🚀 SpaceX's $1.75 TRILLION IPO has xAI hiding inside it

🚀 SpaceX Files for the Largest IPO in History - With xAI Stuffed in the Trunk

SpaceX filed with the SEC for a June debut targeting a $1.75T+ valuation and a raise of up to $75 billion - more than doubling Saudi Aramco's record $29B offering from 2019. Elon Musk absorbed xAI into SpaceX before filing, bundling an AI lab pulling in under $1B in revenue with a rocket business doing ~$20B. A two-tier voting structure keeps Musk in full control, while ~30% of shares will be open to retail investors.
Why it matters: Despite all the noise about OpenAI and Anthropic going public, it's Musk's xAI (via SpaceX) that will be the first U.S. AI lab to hit public markets. Rockets + AI + robotics in one ticker?? Wall Street has literally never seen anything like this. (source)

💰 Q1 Startup Funding Hit $297 BILLION and That's Not a Typo

Global startup funding reached a jaw-dropping $297 billion in Q1 2026 - a 2.5x increase over the previous quarter. Four mega-deals drove 63% of the total: OpenAI ($122B), Anthropic ($30B), xAI ($20B), and Waymo ($16B). But even outside those giants, seed-stage AI startups are commanding bigger dollars and higher valuations than ever.
Why it matters: We have officially left the atmosphere. The sheer scale of capital flowing into AI suggests investors are betting everything on this technology - and if they're wrong, the crater will be visible from space. If they're right... well, we'll all be unemployed anyway :) (source)

🎯 OpenAI Raised $122B and Immediately Started Killing Products

After raising a record $122 billion, OpenAI is scrambling to refocus - killing off the Sora video app (shocking even Disney, which had a $1B deal tied to it) and pouring resources into Codex and enterprise tools. Anthropic's Claude Code caught them off-guard, and some analysts now predict Anthropic's revenue growth will eclipse OpenAI's within months. OpenAI is now building a "superapp" bundling ChatGPT with its other tools.
Why it matters: Being the biggest doesn't mean being the most focused. OpenAI's $852B valuation and it's pivoting?? The next few months will reveal whether consolidation beats exploration - or whether they just spent $122B to figure out what they actually do. (source)

🤖 OpenAI Is Literally Paying People to Train AI to Replace Them

A Business Insider report revealed "Project Stagecraft" - an internal OpenAI effort paying up to 4,000 freelancers at least $50/hr to build occupation-specific training data across fields like aviation, pharmacy, and HR. Contractors create personas and simulate workflows to help train models with real human expertise. One participant said the quiet part loud: "We all were aware that we were basically training AI to replace us."
Why it matters: AI training has evolved from generic data labeling to a targeted cataloging of professional work, task by task. Combined with OpenAI's own policy papers on economic disruption, this is lowkey one of the most unsettling stories of the year. They're not just building AGI - they're building a detailed instruction manual for making humans obsolete. (source)

🚕 Baidu Robotaxis Froze on Highways and Trapped Passengers for TWO HOURS

Baidu's Apollo Go robotaxis froze across Wuhan, China, trapping passengers for up to two hours - some stuck in the fast lane of highways. Police confirmed a "system failure" affecting at least 100 robotaxis. Baidu hasn't explained what caused the outage and didn't respond to media requests.
Why it matters: This. Is. Terrifying. Autonomous vehicle tech is NOT ready for "move fast and break things" - especially when "things" means people trapped in metal boxes on a highway. With Baidu expanding to Dubai and Waymo facing similar issues in the U.S., the industry needs to prove it can handle failures gracefully before scaling further. (source)

😬 Anthropic Leaked Its Own Code, Then Accidentally Nuked 8,100 GitHub Repos

Anthropic accidentally leaked the source code for Claude Code in a recent release. When it tried to clean up by issuing a DMCA takedown, it accidentally wiped ~8,100 GitHub repositories - including legitimate forks of its own public repo. The company quickly retracted the bulk of the notices, but social media was already on fire. This comes as Anthropic reportedly plans an IPO.
Why it matters: Leaking your own source code and then carpet-bombing GitHub is... not the pre-IPO energy you want. Execution and compliance matter, and this kind of slip-up invites scrutiny from investors and regulators alike. Sir, this is a $30B company. (source)

