Game in the Brain

A raw update on Mneme, maps, burnout, and the economics of moving terra-tons

What Got Done Today

Mneme CE World Generator — The Feedback Firehose

My one very vocal book buyer has been stress-testing the generator and the bugs are surfacing fast. I’ve been treating his emails as a live QA stream. Today we closed, investigated, or spec’d a pile of issues from QA.md, which I also completely reorganized by status so the next agent doesn’t have to dig through 28 fixed bugs to find the 4 that are actually open.

Closed today:

• QA-034 — Depression Penalty Timing removed. The “Before Starport / After Starport” toggle is gone. After-starport recalculation is now the only path. Less UI noise, more consistent data.
• QA-032 — The 427 km / 0.18G dwarf world. Closed as stale pre-fix data. The new mass+density pipeline (QA-023) makes it physically impossible to generate such a combo. Service-worker cache ghost.
• QA-031 — Star colour names. No more raw hex codes. Stars now show “Orange-Red” or “Pale Blue” with a big colour swatch.
• QA-035/036 — 2D map fixes. Main world was missing from the map, and total planetary body counts excluded it. Both fixed; the map now renders “★ MAIN” correctly and ship totalBodies rolls include it.

Investigated today:

• QA-029 — “Why is everything Anarchy?” It’s not a bug. The getPowerStructure() table maps rolls ≤7 (58.33% of the 2D6 curve) to Anarchy. Neil’s experience is the statistically expected outcome of the current table design. The real fix is a CE Default vs. Mneme Default preset system, not a code patch. Documented in full in the reorganized QA.md.

Spec updates:

• QA-012 — Batch export limits raised. Approved to bump default to 1,000 and max to 9,999.
• QA.md reorganization. The file is now split into clear sections: Current Open Issues → Root Cause Analysis → Fixed Issues → Proposed Features → Investigations. I also added a Root Cause Analysis block that explains why QA-027, QA-028, and QA-030 are all symptoms of the same underlying problem: CE productivity assumptions vs. Mneme productivity assumptions.

The real monster lurking in the backlog is the economy engine. Neil’s feedback keeps hitting the same wall: the numbers don’t feel right. And he’s correct — they don’t, because Mneme is trying to model a world where a 400,000-person colony can generate “billions” in trade, while CE assumes 1970s–1990s incremental productivity curves. That model breaks when your setting includes agent swarms, robot labour, and Jovian Hammers.

The 2D Map Got Its Own Repo (Finally)

QA-033 was the map button breaking on GitHub Pages while working fine in local dev. The root cause turned out to be exactly what I suspected: Vite’s dev server helpfully redirects broken URLs, but GitHub Pages is a static host — it just 404s silently. The monorepo BASE_URL dance was fragile.

Fix: I extracted the entire 2D map into its own standalone repo:

• Repo:Game-in-the-Brain/2d-star-system-map
• Live URL:https://game-in-the-brain.github.io/2d-star-system-map/
• MWG commit:63a01d12 — the generator now opens the external URL instead of trying to resolve a relative path inside the monorepo.

The README and FRD in that repo include a browser console snippet you can paste straight into MWG DevTools to instantly open the current generated system in the map. No file exports needed.

Enter the 3D Interstellar Map

While fixing the 2D map, I started planning the next thing: a 3D sector map for Cepheus Engine.

Not orbits. Not planets. Just stars.

• Sol at the centre.
• Real stars out to 100 parsecs.
• Galactic orientation: +X coreward, +Y spinward, +Z north galactic pole.
• Click two stars, see a line and the distance in parsecs (switchable to light years).
• A persistent compass always pointing coreward/spinward/tailward/outward.
• A star-count slider so your phone doesn’t melt trying to render 10,000 emissive spheres.
• PWA from day one — cache the catalogues, work offline at the table.

I analysed four existing solar-system repos (solar-system-3d-threejs, solar-system-honzaap, solar-system-simone, Solar-System-3D) and distilled the thesis into three planning docs in /home/justin/opencode260220/3d-interstellar-map/:

• repoanalysis.md — architectural thesis
• frd.md — feature requirements (FR-001 through FR-010)
• qa.md — open issues tracker

The goal is a sector map where Sol is the anchor and you work outward. If the device can handle 100 pc, great. If not, slide it down to 50, 25, or 10. Green FPS = add more stars.

