Individualist | deliberate vector | herder of goat | mower of lawn | chaotic good | animals love me | veteran 📐🐐
Opinions my own
Joined December 2024
- Tweets2.6K
- Following243
- Followers416
- Likes44.5K
Pinned Tweet
AI 'labs' need to come clean and start telling the public what AI is and isn't. They're not doing this because they are using that epistemic gap for their own benefit. Leadership is often filling in the blanks so the people who know less than you feel comfortable and included. That none of you are doing this is shameful and a missed opportunity - not to mention, public perception is biting you in the ass and you may be unaware. 🤨
A year is not a lot of time. It's hard to believe this is happening, but it is. I'm anxious to see how well the new world-model's training translated from FSD to the humanoid, and how they will integrate an LLM with that.
The hardware will be impressive, but it's the OS that's going to steal the show. 🤖
Elon Musk on Optimus:
"We've started early production. The growth of Optimus is an S curve, but the initial curve is a shallow slope. We have the vast majority of the 1 million unit/year line installed at Fremont. There's still some things we need to tweak about the design. There's still quite a lot of work to do with the AI in the robot. I would expect that the AI, the electromechanical design, reliability, usage life, expect these issues to be solved by the end of next year. I remain convinced Optimus will be the biggest product on Earth. I do want to temper expectations that we still have a lot to get through, it will take about a year to get through that."
Today's launch marked the second demo of this new platform. SpaceX has been quiet about the new offering, few details other than what you see here.
The concept is a reliable container for customers to manufacture in microgravity. SpaceX handles the details, customers supply the payload and time needed off-planet.
With starship. SpaceX is targeting $100 per kg cost to orbit. Pharma/biotech and semiconductors all have products that if manufactured in microgravity are worth many times that, some as much as $55k per kg.
The model works because the customer is paying primarily for microgravity dwell time and process conditions, not for bulk mass transport
The program is building muscle now and will eventually launch on Starship which can deliver ~100 tons to orbit each launch. This will be a big deal and it will make what is currently a rare or exotic product, common.
Mark my words, frontier 'labs' are going to inject slowing mechanisms into the user output and experience to slow down progress for users outside of 'Enterprise model access'. This is a form of extraction they cannot resist.
Search died because answers got cheap. What replaces search is not chat.
Chat is just a faster answer.
The thing that actually dies with cheap answers is dwell time - the interval where a mind is not allowed to conclude.
Universities, editors, labs, and good conversations sold that interval and pretended they sold knowledge.
Cheap intelligence removes the interval.
The next valuable instrument is not a smarter model.
It is structured delay: a machine that keeps the first fluent paragraph from being allowed to count as finished work. 🤔
In less than two years, all the wrongs done to entertainment franchises will be righted by everyday people, and especially the youth.
I imagine the franchises that people deemed most harmed will show up first, then the wave will arrive.
Thousands of versions of the same stories will spring out from nowhere, and the planet will ultimately decide what the story-line, character development, narrative, and outcome. A new kind of cultural artifact.
We have democratized filming (VHS), distribution (Internet), production (AI), and soon, telling the story - and because the attention vote is the only one that matters, the future belongs to the best contender.
Entertainment is being returned to the wild.
Let the games begin. 🎬
Musk quotes about Roadster:
2021 (on Rogan):
- "I want it to hover, and I was trying to figure out how to make this thing hover without, you know, killing people."
Feb 2024:
- “0-60 mph < 1 sec. And that is the least interesting part.”
- “There will never be another car like this, if you could even call it a car.”
- “I think it has a shot at being the most mind-blowing product demo of all time.”
June 2024:
- “The new Tesla Roadster can fly.”
Oct 2025 (on Rogan):
- “One thing I can guarantee is that this product demo will be unforgettable. Unforgettable.”
- “Whether it’s good or bad, it will be unforgettable.”
- “I think it has a shot at being the most memorable product unveil ever.”
- “My friend Peter Thiel once reflected that the future was supposed to have flying cars, but we don’t have flying cars. I think if Peter wants a flying car, we should be able to buy one.”
- “Like, is it even a car? I’m not sure. It looks like a car. Let’s just put it this way, it’s crazier than if you took all the James Bond cars and combined them.”
Sep 2026 (All-in pod):
- “It’ll be a banger. Excitement guaranteed. Success is not guaranteed, but excitement is. We actually need an audience to vouch for the fact that this is not AI.”
If this fucker does not fly or melt minds, it will be a huge disappointment. 🤨
Some things to keep my mind:
- A car hovering for a short time - like a hovercraft from the 70's - is not even remotely close to being: "most memorable product unveil ever" since we already forgot about them.
- A car using fan-assisted downforce for grip like the 1978 Brabham BT46B that won F1, is not novel or mind blowing. Another car with this tech does not fit: "There will never be another car like this"
- Using fan-assisted downforce for a gravity defying upside-down demo was performed by McMurtry Automotive with their electric Spéirling hypercar in April 2025. A repeat of this demo is not "I think it has a shot at being the most mind-blowing product demo of all time." Nobody even remembers that.
