technologybriefs
11:08in productionCh. 1 · A testbed, not a computer/ 11:08 · ceiling 15 min
Hardware · Tech history

Random-access memory

Random access wasn’t invented to speed up computers—it was built to prove memory could be both volatile and trustworthy enough to hold a program.

RAM began as an experiment in trust—not speed. The Williams tube proved volatile memory could reliably store instructions. Its mechanism was simple: charge spots, aim a beam, read or write. Its limits were severe: capacity under 1,000 bits, rapid decay, manual tuning. Yet it enabled the first stored-program computer. Core memory improved durability. DRAM enabled scale. RAM is not about density or bandwidth alone. It is about the moment engineers accepted volatility as a feature—not a flaw.

HardwareTech historyComputingHardwareInventionComputing across the network →
Chapters & takeaways4
  1. 1:08
    A testbed, not a computer

    The Manchester Baby wasn’t built to run programs—it was built to test whether the Williams tube could hold them.

  2. 3:14
    How a TV screen became memory

    Charged spots on a CRT face made random access possible—not because they were fast, but because the beam could jump anywhere.

  3. 5:05
    Tiny, fragile, and decisive

    A few hundred bits—smaller and faster than vacuum tubes—was enough to run the first stored program.

  4. 6:39
    Rings over rays

    Magnetic-core memory replaced the Williams tube not because it was faster, but because it was more reliable and rewritable.

Worth your time?

Yes. Study the whole thing.

4.5/ 5
What works
  • It enables arbitrary address access in near-constant time.
  • It replaced sequential-only memory, enabling stored-program architecture.
  • Its evolution—from CRT to core to DRAM—tracks the shift from experimental proof to mass-manufacturable component.
What does not
  • It does not deliver high density without trade-offs in refresh overhead or complexity.
  • It does not eliminate access latency variation across locations.
  • It does not decouple memory from process technology in early forms.
Study it if
  • Hardware architects evaluating memory hierarchy trade-offs.
  • Historians of computing infrastructure.
  • Students learning how abstraction layers emerge from physical constraints.
Skip it if
  • Software developers assuming uniform memory latency.
  • Cloud platform engineers sizing DRAM for production workloads.
  • AI model trainers selecting memory bandwidth for accelerators.
The written brief1 min read

What it is and the problem it solves

Random-access memory is hardware that allows any storage location to be read or written in roughly equal time. It solves the problem of rigid, sequential access in earlier memory forms like delay lines or relay banks.

How it works

The Williams tube stored data as electrically charged spots on a cathode-ray tube. Its electron beam read and wrote those spots in any order. Magnetic-core memory used magnetised rings arranged in a grid. DRAM uses one MOS transistor per capacitor.

What works

The Williams tube enabled truly random access via CRT beam positioning. It was smaller, faster, and more power-efficient than vacuum tube latches. It powered the Manchester Baby’s historic 1948 program run. Magnetic-core memory became widespread. DRAM’s single-transistor-per-bit design scaled silicon memory.

What does not

None of these technologies delivered scalable, reliable, high-density storage at low cost from day one. The Williams tube degraded quickly and required constant refresh. Core memory needed complex wiring and precise magnetic control. Early DRAM suffered yield and reliability issues.

What it changes

It changed computation from sequential, hardwired logic to programmable, stored-instruction machines. It enabled the Manchester Baby’s 1948 program run—the first electronically stored program—and made general-purpose digital computers physically feasible.

Is it worth your time

Yes—if you work on memory architecture, hardware history, or early computing systems. No—if you are evaluating modern memory for deployment, performance, or power: this is foundational context, not current practice.

Same field · Hardware4 of 198
9:56
Haptic technologyHaptic technology delivers programmed mechanical stimuli—not touch. It works where fidelity is secondary to signal: telerobotics, sensory research, and audio-triggered wearables. It fails where nuance matters: texture, temperature, pressure gradients. Its history is one of substitution, not replication.
9:27
Lidar1961Lidar began as Colidar: a satellite-tracking tool that measured distance by laser time-of-flight. It solved one problem well—precise, line-of-sight ranging—and nothing else. Its legacy is hardware lineage, not capability.
11:43
FM broadcastingEdwin Howard ArmstrongFM broadcasting is a noise-resilient analog transmission method that trades bandwidth for fidelity and static immunity. Armstrong engineered it deliberately—not as an evolution of AM, but as its functional alternative. Its mechanism works. Its early promise did not. Its regulatory impact did.
10:54
Chain driveHans RenoldThe chain drive is a mechanical transmission system using a roller chain and sprockets to move torque across distance. Hans Renold’s 1880 bush roller chain solved critical wear and slippage problems of earlier pin-and-link designs—and became the basis for all modern precision roller chains. It enables adjustable gear ratios and higher torque via duplex configurations. But it inherently produces speed variation (chordal effect) due to discrete link engagement. Its first commercial validation came in James Starley’s tricycles using his 1877 differential. It remains essential hardware—not legacy tech—where reliability, torque and spacing outweigh the need for constant velocity.
Up next in Technology

Raytheon

1922 · 11:08

Raytheon isn’t a tech company—it’s a missile factory with a balance sheet.

Play nextBack to Hardware
11:08