AI without the smoke and mirrors

Under the prompt.
Down to the metal.

The prompt box is the visible surface. The real system is model weights, quantization, memory bandwidth, accelerators, PCIe, networking, runtimes, identity, and the software that decides what the model is allowed to do.

magicthellm / stack
$ inspect ai-stack

application   agent / tools
runtime       llama.cpp / vLLM
model         70B Q4
memory        VRAM + KV cache
compute       GPU / CPU
fabric        PCIe / Ethernet
physical      power / cooling

status        abstraction removed
01 / Models

Parameter count is only the beginning.

A model's size tells you roughly how much state exists. It does not tell you whether it will fit, how fast it will run, or how much context you can afford.

FP16

Raw weights

A 70B model at FP16 is roughly 140 GB of weights before runtime overhead.

Q4

Quantization

At an idealized four bits per parameter, the same 70B model falls near 35 GB before overhead.

CTX

Context

Long prompts and conversations consume additional memory through the KV cache.

02 / Memory

VRAM is capacity.
Bandwidth is velocity.

Fitting the model is only step one. Autoregressive inference repeatedly moves model state through memory, so bandwidth can matter as much as raw compute.

CapacityCan the model fit?
BandwidthHow fast can weights move?
KV CacheHow much context can survive?
OffloadWhen PCIe becomes visible
memory-math.md
70B FP16
70B × 2 bytes
≈ 140 GB

70B Q4
70B × 0.5 bytes
≈ 35 GB

plus runtime overhead:
metadata
KV cache
buffers
batch state
context
03 / Hardware

Old GPUs can still be useful.

Not because they are secretly fast. Because large amounts of cheap VRAM can unlock models that simply do not fit on smaller consumer cards.

24G

Cheap capacity

Retired datacenter GPUs can offer unusually low acquisition cost per gigabyte of VRAM.

W

Power penalty

Older silicon often trades purchase price for heat, noise, and much higher energy use.

PCIe

Platform matters

Slots, lane width, cooling, power delivery, and chassis airflow can become more important than the GPU itself.

04 / Networking

100GbE only matters
when the workload earns it.

Single-node text inference moves very little network data. High-speed fabrics matter when the network becomes part of the compute path: model distribution, shared storage, multi-node inference, or distributed training.

10GbEplenty for many labs
40GbEuseful shared fabric
100GbEdistributed workloads
RDMA / RoCElatency-sensitive scale-out
network-check
single GPU inference
100GbE value: LOW

shared model storage
100GbE value: MAYBE

distributed training
100GbE value: HIGH

rule:
name the bottleneck
before buying bandwidth.
05 / Agents + Security

The model should not be the trust boundary.

Tool-enabled AI becomes infrastructure. That means identity, validation, least privilege, segmentation, and auditability matter more than clever prompts.

Identity

Authenticate the requester

Know which person, service, or process is asking the AI system to act.

Policy

Validate tool calls outside the LLM

Natural language output is not authorization.

Audit

Log intent, action, and result

AI that touches real systems should be observable like any other privileged service.

06 / Technical Philosophy
“The abstraction is useful. The leaks are where the engineering starts.”

MagicTheLLM follows the entire path from token to transistor because cost, performance, reliability, and security all live below the prompt box.