The Real Bottleneck: Me, and the Demographics Overhaul

I upgraded Kimi from Moderato to Allegro (5× output, basically) because I’m burning through code faster than I can think. The problem isn’t the tools. It’s that Mneme’s core economics need a ground-up rewrite and I can feel the burnout creeping in.

CE vs. Mneme Productivity

Cepheus Engine / Traveller assumes a gentle productivity slope — maybe 1970s-to-1990s style growth, decade over decade.

Mneme doesn’t live in that world. Mneme lives in:

• The age of Agents and Robots.
• The age of Space Industry, where one human + automation moves and converts 3× more material per person every 50 years.
• Compounding growth modeled roughly as Productivity × 3.3^(TL−8).

This is the difference between buying a car and buying an Orbiter. Between a house and an Interstellar Ship with a habitat ring. Between a construction crane and a Jovian Hammer skimming gas giants. Between a particle accelerator and an 800,000 km Great Serpent harvesting antimatter from radiation belts.

When humans move gigatons and terratons the way we move tons in 2026, the entire frame of reference shifts. Why terraform a world to Earth-likeness when you can just build a billion-habitat cluster? Why fight over a single habitable planet when you can spin up a ringworld in a Jovian orbit?

What This Means for the Generator

The audience is split:

• Some want classic Traveller/CE — gritty, slow, frontier ports, human-scale labour.
• Some want Mneme — post-scarcity-adjacent, automated, where a tiny colony can maintain a Class B starport because robots don’t sleep.

I need to build core mechanics that let the user (or referee) set the TL productivity curve themselves. CE Traveller’s slow 1960s–1990s pace should be the default, but Mneme’s high-compounding mode needs to be a toggle. A preset. A dial.

This ties directly into:

• Under Heaven Demographics 2030 to 2100
• Game Projects Wiki

The economics of Mneme are not bugs. They are design intent that currently lacks a UI.

Tooling Notes: Allegro, DeepSeek, and Antigravity

I moved Kimi to Allegro (~$49 vs Claude’s $20) and the token ceiling is massive — roughly 15× more context than Claude in a single pass. The throughput is absurd. I can feed it entire QA.md files, full repo analyses, and get back implemented code in the same breath.

But.

The IDE approach has friction. The terminal has advantages that the polished UI hides. I’m learning the “antigravity” workflow — using cheaper models (DeepSeek) for grunt work when prompted correctly by a smarter AI, then using Allegro for the synthesis and architecture.

DeepSeek is shockingly capable if you hand it a clean spec written by something with better reasoning. The cost difference is enormous. The trick is the prompt relay — don’t ask DeepSeek to design, ask it to execute a design you’ve already validated.

What’s Next (After I Rest)

I need to step back. The code storm is productive but unsustainable. When I return, the priority stack is:

1. Demographics & Productivity Economics Overhaul — the user-configurable TL curve. This is the big rock.
2. 3D Interstellar Map MVP — scaffold the repo, load the 10 pc catalogue, render Sol + 200 stars, selection + distance lines.
3. Second Brain / Vector DB — I need a hosted, open-source vector store for setting research. The wiki is good for publishing, but terrible for thinking. I need something that can hold the full density of Mneme’s setting material and let me query it semantically.

Links

Repo	Purpose	URL
Mneme CE World Generator	React PWA star-system generator	https://github.com/Game-in-the-Brain/Mneme-CE-World-Generator
2D Star System Map	Standalone animated 2D orbital map (extracted from MWG)	https://github.com/Game-in-the-Brain/2d-star-system-map
3D Interstellar Map	Planned Three.js sector map for Cepheus Engine	https://github.com/Game-in-the-Brain/3d-interstellar-map

If you’re reading this and wondering why Mneme feels different from classic Traveller — that’s the point. CE is a 1970s futures lens. Mneme is asking what happens when that curve goes exponential. The generator just needs a UI that lets you choose which timeline you’re playing in.