In partnership with
☀️ TRENDING AI NEWS 🤖 FlashProve: Anthropic released an Apache 2.0 zkSNARK prototype that proves Llama 4 Maverick replies at 110 tokens/second on one H100. 🏢 Google Search: Google confirmed schools can't use Search without AI, so restricting AI means switching Search off entirely. 🛠️ ChatGPT Search: OpenAI says search queries in ChatGPT now equal 20% of Google's volume. 🚨 Canva Layoffs: Canva is cutting or relocating more than 1,200 jobs, 22% of its global workforce. |
Imagine paying for a premium AI model and having no way to check that the answer actually came from it. That gap just got a serious attempt at a fix, and it's the first story below.
| 🛠️ Tool of the day | |
Gamma - an AI tool that turns a short prompt or outline into a polished presentation, document or page. It helps anyone who needs a clean deck fast, and you can start for free.
| ⚡ Quick hits | |
Netflix is rolling out an AI Discover tab on TV that recommends shows and movies based on your mood. Read more →
Y Combinator CEO Garry Tan says he spends roughly $100,000 to $200,000 a year on AI tokens to run agents for work and daily life. Read more →
Apple has started developer beta tests of Audio Intelligence for Apple Watch, with features like Live Rewind and Siri Recaps. Read more →
🤓 AI Trivia
What does the 'zk' in zkSNARK stand for?
The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇
| 🔐 Anthropic Wants to Prove Which Model Answered You | |
FlashProve brings cryptographic receipts to AI
Anthropic is working on a system called FlashProve that would let anyone check that an AI response really came from the model it claims to be from. The prototype ships under an Apache 2.0 license and uses a zkSNARK to prove a reply came from a committed model without revealing the weights.
The numbers are early but concrete: it proves Llama 4 Maverick at 110 tokens per second on a single H100, with proofs around 19 MB. That is a long way from a free lunch, but it is the first time provable inference looks like something a real team could test.
The bottom line
If you route work through third-party APIs, model substitution is a real worry, and proofs like this could eventually make 'trust us' obsolete. Not production-ready, but worth watching.
Stop feeding the feed.
You’ve put in the work to build an audience. But when the only way to reach your fans is through a social feed determined by algorithms, the platform decides who sees what you make next.
beehiiv gives you the tools to reach your audience directly. On their all-in-one platform, you can turn followers into subscribers, grow your newsletter, and connect it all to your community or podcast. Stop chasing virality on platforms you can’t control. Build a stronger, more sustainable business on beehiiv.
| 🏫 Google Says You Can't Use Search Without AI | |
Schools get a stark all-or-nothing choice
Google announced a slate of AI products for classrooms on Tuesday, along with more ways for schools to restrict them. The catch: access to Gemini, Gemini Notebook, Gemini in Classroom, and Google Search will be entirely off for K-12 student accounts unless administrators explicitly turn them on.
Put plainly, a school that rejects AI loses Search too, because AI is now woven into the engine itself. It's a revealing admission about how entangled Google's core product has become with AI features.
The bottom line
Opting out of AI is quietly becoming opting out of the web's default tools. Expect this debate to reach every school district that wanted a middle path.
| 🔎 ChatGPT Search Is Now 20% of Google's Volume | |
'Ask Chat' is replacing 'Google it'
At a press event on Monday, OpenAI's consumer product lead for ChatGPT, Adam Fry, said the volume of search queries run through ChatGPT now equals 20 percent of the volume run through Google.
The company defines a search as any time a user inputs a query that results in a web lookup, so the comparison is OpenAI's own framing rather than an independent measure. Still, a fifth of Google's scale is a striking claim given how recently chat assistants were treated as a novelty.
The bottom line
Treat the 20% as a directional signal, not a verified stat. But combined with Google's schools announcement, it shows search is being redefined from both sides.
| 🏢 Canva Cuts or Relocates 1,200+ Jobs | |
A 22% workforce reset at the design giant
Canva is cutting or relocating more than 1,200 jobs, or 22% of its global workforce, in a major restructure that includes laying off staff who helped build the Australian-founded design platform.
CEO Melanie Perkins called it 'without doubt the hardest moment in our thirteen-year history' in an email to staff. It lands the same week Squarespace announced it is laying off 19% of its employees, a reminder that restructuring is hitting web and design tools across the board.
The bottom line
Two big web-tool companies shrinking in two days is a pattern, not a coincidence. If you work in design or site-building, it's a good week to look at your own workflow.
| ⚠️ Hinton Warns of a Major Incident This Year or Next | |
Blackouts and bank shutdowns top his list
Nobel winner Geoffrey Hinton said it could turn out that humans would have been better off never developing AI, warning the technology could trigger a large-scale incident this year or next, such as a temporary blackout of financial or power systems.
It echoes another warning from this week: the UK's cybersecurity body said AI-powered malware is 'highly likely' to be used against power stations, airports and hospitals by 2028. If you want to see how we've covered the safety debate, our AI safety tag page has the backstory.
The bottom line
Hinton's timeline is short and his examples are infrastructure, not sci-fi. Whether or not you agree, the pressure on critical-systems security is only going one direction.
| 🛠️ Ship Your Idea Faster | |
Quick plug: if all this AI news has you itching to put a project online, our sponsor 60sec.site is an AI website builder that gets you a live site in about a minute. And for more daily AI coverage, head to dailyinference.com.
| 🌎 Trivia Reveal | |
The answer is zero-knowledge! A zkSNARK lets you prove a statement is true without revealing the underlying data, which is exactly how FlashProve can show a reply came from a committed model without exposing its weights. 🔐
| 💬 Quick Question | |
Do you still reach for a classic search engine, or has a chatbot become your first stop for looking things up? Hit reply and tell me - I read every response!
That's it for today - see you tomorrow with more.
| 🎁 Share Daily Inference | |
Share your link with a friend. One referral gets you the AI Tools Starter Kit. Five referrals get you the AI Insider Briefing.