Compute is componentized: 11 parody GPU cards (Ex-Mining P106, GFX 1060
Dumpster, Radish VII Furnace, RTX 4090 Gigachonk, H100 'The Vault',
B200 BigChungus...) install into carrier slots — Budget PC takes one,
the milk-crate rig four, the half rack eight with rack-only datacenter
cards. Cards carry tok/watts/VRAM/heat individually; total cluster VRAM
must hold the running model version or all token output starves.
Storage devices (SSD shoebox / NVMe hoard / NAS tower) gate training:
no checkpoint space, no progress. Network grows a switch layer with
per-switch throughput caps between devices and uplinks. Shop grouped
by category. Old saves migrate with equivalent card loadouts.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
You name your model at incorporation and release versions 1/1.5/2/3,
shipping each in mini/pro/MAX grades from the training bank (volume vs
margin vs prestige). Compute now splits three ways: slop, training, and
API hosting. A rotating market of third-party models (DeepSlop-236B,
Kimi-Chi-7B, SloPangu-Air...) offers hosting margins that crater ~28%
every time a hot new model drops. Hosting your own grades pays premium
margin and +10% views for MAX — but rolls distillation: a rival lab
clones your weights and guts that grade's margin until you re-ship it
on a newer version. Scraping ops cash ladder (puppeteer scripts →
proxy pool → residential botnet → captcha farm) buys training-rate
bonuses and immunity against escalating IP-ban outages.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three progression structures, three currencies. Model Lab (training
tokens): generation ladder bought from the bank, modality branch
(voice/image/video/AV — richer slop earns more but costs tokens,
bandwidth, watts), one exclusive fine-tune per generation, and a data
source choice: own slop trains 30% faster with a 35% model-collapse
gamble, scraped draws copyright lawsuits, licensed bills per token.
Corporate R&D (perk points from closed VC rounds + cash): Ops, Growth,
Legal branches with capstones Diesel Backup, THE CONVEYOR (burns 10%
of gross income for +50% views forever), Regulatory Capture. Sockets:
device inspector panel with on/off, sell, and two upgrade slots.
All effects flow through a single computeMods() bundle in research.ts.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Raise Seed through Series C for cash against views-delta mandates with
deadlines; success advances the round ladder, failure burns 25% of cash
and permanently dilutes income 15%. Four fictional rival archetypes fire
on a seeded timer: Doomsworth congressional panics (ALGO HEAT spikes),
Hypeman fake demos and Poogle algo tweaks (view debuffs), Freeman Weights
clone drops (API price crash) — with a paid 60/40 CLAP BACK gamble.
HUD: mandate progress banner, rival effect badges, total views stat.
Fix: scene texture list now derives from the catalog (new sprites were
rendering as purple placeholders).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Platforms adapt to your slop (ALGO HEAT): each publish raises spam-filter
wariness which throttles all view rates up to 85%; training a new model
tier slips past the filters. Verified loop: publish -> squeeze -> train ->
relief. New hardware: 20A sub-panel (+1500W), fiber uplink (1Gbps, 8
ports), Cash Furnace (burns $1.50/s for +40% views, breaker-immune).
Fix: New Game wipe no longer clobbered by the beforeunload autosave.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
