Sundar Pichai said on Lex Fridman's podcast recently:
“This is the worst technology will ever be.”
That's been looping in my head all week. It's such a clean way to look at progress — we're literally building upwards from the floor. And if this is the floor… damn, the ceiling is going to be insane.
This week, everything I came across — every founder I spoke to, every deal I read about, every late-night rabbit hole — made me feel like we're hitting some wild inflection point. There's a shift in how companies are built, how work gets done, and even who (or what) is doing that work.
Cursor (Anysphere) is such a great example. They just raised $900M at a $9.9B valuation, and a team of like 100 engineers. Half the Fortune 500 uses it! That's not just a good startup — that's proof that something fundamental is changing. They're not scaling like traditional companies. They're building a product around intelligence as the core resource. I'm obsessed with companies like this. Lean, sharp, absurdly leveraged.
Netic is another one I came across — an autonomous revenue engine for essential services. They just raised $20M. And the crazy part is that sales, which used to be this heavily human, relationship-driven function, is becoming almost fully automatable. Like, you feed it inputs, and it closes deals. That's so cool to me — and also a little unnerving in a way I can't stop thinking about.
But here's the direction I'm most personally excited about: physical agents. I genuinely think this is the next massive breakout space. I spoke to a founder this week who's working on using AI to run scientific labs — not just automating data analysis, but having AI plan and conduct experiments. That made me pause. Because it's not just “AI co-pilot” anymore — it's AI as the scientist. I think that's where we're headed. Labs, warehouses, healthcare settings — places with physical complexity and structured environments are going to be where AI shows up in real, tangible ways.
YC Companies like Sava and Boost Robotics are already pointing there. Boost's approach — combining dextrous manipulation with mobility — opens up so many use cases: healthcare robotics, dark stores, data center maintenance. It's like we're entering a world where you can plug in a smart agent to run a physical environment end-to-end.
A part of me still dreams about a future where we get humanoid UFC fights. Full-on personality-driven robot fighters with storylines. Tell me that wouldn't be a hit!
LuminX is also a signal of where this all goes. They just raised $5.5M to bring AI into logistics and warehousing. Inventory management, routing, fulfillment — all managed seamlessly, without any humans in the loop. Aerones is another one that stood out — they raised $62M for robotic wind turbine maintenance. A whole layer of the industrial world is quietly being taken over by machines, and we're barely talking about it.
Zooming out a bit, there's another thread that I think is way more powerful than people give credit to: using AI to reinvent old-school businesses through roll-ups. The idea is so simple it almost sounds dumb — buy people-heavy companies like law firms or agencies, replace the humans with AI wherever possible, massively improve the margins, use the extra cash to buy more firms, and repeat. But that's exactly why it works. The opportunity is hiding in plain sight. AI is so good at language manipulation — law, HR, contracts, outreach — and no one's applying it to these legacy markets at scale yet. That's going to change. And when it does, I think we'll see AI holding companies that make today's consultancies look like Blockbuster.
And we've got this other layer happening at the infrastructure level — AI companies like xAI are raising $5B in debt to build out compute infrastructure. Samsung is preloading Perplexity on its devices and investing in its $500M round. Meta wants AI to fully automate ad creation, cutting out agencies and giving small businesses tools that used to take entire teams. These aren't just product decisions. They're territory grabs. Everyone's trying to own the user layer — the device, the assistant, the interface. Because if you own the interface, you own the user. Humane.ai's collapse is proof that just “having AI” doesn't matter. They raised $200M, sold for scraps to HP. Why? Because the product didn't solve a real problem. AI isn't the point — it's the engine. You still need a vehicle worth driving.
Also, I feel we have to talk about how all the mass content generation through AI is going to flood the internet. Like, if every ad, article, video, and song is AI-generated — how do people stand out? Does the internet become noise? Or do we actually loop back to real-life stuff becoming premium again — print, events, face-to-face communities. Maybe digital overload forces a return to physical meaning. Maybe we all get nostalgic for what's real again.
Even in defense, we're seeing AI stretch into wild territory. Anduril raised $2.5B at a $30.5B valuation — building autonomous defense systems. I lowkey think the future of defense is straight out of sci-fi. AI-powered fighter jets? Autonomous mission planning? It's happening. And that space is finally shedding its “too scary to touch” reputation among VCs. This stuff excites me, not just because it's high stakes, but because it's pushing the boundary of what we can trust machines to do.
Recently, while building custom AI workflows for a VC fund, they asked if I could help them build an AI that searches for people in their network. While talking to a few other funds, I realized that there's a surprising gap in how VCs manage relationships. One fund was literally using an Excel sheet to track contacts. I ended up writing code for what essentially started looking like a version of Happenstance without even knowing it existed. Now that I've come across Happenstance, I love it!
I was talking to a YC startup this week called Cohesive AI who are building Agentic CRMs for blue collar businesses. The conversation made me realize the enormous extent to which AI Agents can change CRM, and just how good agents are at it!
This all brings me back to that Pichai quote. If this is the worst it's going to be — if these tools, these ideas, these experiments are just the bottom layer — then the stuff we'll see in the next few years is going to make today look primitive. We're moving into a world where intelligence — not capital, not labor — becomes the most valuable input. And that's a huge unlock.
If you're a founder building something — this is your time. We're not just investing in what works today. We're betting on the way things will work. And honestly, I've never been more excited to be doing this.
Let's build.
DMs open if you're working on something crazy or want to share your thoughts. I want to hear from you.
— Archit Lal, GP @ Atland Ventures