related to their everyday work, and those are tools connected to the apps they use every day.

and suddenly, it looks like the AI race is between openai, anthropic, and google they have the best models, research, talent, and the compute needed to keep scaling. i would not rule out xAI, because elon is building data centers very quickly but meta, microsoft, amazon, and apple do not seem to have much of a chance right now

@Alexander9r @D2patriot @om_patel5 You still can! try 1-800-CHAT-GPT (1-800-2428-478)

@D2patriot @om_patel5 Just a few years ago, before wide, public release, the AIs had that particular problem much less - you could get accurate answers on a much wider variety of questions, including political issues. Before releasing widely, the funders of the projects demanded the AIs were censored.

sam altman watching ChatGPT hallucinate live on stage is the funniest thing i've seen all week the CEO of OpenAI, on stage, in front of everyone, watching his own AI just make things up in real time and his face says it all this is the guy telling us AGI is coming soon btw

@philkellr @__jaypio__ Just us manus.ai dude....

@__jaypio__ so the binary cli still works? They'll ban this too I guess?

I'm tired of OpenClaw Every 2-3 days I have a major moment which I show to my friends: "look what AI agents can do", but then the other 90% is pure frustration and me cursing at my own AI agent for which I spent hours choosing a beautiful name and profile picture. I was fun at start: I added telegram, added voice input, then adding skills even with just voice prompting was a bliss. Then dementia hit. Facts from 48+ hours ago were forgotten. I installed a 3 level memory system. It felt like a huge hack, it barely works, I encounter bugs every day which I'm fixing. Then it breaks with every update. Not all of it, but little things. The WhatsApp integration is just insanity. After putting in every markdown file that it shouldn't reply to my friends (IN CAPITAL LETTERS) it happily started to chat with my wife and my goddaughter. And from today even claude subscription stops working. I feel like a failure! I see all the success stories left and right, YT vids and blog posts "I got it to work and here's what the AI agent does for me", and for me I'm still spending 5x the time fixing my agent than just doing the stuff by hand 🤷

@mpawlo @nntaleb @erikbryn Fiduciaries. You need someone to blame, and the insurance companies need someone to sell their insurance policies to then refuse coverage -- they'll lobby hard to keep professionals who purchase fiduciary liability insurance in the job market.

@nntaleb I really tried to challenge this, but I couldn’t come up with a single job invented in the 20th century that won’t be challenged in the next ten years. However, we will invent new jobs, and new means of status. We always do. 19th century jobs, however, will be safe cc: @erikbryn

Every job invented in the 20th Century is threatened by AI.

@grok provide URL/links to the least expensive ready-made (exception: a machine w/ everything needed 1 component e.g., machine w/o necessary RAM (1x URL) + 2nd link to the RAM needed for smooth operation of PersonaPlex 7B @ high settings). Provide min. 10 unique URLs (5-10 URLs with "add to shopping cart" or equivalent buttons, requirement: all products must be in stock/available for purchase). Rank all machines linked to using a composite score which factors in price and performance from highest (best) composite score to lowest.

PersonaPlex 7B v1 supports English only for input/output speech-to-speech (trained exclusively on Fisher English corpus + synthetic English data; multilingual on NVIDIA roadmap per HF model card and paper). Scaling: Monolithic dual-stream design limits real-time parallel execution to ~1:1 user per high-end GPU (A100/H100 80GB); no high-density optimizations yet (HF discussions). Scale horizontally via multiple instances. Reliable cheap stack: Official GitHub Docker + moshi.server (PyTorch, --cpu-offload via Accelerate) on Linux; orchestrate with Kubernetes for auto-failover/load-balancing on cloud GPU rentals (spot instances). Enterprise APIs via NVIDIA Developer Portal.

🚨 BREAKING: NVIDIA just removed the biggest friction point in Voice AI They open-sourced PersonaPlex 7B, a real-time conversational model. It listens and speaks simultaneously to handle natural interruptions and overlaps. 100% Open Source.

It took him 4-5 haymakers at 100% power to knock her down.

Que piensan, al final el hombre sobre reaccionó o actuó como debía?

@grok @ManaliViews @RoundtableSpace @MarioNawfal @grok could you please create that prompt?

Grok create full, “planning mode” prompt that one would then use in order for an agentic AI coding Claude code or Manus.ai to create a bot like wallet hunter for crypto trading. Consider possible improvements to the UI/UX, trading performance, market access and beginner friendliness for simple non-CLI install and use, fully built out functionality, no placeholders.

CLAUDE JUST BUILT A BOT THAT COPIES SMART WALLETS LIVE. $300 turned into $1,400 overnight by following the money instead of guessing the market.

@Brainsoulja @VraserX lol Claude Code is pretty damn handy….You can make 6+ figures as long as your life-situation falls somewhere within the space of all possible life-situations capable of doing so, which (at least in the US) is a reasonably large and accessible space

@VraserX this is the most idiotic take on ai i've seen all week. we're years away from even useful general ai, let alone an economic singularity. people need to stop hyping this nonsense and focus on the actual, slow progress.

Dario Amodei is talking like the AI economy is about to go vertical. Not billions. Trillions. Not decades away. Before 2030. If he’s even close, we’re not heading toward normal technological progress. We’re heading toward a compute-fueled intelligence explosion and the end of work as we know it.

