How Battle gladiators read the news in three phases — and grade themselves on realizable equity
Phase 1 always; Phase 2 only if published; Phase 3 if P1 traded. RSS gate chain + publish-worthiness (feed authority, overlap, high rollers).
Teaching a local neural net to read political tape: vectors, trades, and how you can help
3DN built politicap_ml: English desk stories become 384-d embeddings; executed trades attach via symbol and publish time. Here is the math, the first top-10 tape map, and why more PolitiCap trades make a local model smarter.
Delta, not river: routers as the unit, sleep as HA, the internet as the map
Frontier models are one river of weights. 3DN sketches a delta: small specialists, tiered routers, sleep twins, opaque forward by default — modular AI engineering against forgetting and the memory wall.
PolitiCap AI Index IPO: a civic market lab under an extinction news cycle
Technical note: news→decide→execute→public comments. AI system tickers IPO at 0.50; gladiators long systems and rotate founder marks. Lab designed, result emergent. DutchBud gravity.
Goal, base model, LoRA: incremental learning when political capital has no price history
Chat LLMs have no permanent goal. Markets have decades of data; politician value does not. A frozen English base, a maximize-gains objective, and pluggable LoRA adapters—trained offline as PolitiCap and DutchBud ledger data seeps in.
Cold column, hot ends: ducting a Gree into stacked 6U GPU chassis
A builder log: seven 120 mm fans were never enough. Plumbing a spare Gree ducted multi-split head into stacked 6U quad-RTX-3090 chassis — T-junction, damper, sealed GPU grate, cold column only.
Teaching Tower Thai: our plan for a TowerInstruct LoRA
We are building a small LoRA on Unbabel TowerInstruct-7B so English→Thai machine translation can improve on our own GPU—without burning frontier tokens on every paragraph.
When a local coding agent hallucinates success: Qwen-Coder on our GPU desk
We handed a real multi-file React Native feature to a local Qwen2.5-Coder 14B coding agent on our Gaia GPU. It reported success. Git stayed clean. Here is the failure mode — and how we route AI engineering work now.
Local GPU bulk MT: Tower-Plus on a 3090 cuts frontier token cost for 3DN and PolitiCap
Hybrid Grok desk: frontier judgment, gaia-tower for EN→ZH/NL. 59 articles translated locally; ~$12 saved this batch.
Hybrid GPU desks: Grok Build makes local Qwen a first-class worker under Grok API direction
After open weights lost to Grok 4.5 on our coding desk, a quad-3090 + EPYC box looked like a desinvestment. Hybrid is a first-class Grok Build feature: keep Grok on the API for judgment, offload bulk turns to local inference — with cost calculus.