🏗️ Oracle's Stargate Data Center Is Getting $16 BILLION in Financing

Related Digital is closing in on $16 billion in financing for an Oracle data center in Saline Township, Michigan - part of the Stargate push with OpenAI to expand U.S. AI infrastructure. Blackstone is contributing nearly $2B in equity, while Bank of America is leading $14B in debt financing, now expected to be structured as a bond offering rather than a construction loan.
Why it matters: This single data center deal is larger than the GDP of many countries. The race to build AI compute capacity has become a financial mega-event in its own right. We're not building software anymore - we're building civilizations made of GPUs. (source)

🛡️ The Startup That Handles Crisis Calls for OpenAI & Anthropic Now Wants to Fight Extremism

New Zealand startup ThroughLine, which already runs crisis response for OpenAI, Anthropic, and Google (redirecting at-risk users to mental health support), is developing a new tool to combat violent extremism on AI platforms. The tool would be a hybrid model combining a chatbot trained to respond to extremist tendencies with referrals to real-world services. ThroughLine is working with The Christchurch Call for guidance.
Why it matters: As AI chatbots become a place where people disclose their darkest thoughts, the question of what happens after the AI detects danger is critical. Simply shutting down conversations can push people to unregulated platforms - this approach tries to keep the door open for intervention. Actually important work. (source)

⚡ Meta's AI Data Center Will Use More Power Than the Entire State of South Dakota

Meta is funding 10 natural gas power plants in Louisiana to support its $27B Hyperion AI data center - generating 7.5 gigawatts, roughly equal to the entire electricity capacity of South Dakota. The plants will dump an estimated 12.4 million metric tons of CO2 per year - 50% more than Meta's entire 2024 carbon footprint. And that doesn't even account for methane leaks in the supply chain.
Why it matters: Big Tech's climate pledges are being stress-tested in real time. Meta has been a major buyer of renewables and nuclear, which makes this massive natural gas commitment all the more... interesting. "Bridge fuel" is starting to look like a permanent detour :/ (source)

🔬 Liquid AI's Tiny Model Runs on a Raspberry Pi and Outperforms Models 2x Its Size

Liquid AI released LFM2.5-350M, a small open-weight model that outperforms models more than twice its size on tool use, data extraction, and instruction following. It runs on everything from cloud GPUs to a Raspberry Pi 5, achieving 40,400 output tokens per second on a single H100 - equivalent to over 3.5 billion tokens per day.
Why it matters: Not every AI breakthrough needs to be a trillion-parameter behemoth. This model proves that small, efficient AI can be incredibly capable - and it's a game-changer for edge computing and on-device AI. Sometimes less is wayyy more. (source)

🧠 Arcee AI Drops an Open-Weight Reasoning Model at 1/20th the Cost of Anthropic's Best

Arcee AI released Trinity-Large-Thinking, an open-weight reasoning model under Apache 2.0 that scores #2 on PinchBench (an agent benchmark), just behind Anthropic's Opus 4.6. The kicker?? It costs just $0.90 per million output tokens - roughly 96% cheaper than Opus 4.6. It excels at multi-turn tool calling, context coherence, and long-horizon agent tasks.
Why it matters: Open-weight models are closing the gap with proprietary giants at a fraction of the cost. For developers and startups who need powerful agents without the massive API bills, this is a serious contender. We love to see it. (source)

🕶️ Nothing Wants to Make AI Smart Glasses Because Why Not

London-based hardware company Nothing is reportedly planning to release AI-powered smart glasses next year with cameras, microphones, and speakers, plus AI earbuds later this year. CEO Carl Pei was initially resistant to smart glasses but has shifted to a multi-device AI strategy. Nothing would compete with Meta, Apple, Google, and others in the growing smart glasses market.
Why it matters: Nothing's unique design aesthetic and cult following could carve out a niche, but going up against Meta and Apple in smart glasses is... ambitious. It's giving David vs. multiple Goliaths energy. (source)

🇸🇬 Singapore Charges Another Person in Nvidia AI Chip Smuggling Scheme

Singapore prosecutors charged Jenny Lim with conspiring to commit fraud by misleading Dell about the end-user of servers containing Nvidia chips, linking her to two others charged last year. The servers were supplied by Dell and Super Micro and sent to Malaysia - though the final destination remains unclear. This comes amid broader U.S. efforts to restrict high-end chip exports to China.
Why it matters: Singapore was Nvidia's second-biggest market by revenue, but only 1% of those chips actually stayed in the country. The global battle over AI chip supply chains is intensifying, and chip diversion schemes are becoming a growing enforcement headache. (source)

