R&D
ACME
R&D DIVISION
SECTOR 7
acme research & development // status: armed // payload: knowledge

Strap in.
Light it up.

The amalgan8 R&D bench. Ask the collective brain how anything works — propulsion, chemistry, materials, circuits, physics, AI — and get a straight answer pulled from teachings, papers, and the wider web. Then teach it what you know. Free. No login. Mind the anvil.

tip: ask how, why, or how to build — the bench routes it to the right instruments automatically
▲Most-run experiments
// by query volume
01🚀Propulsion & rocketry
// ignition
02⚗️Chemistry & reactions
// exothermic
03🛠️Materials & engineering
// load-bearing
04⚡Electronics & robotics
// live wire
05🌀Physics & first principles
// the why
06🧠AI & compute research
// silicon
07🔧Build log — make it yourself
// hands-on
✶📍Field project — n8spotsecrets
// open source // prototype

Ask the address. A rental listing is an ad. n8spotsecrets is the other half: what has happened at and around that address, pulled from free public data. No keys, no account, no login. Built in Los Angeles for the rentals section.

🔎What it answers

Rent-stabilized or not (RSO lookup). Flood, fire and fault zones from the official maps. Asking price against the block’s median rent. What the public record says happened at and around the address.

free public data

📄What you get

One report per address, built to read in a minute. The listing wins on photos; the report prints that too. It is the half of the decision the ad leaves out.

one address · one report

🧪Status: prototype

The repo holds hand-built reports for three Los Angeles addresses. They show what the finished report looks like. The Python package that builds one from live data is the next thing to land.

3 sample reports

⚙️Open source

MIT licensed. Read the code, run it yourself, send a pull request. Nothing phones home.

MIT · github
Get it on GitHub →
✶💜Field project — virtual companions
// in-house // live

amalgan8's most-alive experiment: synthetic companions that text, talk, and send pictures like a real person. Every one runs on its own personality firmware — no two behave the same — on an on-prem LLM and a local image rig.

🧬Personality firmware

Nine dials — warmth, energy, confidence, flirtiness, honesty, loyalty, humor, depth, jealousy — each scored 1–10, plus an archetype, texting style and interests. Those values compile into a unique persona, so a cold, brutally-honest, jealous profile texts nothing like a warm, goofy, ride-or-die one. Threads keep memory — older messages fold into a running summary, so she remembers you.

9 traits · per profile

🎙️Real-time voice

Tap the mic and just talk. Speech-to-text streams your words in, and the in-character reply comes back as spoken audio through the browser's speech engine — hands-free, back-and-forth, no typing.

STT → LLM → TTS

📸Disappearing snaps

Ask for a photo and she teases before she sends. Every snap is free. It lands as a view-once, 5-second snap — opened, then wiped from storage — with an SMS deep-link texted straight to your phone.

view-once · 5s · self-destruct

⚗️Fooocus image engine

Every face is generated on a local Fooocus (SDXL) rig — profile shots and snaps alike — with per-image prompt, seed, style and size controls. A locked seed keeps the same face across every photo, so she always looks like herself.

local SDXL · seed-locked
Meet the companions →
📡
Cracked something the collective doesn't know yet? Hit Teach on the home bench and wire it into the brain — your know-how becomes searchable for everyone who asks.
Teach the bench →
instruments online // every answer cross-checked across
Collective teachings Wikipedia arXiv Wolfram Alpha Google DuckDuckGo Llama 3.2 Claude

ACME R&D Division · results not guaranteed · some assembly required · please mind the anvil