Weekly Dose of Optimism #184
Solar and Brains' Big Week, Zero Point Energy, Replit Agent 4 + Funding News + Nietzsche
Hi friends 👋,
Happy Friday! Here in New York City, it was the sunniest of times, it was the snowiest of times, but most of all, it was the incredibly hard to keep up with all of the incredible stuff getting done and funded of times.
Big week in robotics, solar, brains, world models, manufacturing, flight, AI, and even affordable luxury.
Let’s get to it.
Today’s Weekly Dose is brought to you by… Ramp Sheets
A few weeks ago, I wrote that I wrote that “a Ramp engineer with an AI will build something better than a CFO with an AI.” Ramp Sheets is exactly what I’m talking about.
See, a surprising amount of my high-leverage work ends up in a spreadsheet.
A deck comes in. A CSV export lands in my inbox. I want to benchmark a market, pressure-test a company, or run an analysis for a Deep. Dive. Usually, that means a lot of manual cleanup and model-building before I get to the actual insight.
Ramp Sheets skips a lot of that.
It’s an AI-powered spreadsheet from Ramp that lets you upload a PDF, pitch deck, CSV, or Excel file and ask it to do the work: build a model, research competitors, clean messy data, and help you get to an answer faster. For founders, investors, and operators, that means less spreadsheet mechanics and more actual thinking.
Best part: anyone can use Ramp Sheets today for free. You don’t need to be a Ramp customer to try it (but why wouldn’t you be?).
(1) Swift Solar Acquires Meyer Burger to Build Gigawatt-Scale Solar Factory
“If your technology is so good, why aren’t you using it to compete?”
In Power in the Age of Intelligence, I wrote that that’s a question I’d like to see more investors ask. If the technology you’re building with is so much better, why in the world are you just selling it to someone else as a component? Why not just make a better end product?
Joel Jean at Swift Solar said, “Ok bet.”
I met Joel a few years back on an introduction from his brother Neal, the co-founder and CEO of Beacons. Joel, like Neal, is a genius, with a PhD in electrical engineering from MIT, and he was putting that genius to work by making silicon-perovskite tandem solar cells work.
The promise of perovskite is that they’re much more efficient than solar panels, meaning they can convert more incoming sunlight into electricity. Current commercial silicon panels are 22% efficient, with a lab record of 26.8% and a theoretical limit of 29%. Swift Solar’s panels are already at 28% efficiency. China’s Longi, the world’s largest manufacturer of monocrystalline silicon wafers, recently hit 34.8% with a silicon-perovskite tandem solar cell, and the theoretical limit, for which Swift is gunning, is 45%.
More efficiency means more watts per panel, which means more watts per acre and a lower balance-of-system cost; 30–40% efficiency modules could cut solar system costs 20–40% while doubling power density.
This all sounds great, but the problem is that tandems tend to degrade quickly outside, where the sun is. Swift thinks that they’ve cracked that, and that China hasn’t yet, which gives the west an opportunity (certainly not guaranteed, probably not even likely, but an opportunity!) to leapfrog back over China in solar production.
So if you have the magic, degradation-resistant tandem cells, what do you do? Use it to compete.
Swift announced that it acquired the core assets of Europe’s leading solar manufacturer, the German company Meyer Burger, to “build the next generation vertically integrated US solar manufacturer.” Meyer Burger’s former CEO joined Swift to lead the manufacturing effort.
Solar panel manufacturing is one of the most cutthroat, competitive industries in the world. China keeps driving costs lower and lower and lower. It’s been all about scale. Meyer Burger itself went bankrupt after having sold solar cell manufacturing equipment to China, only for China to reverse engineer and mimic the technology.
But I love it. Pressing a technological advantage into vertical integration and competing directly is the kind of thing that gets my heart racing. And this echoes a deal that hits close to home. A few weeks into Puja’s time at Harry’s, she told me the 10-month-old startup was acquiring one of the world’s oldest and best blade manufacturers, Germany’s Feintechnik. That has been an incredibly successful partnership.
