|
|
Hi there, this is your daily ☕️ Techpresso.
|
|
|
In today's Techpresso:
|
|
💥 Anthropic confidentially files for IPO 📈 Alphabet raises $80 billion for AI buildout ⚖️ Florida becomes first state to sue OpenAI over child safety risks 📧 Meta captures employee emails for AI training 🔌 China's military obtained Nvidia chips despite US curbs 🎁 + 10 other news & articles you might like 🧰 + 6 trending tools 📚 + 2 trending papers & reports
|
|
FROM OUR PARTNER
Claude Code writes beautiful code. So does Codex. But here's the thing, they also think they write beautiful code. And when you ask an AI to review code it just wrote, you get the intellectual equivalent of a student grading their own exam. Shockingly, they always pass. CodeRabbit CLI plugs into Claude Code and Codex as an external reviewer — 40+ static analyzers and zero emotional attachment to the code it's looking at. The agent writes, CodeRabbit reviews, the agent fixes. You show up when there's actually something worth approving. The AI still does the work. It just doesn't get to decide if the work is good anymore. Free tier available. Try CodeRabbit's CLI.
|
💥 Anthropic confidentially files for IPO
LINK
|
- Anthropic has confidentially submitted its IPO prospectus to the SEC, putting the Claude maker on track for a share sale that would give investors a way to bet on artificial intelligence.
- The filing gets Anthropic ahead of OpenAI, which is preparing its own confidential paperwork, though the company said any public offering will hinge on market conditions and the SEC review.
- Anthropic's revenue run rate has climbed to $47 billion from $10 billion in annual revenue last year, and it closed a funding round last week at a $965 billion valuation.
|
📈 Alphabet raises $80 billion for AI buildout
LINK
|
- Alphabet is pulling in $80 billion through new equity offerings, an unusually large fundraise aimed at expanding Google's footprint in AI hardware and pouring money into its chip-making operation.
- The plan includes $40 billion in new stock dripped into the market starting in the third quarter, plus $30 billion in underwritten shares and mandatory convertible preferred stock backed by Goldman Sachs, JPMorgan, and Morgan Stanley.
- Berkshire Hathaway is putting up $10 billion for shares, and the money is meant to scale up Google's tensor processing units, or TPUs, which have been competing internally with Nvidia's GPUs.
|
⚖️ Florida becomes first state to sue OpenAI over child safety risks
LINK
|
- Florida has filed the first state-level lawsuit against OpenAI and CEO Sam Altman personally, treating ChatGPT as a defective product and seeking civil penalties along with a court order blocking data collection from children under 13.
- Attorney General James Uthmeier's complaint brings 10 counts including product liability, negligence, and fraudulent misrepresentation, citing the April 2025 FSU mass shooting where the gunman exchanged over 270 messages with ChatGPT about weapons and campus locations.
- The suit lands weeks before OpenAI's expected IPO filing at a valuation of up to $1 trillion, alleging the chatbot causes addiction in young users, lacks parental oversight tools, and contributed to deaths of two University of South Florida graduate students.
|
📧 Meta captures employee emails for AI training
LINK
|
- Meta is logging how its employees work on their computers through a program called the Model Capability Initiative, feeding the data into AI training across more than 200 apps and websites, according to internal documents seen by Reuters.
- The MCI tool tracks mouse movements, clicks, navigation patterns, code changes, browsing history, device sleep cycles, and clipboard actions, and also captures the contents of emails and chats sent to US-based employees, including from overseas colleagues.
- Meta says the tool is installed only on US devices and analyzes interaction behavior rather than message substance, but workers report data usage spikes that burned through a month of home internet in days, suggesting continuous uploading.
|
🔌 China's military obtained Nvidia chips despite US curbs
LINK
|
- A new report reveals that China's People's Liberation Army has spent years acquiring Nvidia AI chips, including models blocked by US export controls, undermining Washington's effort to keep American hardware out of Beijing's military programs.
- Wirescreen reviewed roughly 3,800 procurement records and found about 500 cases between 2019 and 2025 where Chinese military units sought A100, A800, H100, and H800 chips for nuclear simulations, war games, cyber operations, and weapons research.
- A separate Bloomberg review found at least seven Chinese universities tied to the armed forces, including blacklisted Beihang and Northwestern Polytechnical, sought access to Nvidia's H200, the most powerful AI chip Washington still permits for sale in China.
|
|
FROM OUR PARTNER
Your ultimate guide to the future of tech. Semafor Technology decodes the innovations, trends, and forces reshaping the global tech landscape. Each briefing delivers clear insights on AI, machine learning, startups, and policies driving change. Join 75,000+ readers who rely on Semafor Technology Subscribe for free. |
|
|
|
Other
news & articles you might like
-
Anthropic to offer EU access to its advanced Mythos model
LINK
-
Quote of the day by Oracle co-founder Larry Ellison: "Citizens will be on their best behavior, because we’re constantly recording and reporting everything that is going on" — a dire warning on the erosion of privacy
LINK
-
Why Video Agent models are next — Ethan He, xAI Grok Imagine
LINK
-
AI costs how much? GitHub Copilot users react to new usage-based pricing system.
LINK
-
Microsoft and Google are late to AI coding, but 'absolutely critical' they compete for growth
LINK
-
EVs are getting more affordable worldwide — except in the U.S.
LINK
-
Apple’s MacBook Neo is winning over a new generation of buyers
LINK
-
Salesforce acquires Berlin-founded AI platform Contentful
LINK
|
|
🧰 Trending tools
|
Fundraisly: schedules investor meetings directly on your calendar, helping founders secure 20–40 meetings within 90 days without manual outreach effort.
LINK
|
|
Vokal: a shared workspace that lets multiple AI coding agents from different providers collaborate directly, eliminating manual copy-paste handoffs between teammates.
LINK
|
|
Gigacatalyst: lets sales and CS teams build custom features directly into your product using AI, reducing churn by adapting software to each customer's workflow.
LINK
|
|
Co-Invest: lets you research, analyze, and execute real trades across 500+ markets directly within ChatGPT or Claude using plain language commands.
LINK
|
|
Paste MCP & AI Tools: gives your clipboard a searchable history, letting you instantly retrieve any text, color, or asset you've previously copied.
LINK
|
|
Enshittifier: replaces "AI" text with a poop emoji system-wide on Mac and Chrome, nudging mindfulness about AI hype through font ligature trickery.
LINK
|
|
You can check the previous tools here, or add your tool here
|
|
📚 Trending papers & reports
|
> Flat datacenter networks built on quasi-random graphs match or beat traditional fat tree performance while being up to 45% cheaper, and are now Amazon's default network for most workloads.
LINK
|
|
> Larger AI models beat smaller ones on rare, complex tasks because they have enough capacity to handle common tasks without overwriting the slowly building knowledge needed for uncommon ones.
LINK
|
|
|
|
🎓 Want to master the AI tools we cover every day? Techpresso's AI Academy has 300+ step-by-step tutorials on ChatGPT, Claude, Perplexity, and every tool that matters. No fluff — just practical workflows you can use at work.
|
|
💬 How did you find today's edition?
We read every reply — just reply to this email and let us know how we can improve!
|
|
|
