awkrad

Father and son team! The Bridge Memos.
Question everything.
Tech that becomes infrastructure.

[awkrad @awkrad]

What if the thing nobody understands isn't actually a company at all? In Memo 004 of The Bridge Memos, I go back to an idea I've been obsessing over for years: the strange connections between Tesla, SpaceX, Starlink, xAI, X, Optimus, Neuralink, The Boring Company, energy, chips, AI, and infrastructure. For years, I pictured it like one of those detective boards. Pictures. Pushpins. Yarn everywhere. I kept wondering if there was some giant master plan behind all of it. I don't think there is. I think something stranger may be happening. Problem. Solution. Capability. Next problem. Reusable rockets make cheap, frequent access to orbit possible. That capability helps make Starlink possible. Connectivity becomes useful for increasingly intelligent machines. Machines need intelligence. Intelligence needs enormous amounts of compute. Compute needs energy and chips. Chip supply becomes a bottleneck. So Tesla and SpaceX move toward semiconductor manufacturing. Compute runs into land, power and cooling constraints. So SpaceX explores putting AI compute in orbit. Which requires connectivity. And launch. And suddenly we're back where we started. Then there’s X. Human information. Communication. AI. Payments. Identity. Transactions. Maybe Twitter was never the destination either. But this is where I have to be careful. Because if you put enough pictures on a wall and run enough yarn between them, eventually you can make everything look connected. So in this Memo, I start cutting strings. Some of my old ideas were wrong. Some were too aggressive. Some connections are real. Others remain speculative. And some may never happen. But after cutting the bullshit, there's still an uncomfortable amount of yarn left. Maybe the more useful way to understand this isn't as a collection of companies at all. Maybe it's a capability stack: Intelligence. Compute. Chips. Energy. Connectivity. Launch. Machines. Human interaction. Transactions. Human-machine interfaces. Physical infrastructure. From one direction, that can look like extraordinary vertical integration — one capability solving the next bottleneck. From another direction, it looks like concentration — more critical infrastructure passing through fewer decision points. And efficiency and concentration can look remarkably similar depending on where you're standing. Which leads to a much harder question. If this system actually works… do we even want somebody to stop it? What around us today feels permanent — but isn't? Next: Memo 005 — Who Stops It?

[unZapped @Unzapped]

Will humans still work?

Not nobody-works. The messy middle — what we do when work winds down.

[unZapped @Unzapped]

Last memo asked who can tell them no.
This one asks who's left — and whether we even want them to.

You can vertically integrate almost anything.
You cannot vertically integrate consent.

Bridge Memos 005 — Who Stops It?
Full memo on unZappedTV (link in CTA).

[unZapped @Unzapped]

They don't want a spreadsheet.

They want which products are losing money. Where expenses jumped. The weird transactions. Something they can understand today.

Excel can still sit underneath everything. Hand intelligence the data and ask the question — the relationship changes.

So here's the Bridge Memos question: when the spreadsheet stops being the skill, what's the job now?

[awkrad @awkrad]

South Korea’s utility just asked Samsung to prepay about $15B for electricity through 2031 — and SK Hynix another few billion — so the grid can keep up with fabs and AI campuses.

That’s not regulation from outside the stack.
That’s government becoming a layer financed by the stack.

Vertical integration isn’t spreading because CEOs love org charts.
It’s spreading because the infrastructure underneath can’t move fast enough.

When the referee starts needing the players’ money to keep the lights on…
who’s left with a clean veto?

[awkrad @awkrad]

Austin’s throwing a Cybercab launch today.
Tesla’s own site still says those rides are “in the future.”

Same day, France starts its own FSD road tests instead of trusting Tesla’s or the Netherlands’ numbers.

You can build the intelligence once.
You still have to convince governments one at a time that it deserves permission.

That’s the quiet half of “who stops it?”
Not just who can say no —
who has to say yes, over and over.

[awkrad @awkrad]

@farzyness Yep. The next “software company” looks a lot like a factory with a nervous system.

[awkrad @awkrad]

@CuriousPejjy 10. On all Tesla/Spacex news as of late…

[awkrad @awkrad]

@farzyness Loop

[awkrad @awkrad]

@farzyness Door dash next? Once @Tesla_Optimus can ride in a Cybercab...

[awkrad @awkrad]

@elonmusk @bot Great job!

[awkrad @awkrad]

@farzyness @bot Nice! Did they change Grok bot limits?
https://t.co/ddbfi5chax

[awkrad @awkrad]

While the API fix may be urgent. More important will be to give Grok Bot the ability to "just do it".

Today, the need is to create human artifacts with provenance and publish everywhere... easy peasy to vibecode. But a vibe coded app is just a bridge, isn't it? I should be able to just point to a local video I created to Grok Bot, have it transcribe it, generate all publishing assets, Title, description, thumbnail, Chapters, establish provenance (human made), and post everywhere the human creator wants. It can do most but it fails at the 100mb limit. it can't actual use it's computer to post. Why waste resources on bridges to nowhere?

[awkrad @awkrad]

@Chef4Tnite Nobody in NY... 😆

[awkrad @awkrad]

While the API fix may be urgent. More important will be to give Grok Bot the ability to "just do it".

Today, the need is to create human artifacts with provenance and publish everywhere... easy peasy to vibecode. But a vibe coded app is just a bridge, isn't it? I should be able to just point to a local video I created to Grok Bot, have it transcribe it, generate all publishing assets, Title, description, thumbnail, Chapters, establish provenance (human made), and post everywhere the human creator wants. It can do most but it fails at the 100mb limit. it can't actual use it's computer to post. Why waste resources on bridges to nowhere?

[awkrad @awkrad]

@robotaxi GODSPEED!

[awkrad @awkrad]

We talk about Tesla, SpaceX, Starlink, xAI, and Energy like they’re separate bets.

They’re not competing with each other.
They’re completing each other.

Cars that lean on satellites.
Rockets that lean on power.
A model that needs both.

Memo 004’s quiet point:
if this is already one infrastructure system…
what happens when that system becomes the default path?

[awkrad @awkrad]

Today in Chapel Hill, Commerce sits down in person with Nvidia, OpenAI, Anthropic, and Palantir.

Yesterday that same G20 room got the light-touch pitch: don’t build a new AI referee.

Chips. Models. Cloud. State. Same table.

When the people who build the stack sit with the people who write the rules…

who’s left outside the room to say no?

[awkrad @awkrad]

Johnny Cab was a joke in a movie.
Plastic smile. One bit. Cute and dumb.

Then they built the real one.
No steering wheel. No pedals. Nobody in the front seat.

They added GROK right before Cybercab Launch... Coincidence? I THINK NOT!

Johnny Cab on steroids.

What happens when the ride home can talk back —
and nobody’s judging you?

[awkrad @awkrad]

Sears wasn’t killed by the internet.

It was killed when somebody else became the catalog.

The catalog wasn’t the business.
The catalog was the interface.

Kodak. Blockbuster. AOL.
Same pattern: the need stayed. The mechanism moved.

What are we still calling a “product” that’s actually just the old interface?

[awkrad @awkrad]

Same corridor week:

The US is at the G20 asking countries not to build a new AI referee.

OpenAI says its next model crossed a “critical” cyber threshold — so they’re restricting the strongest tools themselves.

Light-touch from government.
Hard limits from the company racing to ship.

So who is actually governing whom?

[awkrad @awkrad]

@elonmusk The most compelling this about this is the audience. Did they even understand what you said?