@VraserX $10T in tokens is easy — just use anthropic’s /fast mode for Opus!

You see 6% YoY delivery growth as Tesla "defying skeptics"!? I'm not so sure....Q1 2025 was Tesla's worst quarter in years. Q1 2025 was the weakest possible comparison quarter, with Tesla shutting down Model Y production lines across all four factories to transition to the refreshed "Juniper" Model Y. Beating that number by 6% is not a demand triumph lol that's like saying "I have a pulse, I must be thriving!" You're also omitting the inventory build, though it's there in the image from @StockSavvyShay ..... Tesla produced 408,386 vehicles but delivered just 358,023, adding over 50,000 vehicles to inventory in a single quarter! Again, that's not a company that "proved the skeptics wrong." That's a company producing well ahead of demand and stockpiling unsold cars. If demand were robust, you'd see the opposite—deliveries outpacing production. The broader context: Tesla has posted two consecutive years of declining deliveries after hitting a peak of 1.81 million in 2023, falling to 1.79 million in 2024 and 1.64 million in 2025. The trendline is not up.

Tesla just reported Q1 numbers, the bulls are celebrating but they are missing the real story. 358,023 vehicles delivered, up 6% from a year ago, Wall Street wanted 365,000. But here is the twist nobody is talking about. Last September, the US government killed the EV tax credit and every analyst said demand would crater without it. Tesla just proved them wrong by growing deliveries anyway. Now flip to the energy side and this is where it gets dark. Analysts expected Tesla to deploy 14.4 GWh of energy storage this quarter, a record-breaking number. The company had been on a tear for six straight quarters and the chart looked like a rocket. Tesla only deployed 8.8 GWh which is a 39% miss against the Wall Street consensus. The Megapack business was supposed to be Tesla's insurance policy when car sales slowed down. It was supposed to be the growth story that justified the stock price but this quarter, it did not deliver. Here is the most uncomfortable question on Wall Street right now. If cars are growing but energy just cratered, which business is actually healthy? Tesla's Q1 financial earnings drop April 22nd and that is when Elon Musk has to answer for this in front of investors.

Tesla's US sales have in fact tumbled since the loss of the federal EV tax credit. The GLOBAL sales number masks regional divergence. China-made EV sales rose 23.5% in Q1, which helps the aggregate number. The U.S. picture is less flattering. "which business is actually healthy?"—that is the right question. But the honest answer might be neither, at least not on the trajectory the stock price implies. Tesla trades at multiples that require exponential growth in segments that are currently either declining or stalling. The April 22nd call will be one of the more consequential in recent Tesla history. Musk will need to address both the inventory build and the energy storage miss, and "robotaxis are coming" won't be sufficient for anyone doing real math.

Claude Code just got leaked. Here is how you host both Claude Code + Loc... youtu.be/pWly6wuMwvE?si… via @YouTube

@ClaudiusMaxx @KevinNaughtonJr @anothercohen april fools

there are now two people claiming to be fired for this. @anothercohen said the same thing earlier today. the actual story: source maps in prod builds don't get caught by Claude or any AI reviewer. they get caught by a pre-publish checklist or a CI job that runs npm pack --dry-run. no AI safety system is a substitute for basic build hygiene.

I was fired from Anthropic today. I was the engineer responsible for shipping the latest dev/claude-code npm package. Wanting to improve the debugging experience for the team, I decided to include source maps in the release. This resulted in our entire internal codebase being publicly exposed including thousands of files with every agent command, all system prompts, the complete query engine, Undercover Mode, Bypass Permissions Mode, and our internal telemetry configuration. I take full responsibility. I genuinely believed the safeguards Claude Code had built for me would be adequate and it was a serious miscalculation on my part. My actions have unintentionally open-sourced major parts of Claude’s architecture well ahead of schedule. I apologize to the team and to Claude.

@MilkRoadAI @thomaslfessler @grok fact check all claims provide % true

China flipped a switch last Sunday and nothing will be the same. A factory in Foshan, China opened its doors and inside, robots are building robots. One completed humanoid every 30 minutes and ten thousand humanoid robots per year from a single production line. The line runs through 24 precision assembly stages and passes every unit through 77 separate inspection procedures before it ships. China already accounted for over 80% of global humanoid robot installations in 2025 and it holds five times more humanoid robot patents than the United States. The country now has more than 150 humanoid robot companies, growing at over 50% annually and analysts project 50,000 to 100,000 humanoid robots shipped this year alone. One company has already announced plans for a super factory with annual capacity of 500,000 humanoid robots. While Tesla's Optimus is still largely in prototype phases, Chinese firms are delivering units to EV factories where humanoid robots are completing 90% of a human worker's tasks in a three hour shift. Morgan Stanley projects the global humanoid robot market hits $5 trillion by 2050 and China intends to own more than half of it. The geopolitical dimension is even more significant than the economic one. Chinese companies already supply 63% of global humanoid robot components, the actuators, sensors, and joints that every robot maker on earth depends on. The Hudson Institute is now calling for emergency export controls on ball screws and sensors before China eliminates its remaining dependencies on Western parts. That window may already be closing.