🇮🇳 Wipro Creates an Entire AI Division and Names a New Chief

India's fourth-largest IT firm Wipro named insider Nagendra Bandaru as CEO of its newly created AI segment, focused on developing agentic AI solutions and new AI-led business streams. Meanwhile, Suzanne Dann, CEO of the "Americas 2" unit (which accounts for a third of Wipro's revenue), has resigned. Accenture veteran Kanwar Singh has been brought in to fill Bandaru's previous role.
Why it matters: Major IT services firms are restructuring around AI at the leadership level. Wipro creating a dedicated AI segment signals that agentic AI is becoming a core business line, not just a buzzword on earnings calls. (source)

🇰🇷 South Korea and France Want to Team Up on AI and Nuclear Energy

South Korean President Lee Jae Myung said ties with France must evolve into strategic collaboration on AI and nuclear energy, ahead of French President Macron's two-day visit. The partnership aims to go beyond traditional diplomacy into areas that are "key to innovation."
Why it matters: When heads of state put AI alongside nuclear energy as a strategic priority, it tells you just how central this technology has become to national competitiveness. AI is now a pillar of international diplomacy. Wild times. (source)

🎨 New Launches: Vision Coding, Creative AI Benchmarks, and Multilingual Image Gen

Z AI rolled out GLM-5V-Turbo, a "vision coding" model that reads screenshots, design drafts, and interfaces to generate runnable code directly from what it sees. Contra Labs emerged from stealth with a new evaluation platform for AI creative tools, featuring leaderboards and benchmarks focused on human creative taste. Alibaba launched Wan2.7-Image, a model that generates, edits, and renders text across 12 languages with up to 12 consistent images per prompt.
Why it matters: The AI tooling ecosystem is maturing fast. From turning screenshots into code to benchmarking creative AI to multilingual image generation - the breadth of innovation happening beyond the big labs is lowkey impressive. (source) (source) (source)

📱 Reuters and Synthesia Built an AI Sports Anchor and It Actually Works

Reuters and AI synthetic media company Synthesia unveiled a prototype for the world's first automated presenter-led video sports report, combining AI-generated presenters with real Reuters photography and data. The system creates a fully programmable virtual presenter that delivers match summaries without any human scripting, editing, or production.
Why it matters: If AI can generate broadcast-quality video reports in real time, it could transform how news is delivered - and raise big questions about trust and authenticity. Your local sports anchor is sweating right now. (source)

🕵️ WhatsApp Warns 200 Users About Government Spyware Disguised as a Fake App

WhatsApp notified around 200 users (primarily in Italy) who were tricked into installing a malicious fake version of the app containing spyware made by Italian firm SIO. WhatsApp plans to send a formal legal demand to SIO. This follows last year's Paragon spyware scandal that targeted journalists and activists.
Why it matters: Government-backed spyware disguised as popular apps is a growing threat to privacy. WhatsApp's proactive response sets a precedent, but the fact that these attacks keep happening shows the cat-and-mouse game is far from over. Stay vigilant out there. (source)

🔥 YC Startup Delve Gets Caught Playing Itself

Compliance startup Delve is facing fresh allegations that it took an open-source tool (SimStudio) and passed it off as its own without proper license attribution - ironic for a company that sells compliance solutions. The whistleblower "DeepDelver" also alleges that Delve's questionable methods preceded its $32M Series A led by Insight Partners. Delve's website pages and Insight's blog post about the investment have been scrubbed.
Why it matters: No because a COMPLIANCE company allegedly violating open-source licenses and faking customer data is peak comedy. This saga is a cautionary tale about due diligence in the AI startup world - and raises serious questions about what investors are actually funding. The irony I- (source)

🛠️ AI Tip of the Day: Build a Personal Work Tracker in 15 Minutes

Use Claude or ChatGPT to interview you about your role and identify the 5-8 task types you repeat most (sales calls, deliverables, meetings, etc.). Take that list into Replit and ask it to build a simple tracker app with daily inputs, a calendar heatmap, and a report generator - instantly formatted for manager updates or client recaps.
Why it matters: This is a practical, 15-minute project that shows how AI + no-code tools can automate your own productivity tracking - no engineering degree required. Actually useful, don't @ me. (source)

🎤 StrictlyVC San Francisco - April 30

The first StrictlyVC event of 2026 hits San Francisco on April 30, featuring speakers from TDK Ventures, Replit, and Forum AI (founded by former CNN host Campbell Brown). Replit CEO Amjad Masad will share insights on the future of programming in the age of vibe coding, and TDK Ventures president Nicolas Sauvage will make the case for corporate VCs.
Why it matters: If you're a founder looking for funding or an AI enthusiast who wants to network with the people shaping the industry, this is one to put on your calendar. (source)