Sign Up · Advertise |
|
|
Welcome, humans. |
In case you missed it on Friday, Bumble is launching an AI dating assistant called "Bee" that writes your profile, generates conversation starters, suggests date ideas, and plans outings. Meanwhile, Tinder held its first-ever product keynote and launched IRL curated events, video speed-dating, and something called a "Does This Bother You?" safety LLM. |
Both apps announced AI features the same week. Two companies whose entire business model is getting strangers to talk to each other just admitted the talking part is so broken they need robots to help. At this rate, two AI agents will fall in love, plan the wedding, and your only job will be showing up. AI-ranged marriages! |
Here's what happened in AI today: |
😺 Businesses are picking sides in the AI platform war, and the data says Anthropic is winning 📰 Ai2's CEO stepped down because nonprofits can't compete at frontier scale 📰 Facebook Marketplace now lets Meta AI auto-reply to buyers for you 🍪 LogClaw deploys an AI site reliability engineer inside your VPC 🌟 Ethan Mollick graphed every major AI benchmark and the curves all look the same
|
… and a whole lot more that you can read about here. |
P.S: Want to reach 675,000 AI-hungry readers? Click here to advertise with us. |
|
😺 Over $2B in AI Funding Hit in a Single News Cycle. Here's Where the Money Went. |
If you want to know where AI is heading, don't read the research papers. Read the term sheets. |
This past week, over $2B in AI funding and valuations landed in one news cycle (and it wasn't from OpenAI or Anthropic; shocking!). Three patterns jump out. |
Pattern 1: The "boring" verticals are getting the biggest checks. |
Rox AI hit a $1.2B valuation for autonomous sales agents that plug into your CRM and deploy AI workers to monitor accounts, research prospects, and update Salesforce (Sequoia, General Catalyst) Oro Labs raised $100M for AI-powered corporate procurement (automating vendor selection and contract negotiation) Wonderful raised $150M at a $2B valuation for AI customer service agents operating in 30+ non-English markets Waiv raised $33M for AI-powered blood testing that detects cancer early
|
Nobody's writing viral tweets about procurement automation. But that's where the biggest checks are going. |
Pattern 2: Agent infrastructure is the new cloud. |
Gumloop raised $50M to let non-technical employees build AI agents (Benchmark led) Qdrant raised $50M for vector search (the retrieval layer that lets agents find relevant information) Onyx Security raised $35M to monitor, govern, and correct AI agents inside enterprises HydraDB raised $6.5M for agent memory infrastructure (so agents can remember context across sessions)
|
If agents are the new apps, this infrastructure is the new cloud. Search (Qdrant), memory (HydraDB), governance (Onyx), building (Gumloop). The VCs are funding the picks-and-shovels layer. |
Pattern 3: The creative AI arms race is global. |
PixVerse raised $300M backed by Alibaba for AI video generation, becoming a unicorn Axiom raised $200M at a $1.6B valuation for "Verified AI" using formal mathematics (their prover got a perfect Putnam score, only the fifth in a century) Bold raised $40M in Israel for on-device AI cybersecurity agents
|
A Chinese-backed video AI unicorn and a math-verification startup at $1.6B in the same 24 hours. Neither is US-only. |
Why this matters: When over $2B moves in a single news cycle, and it's not just OpenAI or Anthropic doing another mega-funding round, it's a signal. Agents are the next platform, we hate to say it but boring verticals are the biggest opportunity, and the global AI race has more players than Silicon Valley wants to admit. After all, if the world is moving towards a world where agents do all the work, they're going to need a lot of infrastructure and services… And no, we're not talking about Rent A Human. Although… if one needed a career pivot… there are worse ideas. |
|
|
|
|
|
Explore Brand Radar |
|
🎓 AI Skill of the Day: Use AI to research a company in 5 minutes |
Whether you're prepping for a job interview or curious about one of today's funded companies, AI can compress hours of research into minutes. The key: give it a structured framework. |
Research [COMPANY NAME] and give me a briefing with these sections: 1. What they actually do (one paragraph, no jargon) 2. Business model (how they make money, who pays) 3. Key numbers (revenue, funding, headcount, growth rate) 4. Who they compete with and how they're different 5. The biggest risk to their business 6. One thing most people get wrong about them Use only information you can verify via web search. Flag anything uncertain. |
|
|
Our favorite insight: section 6 ("one thing most people get wrong") consistently produces the most interesting output. It forces the model to find a contrarian angle rather than restating the company's own marketing. Try it on any company in today's funding roundup. |
Now, if you want something more builder friendly, go read this: It's Ben of Ben's Bites sharing his personal Codex stack, skills he uses, tools he prefers, a skill he built to recreate Anthropic's visual design mechanism in Codex (to install, just give the repo link to Claude Code or Codex and ask it to install), and a project he made where you can learn how to "become a builder" through an interactive process (love this). |