There’s something about a the US innovation / German manufacturing tandem.
(2) Reflect Orbital Built a Space Mirror
Ben Nowack for Reflect Orbital
Imagine you are a photon that does not want to get turned into usable electricity on earth. It’s been a really bad week for you.
First, Swift’s tandem cells. Before, there was a 78% chance you’d escape once you hit the panel, but that looks like it’s going to fall. But at least you have nighttime, right?
At night, you can just fly off into space at the speed of… well, you know. Run away from Earth as fast as is physically possible.
Not so fast, says Reflect Orbital. CEO Ben Nowack just announced that the company had made a solar mirror, one of four huge guys that will go up on each satellite, stop the photons’ getaway, and redirect them down to Earth.
Sometimes, they’ll be used to light up a concert or a remote worksite in the evening (imagine how much less creepy True Detective: Night Country would have been if Reflect existed then), sometimes, they’ll shine on crops to help them grow faster, and mostly, they’ll bounce into solar panels that capture them and turn them into electricity.
Efficiency is one of solar’s challenges, but a bigger one is one that seemed almot insurmountable: the sun only really shines strongly enough for max solar output for 4-6 hours a day. What if it could shine on solar farms - out in the desert, away from homes (where it wouldn’t imapact our sleep, Huberman) - round the clock?
Solar farm economics would get ridiculous, solar would get much cheaper, we’d get a lot more of it, and we’d be able to power all sorts of new, better electric things.
Mostly, though, I’m sharing because that video is a bright spot in a sea of sameness.
(3) The First Multi-Behavior Brain Upload
Dr. Alex Wissner-Gross for Eon Systems
As big a week as it was for solar, it may have been an even bigger week for brains.
A couple of weeks ago, an Australian company Cortical Labs grew brain cells on a microchip and taught it to do the first thing hackers try to do on any new computer, from TI-83s to ATMs to pregnancy tests to vape screens: play Doom. And yup, it could play Doom.
On Tuesday, Dr. Alex Wissner-Gross shared a video made by a company he works with called Eon Systems, that one above, that shows a virtual fruit fly being controlled by a brain with 140,000 neurons and 50 million synaptic connections running on a computer, mapped from a real fly’s actual connectome.
The Eon team took the adult Drosophila brain connectome, the complete wiring diagram of a real fly brain, painstakingly mapped by the broader neuroscience community, combined it with a physics-simulated fly body built from an X-ray scan of an actual fruit fly, and got the two to talk to each other. Then sensory information from the virtual world enters the brain model, the brain’s neurons issue commands, and the body moves. It does all of the things a fruit fly would do in response to stimuli. It navigates toward food by “taste.” It grooms itself when virtual dust activates its “antennal mechanosensory circuits.” It “eats.”
The team is careful not to oversell the accomplishment, and Dr. Doris Tsao at Astera Neuro has a good list of thoughts on connectonomics and uploads here which basically says, “This is awesome, but we’ll get further by understanding instead of blindly mapping, and anyway, it’s going to be really expensive to map bigger and more complex brains.”
But still… this is sci-fi stuff. I don’t even know if it’s good. But it’s sci-fi!
Meanwhile, in regular old science, two very cool brain papers:
Isotonic and minimally invasive optical clearing media for live cell imaging ex vivo and in vivo. A team of Japanese scientists basically figured out a way to “make brains see-through without killing them.” Kording Lab explained: “The mechanism of this is so obvious and simple and yet brilliant. Match the refractive index inside and outside of cells - and tissue becomes transparent. Because the scaling of scattering with object size, small things don't matter much. So cool!”
Scientists revive activity in frozen mouse brains for the first time. A team in Germany made progress on cryopreserving and thawing mouse brains that leaves some of the processes necessary for brain functioning like neuronal firing, cell metabolism, and brain plasticity intact. A researcher at another lab cautioned that we’re still a long way away from freezing ourselves and waking up in the future, but he also said, “This kind of progress is what gradually turns science fiction into scientific possibility.”
