Hey Hey! Tuesday again.

Here’s what my google search history is looking like post-thanksgiving:

Basically, I tumbled down some rabbit holes. And now you're coming with me.

Q1: To make a spring STIFFER, engineers use:
Thicker wire

Q2: To make a cantilever beam deflect LESS under load, engineers:
Double the height (deflection ∝ 1/h³)

Q3: In a pressure vessel, stress concentrates at:
Flat sections (why tanks are cylindrical)

Congratulations Smith, Jeff, Christian, Todd, David, Adam, and Gary!
Make sure to check your inboxes!

But first - Mystery gift! You know what to do.

Q1: High-bandwidth BCIs fail first due to…
A) heat
B) RF noise
C) power delivery

Q2: What limits neural bandwidth faster?
A) power
B) heat
C) tissue response

Q3: Add pinned atoms around a liquid metal droplet. Solidification…
A) speeds up
B) slows down
C) refuses to happen

Know the answers?
Reply HERE 👇

WANT TO REACH AN AUDIENCE OF OVER 63K ENGINEERS?
CLICK HERE TO SEE HOW WE CAN PARTNER TO HELP!

Quantum computers store information in atoms. To read or change that information, you need to hit each atom with a laser beam tuned to an exact frequency. 

The problem? Every laser needs a box-sized device (called a modulator) to control it. And quantum computers need thousands of lasers.
Commercial real estate agents loved this setup.

Researchers replaced the box-sized modulator with one that's 100 times smaller than a human hair (and uses 80 times less power).

It uses tiny vibrations (billions per second) to manipulate laser light with extreme precision. And because it's made using the same manufacturing process as computer chips, they can pack thousands onto a single piece of silicon.

Went from "we need three floors" to "fits in your wallet."

Trust physicists to ruin the lab equipment leasing industry.

Read more here

AI models need massive computing power spread across data centers. 

Problem? Those data centers don't talk to each other fast enough to function like a single unit. It's like trying to have a conference call where everyone's on a 5-second delay.

China just activated a nationwide optical network that links data centers across 1,243 miles and achieves 98% efficiency of a single data center - meaning scattered facilities now work almost like they're in the same room.

The system uses a "deterministic network" - think of it like a train schedule for data where every packet arrives exactly on time with guaranteed bandwidth and near-zero packet loss.

No "sorry, can you repeat that?" moments between packets.

They tested it against regular internet: a 72-terabyte dataset transferred 621 miles in 1.6 hours. Over normal internet? Nearly 2 years.

AI model training gets more than 2x faster - each iteration takes 16 seconds instead of 36+ seconds on standard networks.

Turns out the secret to better AI is building a just-in-time system for data.

Read more here

Scientists at Nottingham and Ulm universities melted tiny droplets of platinum, gold, and palladium on graphene and watched what happened atom by atom using an electron microscope. 

Some atoms kept moving (that is - turned liquid). But some atoms? They just stayed put, locked in place by defects in the graphene surface. The introverts of the atomic world.

This is what actually went down: when enough stationary atoms form a ring around the moving liquid atoms, they create an "atomic corral" that traps the liquid inside. The trapped liquid stays liquid even at temperatures way below its known freezing point.

This is the first time anyone's corralled atoms like this. Scientists have done it with photons and electrons before, but never with actual matter.

And if you change where those stationary atoms appear, you change when and how freezing happens.

The discovery could change how catalysts work - and this liquid-solid hybrid state might explain why some catalysts perform better than expected.

Turns out a few stubborn atoms can peer-pressure an entire liquid into ignoring physics.

Read more here

Researchers at the University of Pennsylvania and University of Michigan have developed the world’s smallest autonomous robot (less than 1 millimeter in size) that can sense, think, and act on its own.

Here's the weird thing about brain-computer interfaces (BCI’s): we've always had the software to decode what the brain is saying.

The problem was never the algorithms. It was getting enough signal OUT of the brain in the first place.

Current BCIs are bottlenecked by hardware that's too bulky, too hot, or too limited in how many neurons they can listen to at once. 

You've got these chunky electronics boxes mounted on people's skulls, wires threading through holes in bone, and maybe 100-200 channels if you're lucky. And the brain has 86 billion neurons firing constantly. 

We're listening with the equivalent of a tin can on a string.

Here came Stanford researchers who just said "screw it" and put the entire BCI - amplifiers, analog-to-digital converters, wireless radio, power management, everything - onto a single silicon chip thinner than a human hair.

This thing is 50 microns thick. And flexible enough to drape over your cortex like a contact lens sits on your eye.

On that one chip: 65,536 electrodes. 1,024 channels recording simultaneously. 16,384 stimulation channels. A 100 Mbps wireless radio streaming data out.

For context, current state-of-the-art BCIs have maybe 200-300 channels and stream at kilobits per second. This jumps straight to 1,024 channels at 100 megabits per second.

This can’t be called a new version, it’s entirely new tech altogether.

HOW THEY ACTUALLY BUILT THIS

The breakthrough is mixed-signal BCD processing - a manufacturing method that lets you put high-voltage analog circuits (for amplifying tiny neural signals) right next to dense digital logic (for processing and transmission) on the same piece of silicon.

In simple language, Neural Signals get amplified right where they’re recorded, digitized immediately and streamed wirelessly to a small wearable. And that wearable then provides power for the whole operation and forwards any data like a network device.

No skull-mounted electronics boxes that make you look like Magneto at an X-Men themed Halloween party. Just a chip sitting on your brain, talking to a relay in your pocket.

WHY BANDWIDTH IS THE ENTIRE GAME

With 1,024 channels and 100 Mbps data rate, you can finally ask questions that were impossible before:

The limitation was "Are we even giving it enough data to work with?"

Well, now we might just be.

ARE THESE CHIPS READY FOR MEDICAL USE YET?

Short Answer, nope.

More work needs to be done to overcome these pitfalls: Heat dissipation in living tissue, long-term biological response, and the difficulty of manufacturing flexible chips at medical-grade reliability. 

This isn't getting approved for clinical use next year. But the constraint has shifted. 

For the first time, a brain-computer interface feels less like a fragile research prototype and more like a scalable infrastructure, something you can manufacture, scale, and connect directly to modern AI.

The line between what the brain computes and what machines compute just became an engineering problem. 

And engineers are very good at solving those.

Mechanical Design Engineer - Plastic Extrusion Machinery LLC
Engineering translator turning customer wishes into 3D models that production won't curse at.
Apply Now

Senior Systems Engineer - Auriga Corporation
Train traffic cop making sure locomotives don't accidentally meet on the same track.
Apply Now

Autonomy Engineer III/IV - Zone 5 Technologies
Teaching drones to think for themselves without turning Skynet into reality.
Apply Now

WANT TO REACH AN AUDIENCE OF OVER 63K ENGINEERS?
CLICK HERE TO SEE HOW WE CAN PARTNER TO HELP!