Want more tips like this? Check out our AI Skill of the Day Digest for this month. |
Have a specific skill you want to learn? Request it here. |
|
Three NEW Podcasts from this week… |
An open video model you can run on an 8GB GPU & edit for free, how Canva's AI tools hit 24 billion uses, and just released today, the battery CEO whose AI compresses 8 years of testing into 2 weeks—with GM, Honda, and Hyundai already on board. |
Click any image below to watch on YouTube, or here for Spotify or Apple Podcasts. |
|
|
🍪 Treats to Try |
|
|
|
📰 Around the Horn |
Ai2 CEO Ali Farhadi stepped down after two and a half years because nonprofits can't compete at frontier-scale model development against tech giants spending billions on compute. Board chair: "The cost to do extreme-scale open model research is extraordinary." Facebook Marketplace rolled out Meta AI auto-replies that answer buyer inquiries using your listing details, plus AI-generated draft listings from photos and suggested pricing (3.5M listings posted daily in the US and Canada). Amazon employees report that AI tools are increasing their workload rather than reducing it, and a new independent study confirms their suspicions. Fargo police jailed an innocent grandmother for over five months after faulty facial recognition AI wrongly identified her as a fraud suspect who had never been in North Dakota. Want absolutely EVERYTHING that happened in AI this week? [Click here!](link to ATH digest)
|
Want absolutely EVERYTHING that happened in AI this week? Click here! |
|
|
Focus on the things you love. Vellum can do everything else. |
|
Vellum built a personal AI that actually knows who you are, what you're working on, and what you need before you need it. Private and secure by default. |
Here's what we love: Neuron readers get early access: join the live demo on Lenny's AI-Native PM Series to get it. Register here. |
|
🌟 Sunday Special: |
The AI coding divide nobody wants to talk about |
A blog post called "Grief and the AI Split" hit the top of Hacker News this week and names something a lot of developers have been feeling but couldn't articulate: AI-assisted coding is revealing a split, where some programmers are "craft lovers" who find meaning in the elegance of code itself, and others are "result chasers" who learned to code because they wanted to make things happen. |
Both are valid. But they're experiencing AI very differently, and pretending otherwise is causing unnecessary grief. Honestly, this divide explains all of the pro vs anti AI hate out there across every segment AI is touching. IMO, it's the defining divide of AI, period. |
Whether you're a craft lover of art, writing, music, code, business processes, even the idea of "doing work." There are those who love the craft, and those who want the result of the craft. AI is a faster path to that end result (whether you're happy with that result is another matter). |
Counter take (from the same thread) which captures the code component of the argument with a bit more nuance: |
"I think the article misunderstands completely. "Craft" coders are chasing results too — we're just chasing results that last and that can be built upon. …I think the real divide we're seeing is between people who saw software as something that is, fundamentally, improvable and understandable; and people who saw it as a mysterious roadblock foisted upon them by others, that cannot really be reasoned about or changed." | | | | -wiml |
|
|
So if you can expand out from there to the rest of the craft vs result debate, there are people who still see value in building a skillset (i.e. craft, or skills that make up a career in a given field) that is enduring, with an educational background and experience that can be built upon. |
The problem is: as valuable as it is on an individual level to cultivate the ten thousand hours needed to master something, the market only rewards results, and at a blistering pace that changes all the time. You can spend ten thousand hours to learn something that by the time those ten thousand hours are up, the market no longer requires, or that you're not globally the best at, and you can't monetize. Don't necessarily have a fix for this (well, actually I do), but it's a good debate to have. |
Are AI models actually getting better? |
Ethan Mollick graphed the most critical AI benchmarks this week and found the improvement curves all have a similar shape over the past year. Better, yes. But at a consistent pace, not exponential leaps. |
entropicthoughts showed LLMs have produced zero improvement on strict SWE-Bench merge rates for over a year. And AI2 released "Deep Research, Shallow Evaluation" arguing the benchmarks themselves are flawed. |
Same takeaways abound: the models are good and getting better slowly. The explosive gains come from systems built around them. Which, if you read Friday's newsletter, is exactly the thesis. |
|
|
|
|
| That's all for now. | | What'd you think of today's email? | |
|
|
P.S: Before you go… have you subscribed to our YouTube Channel? If not, can you? |
 | Click the image to subscribe! |
|
P.P.S: Love the newsletter, but only want to get it once per week? Don't unsubscribe—update your preferences here. |
No comments:
Post a Comment
Keep a civil tongue.