Those quotes are all statements made by a CEO of a public company, they can carry consequences - meaning: if this fucker does not fly, shareholder lawsuits will.
So, given all the claims made by the man himself and stacked up against actual history, I can reasonably assume that the Roadster is capable of doing something so wild and novel as meaningfully 'flying.'
Imagine how world changing that would be?
I've been working on open source intelligence platform that leverages a novel ontological and systems approach to pattern modeling. But that is not the interesting part. The interesting part is that the system is saying OpenAI will cease to exist within a 0 to 8 month time-frame. Let's see if it's right. 🍿
Right or wrong I will not delete this post.
"As the capability lattice proliferates across other labs, open efforts, and ordinary daily use, independent observation of what the systems actually do and do not do closes the lag. The extreme framing loses its ability to sustain exceptional structures around one entity. That is the dead-man-walking status: the instance ends in irrelevance and stasis while the underlying compute continues."
Here we are, standing at the precipice of the collapse of AI and it's re-birth as SI - and I don't mean a name change. Buckle up, buttercup.
It's easy to miss the gambit Musk is running here. Everyone is digging in the wrong place and he eggs them on. The best play in SI is not smarter frontier's, the best play is getting the world to zig when the actual future is a zag.
Really fun way to spend $10 as a kid tbh. I loved it!
And you do learn things if you play many video games.
My lowest point was years ago while going through chemotherapy. I had a very rough night. I had thrown up all I could that day, and the pains in my abdomen were excruciating. I lay in bed, unable to move enough to reach the meds on my dresser next to me, and I was alone.
Sometime after dark, a coyote showed up and began howling just outside my bedroom window. Usually they don't stay in one spot very long, but this one did, occasionally yipping and making noise. I spoke to it several times, thanking it for visiting me.
I don't remember what time I fell asleep, but I woke up to the coyote yipping around the time the sun came up. That beautiful creature stayed with me the entire night, keeping me company. So no, I will never forget who was there for me when I was at my lowest.
Tesla merging with SpaceX will be the modern equivalent to Bell Labs we've been missing for 50 years. Except this time it will be vertically integrated and mission driven with AI supplementing the future-forward research bits.
ENGINEERING NEWS 🗞️ Tesla has found a brilliant way to fight wireless charging's radio leak without new hardware ⚡️
Wireless EV charging works by having a coil in a floor pad beam a magnetic field up into a matching coil under the car. But not all of that field ends up where it should.
Some of it radiates outward as radio emissions, and the FCC's rules cap how much an inductive charger can emit. Until a charger stays under that cap, it can't be certified for sale.
Staying under it usually comes at a price. Tesla's patent application notes that emission limits can shrink the hardware design space, hold back performance, or even force a major redesign. And any energy that leaks away is energy that never reaches the battery.
The key insight lies in how the two coils interact. Their currents flow in roughly opposite directions, so their magnetic fields largely cancel each other out vertically. Sideways, though, they add up, and that horizontal sum is what a distant antenna picks up as leakage.
Tesla team models this leak as the combined effect of both coil currents, each weighted by how strongly its coil turns current into a magnetic field (its inductance). It even corrects for the slight tilt caused by the car's larger rear tires, which angles the car's coil relative to the floor.
That combined leak depends on the strength and timing of each current, and both are set by the charger's electronics. So it can be shrunk the same way noise-cancelling headphones work, by shaping one signal to better offset the other.
To find the right shaping, a controller takes emissions readings from a loop antenna, for example at 10, 20, and 30 meters away, and can extrapolate them to estimate the leak at 300 meters.
It then uses a circuit model to predict each coil's current from the knobs it can turn. Those include the DC voltage feeding the floor pad, and the switching circuits on both sides that chop power into rapid pulses to drive the coils: how long each pulse stays on, how fast they repeat, and how the two sides are timed against each other.
From there, the controller searches for the settings that minimize leakage while still delivering the power the battery asked for, staying within safe voltage and current limits, and keeping the transistors switching at the instant their voltage hits zero, which avoids wasted energy. It then steers the charger toward those settings.
The usual cure for a leaky charger is more hardware, like heavier shielding or a costly redesign. Tesla flips that around. Borrowing the trick behind noise-cancelling headphones, it turns the charger's two fields against each other, and the patent says it does this without sacrificing charging power or adding cost. The stray field that threatened certification ends up tamed by the very currents that created it.
An interesting aspect to consider here is how the web changes when a physical robot can access (and record) information the same way a human can. All the current content obfuscation, auth, and challenge/response protection mechanisms become useless. Why? Because the simple rule is either a human can access it (and now a robot), or it cannot be accessed at all. Existing protections were built on the architectural assumption that automated bots do not interact with the rendered page the same way a human does. Games are a fantastic training ground for this capability because of the rich pixel content and decision making.
