// about
Most people spend 8–12 hours researching AI hardware — reading conflicting Reddit threads, watching outdated YouTube videos, and still ending up with parts that don't play nice together.
We fixed that.
informitIV kits are compatibility-tested hardware bundles paired with one-click setup scripts. Every kit goes from box to running local AI in under an hour. No cloud. No subscriptions. No guesswork.
We also track the AI space daily — cutting through the noise to surface what actually matters.
// kits
The average AI hardware setup takes 8–12 hours of research. We've already done it. Every kit is compatibility-tested and ready to run.
Hardware list, setup script, and model recommendations — all in one place. Order the kit, run the script, start using local AI.
No API bills. No data leaving your machine. No rate limits. Local AI runs on your hardware, under your control, forever.
Apple Silicon local inference, out of the box.
The fastest path to running local AI models. M4 Mac mini + accessories + a setup script that installs Ollama, pulls Llama 3.1 8B, and wires up your agent stack. Zero cloud. Zero latency.
Installs Homebrew, Ollama, pulls llama3.1:8b + mistral, configures OpenClaw agent stack.
$832
RTX-accelerated local inference for serious workloads.
Purpose-built for running 30B–70B parameter models at speed. NVIDIA RTX 4070 Super + high-RAM mini PC + a Windows setup script that gets you running in one click.
Installs Ollama for Windows, CUDA drivers, pulls Mixtral 8x7B + Llama 3.1 70B Q4, configures agent stack.
$1,686
Offline AI on a $150 budget.
Raspberry Pi 5 running lightweight quantized models — completely offline, totally private. Perfect for home automation, private assistants, and edge inference experiments.
Installs llama.cpp on Raspberry Pi OS, pulls Phi-3 Mini Q4 + Gemma 2B, sets up local API endpoint.
$147
Multi-model, multi-agent. Built for builders.
Maximum RAM, maximum models. Mac mini M4 Pro 48GB + ultrawide + peripherals tuned for long sessions. Runs 5+ models simultaneously with full agent orchestration.
Full dev environment: Ollama + 5 models, OpenClaw agent stack, llama.cpp, Python ML env, VS Code extensions.
$2,255
// signal
An unsecured internal data store at Anthropic exposed the name of its next frontier model. Along with it: details of an invite-only CEO event. Mythos is expected to push significantly beyond Sonnet 4.6 on agentic reasoning.
A new compression algorithm shrinks LLM memory footprint by 6× with no degradation. If it holds in production, models requiring 48GB could run in 8GB. Big news for local inference.
Apple's Google deal grants complete Gemini access in Apple data centers — used to distill smaller, device-optimized AI models. This is how Apple Intelligence gets built without building a frontier model from scratch.
Nvidia's CEO declared AGI achieved this week. The problem: no agreed definition exists. Whether it's a scientific milestone or investor optics, the claim signals how fast AI's capabilities narrative is shifting.
// build log
No spam. Just signal. Unsubscribe anytime.
✓ You're in.
Flux will keep you posted.