OK, so maybe all of this is just very sci-fi.
(4) Emergent Quantization from a dynamic vacuum, or Zero Point Energy
Harold “Sonny” White et al, published in APS Physical Review Research, via Andrew Côté
Maybe my favorite rabbit hole I’ve fallen down over the past few years is the Zero-Point Energy Rabbit Hole, which I discovered via American Alchemy Magazine.
The idea is that at absolute zero, classical physics says everything should stop moving, but the Heisenberg uncertainty principle prevents any particle from simultaneously having a precise position and zero momentum, so it has to keep jiggling. That constant jitter is ZPE.
The vacuum, which we think of as empty, is a seething foam of energy, with virtual particles constantly popping into and out of existence. This sounds like woo, but it’s been experimentally confirmed via the Casimir effect: two uncharged metal plates placed extremely close together in a vacuum attract each other, because ZPE fluctuations are slightly suppressed between the plates compared to outside, creating a net inward pressure. This has been measured in real labs like Los Alamos.
Quantum theory predicts, and experiments verify, that empty space contains an enormous amount of ZPE. The amount of energy supposedly packed into the vacuum would constitute a seemingly ubiquitous energy supply, a “Holy Grail” energy source, which is really just about as close to a “Holy Grail” as we can get. You can do a lot with unlimited energy.
Stuff gets wilder from there in a chain of logic that leads from ZPE to mass, inertia, and gravity as engineerable properties of the vacuum to the ability to actually engineer the vacuum to create phenomena like antigravity, which is what some believe Thomas Townsend Brown did.
Anyway, a team led by Dr. Harold “Sonny” White, a physicist and aerospace engineer with experience at Boeing, Lockheed Martin, and NASA who is now the CEO of ZPE company Casimir, showed that “Instead of random noise, the quantum vacuum can have structure, meaning waves, dispersion, resonances,” per Andrew Côté, who added that “their analytical model reproduces all the 'larger scale' predictions and measurements of quantum mechanics exactly.”
Andrew said that while the difference sounds subtle, “it means the vacuum can have local structure, that it is in effect a dynamic medium,” which means that, potentially, we can pull energy out of it. That is what Casimir is trying to do.
Now look, Casey Handmer replied betting $1 million that it wouldn’t work, and who am I to bet against Casey Handmer.
But 1) there are a lot of smart people on the other side of that bet, like Andrew, Sonny, Hal Puthoff, etc… and 2) even if there’s a 1% chance there’s something there, it’s hard to think of anything bigger.
Amjad Masad for Replit
Look, I’m biased. I’m an investor in Replit. I wrote about them in 2021 and predicted they’d be a $100 billion company. This week, they raised at a $9 billion valuation, so we’re getting there.
More importantly, though, they launched their new product, Agent 4. While I love Replit, and think Amjad is one of the best CEOs out there, I’ve been a little worried about vibe coding as a category. It’s so competitive, and won’t the big labs just do it? Won’t Claude Code kill everyone else?
Maybe. Certainly a possibility. But you should watch this video. By focusing maniacally on this one software creation product, the same thing they’ve been focused on making easy, from the infrastructure up, for over a decade, before AI was even a thing, they’ve built what I think is a truly differentiated product, and one that feels like one of the first good AI-native products I’ve seen.
You can collaborate in teams, set multiple agents to work at once, workshop designs, give feedback on specific ones, tweak them, iterate until you get something good, even doodle to show the AI what you want, and when you’ve found a design you like, you can keep it consistent across website, mobile app, presentations, docs, and more. It feels like a workspace for teams, where some of the teammates, the ones doing most of the grunt work at the humans’ direction, happen to be agents.
I’m going to be playing with it this weekend. If you have some time, build something.
EXTRA DOSE: Enormous Week of Fundraising + Nietzsche as Mystic for not boring world subscribers (cmon, join us)
Keep reading with a 7-day free trial
Subscribe to Not Boring by Packy McCormick to keep reading this post and get 7 days of free access to the full post archives.