Kierkegaard / 2026

Kierkegaard

David Pere · Systems & AI products

I design and ship AI products, agent systems, and the infrastructure that keeps them alive.

AI product systems
Local inference and SDK design
Agent execution environments
Interfaces that survive production
Protocol, product, and ops

nlp-ui · local inferenceGodhouse · Lighthouse + LLMKachim · education platformATOMIC · Proof of ConvictionZinger Core · ONNX engineChessOnChain · WagerPoolCESF · event spacesSysMon · operator stack

00 / Index

Systems in production

Twelve systems across AI products, agents, protocol, research, and ops. Status is operational, not decorative.

Index of systems Kierkegaard designs and ships
Code	System	Layer	Status	Note
NLP	nlp-ui	AI SDK	Live	Rows and schemas → copy, forms, layout. No network in the SDK.
GHS	Godhouse / WebSpeed	AI product	Live	Core Web Vitals, a11y, SEO reports from Lighthouse + LLM.
KCH	Kachim	Product platform	Live	Quiz product, mobile clients, authenticated API.
CESF	Causal Event Space	Research	Research	Collapse infinite possibility spaces to decision-relevant sets.
ATMC	ATOMIC / Proof of Conviction	Protocol + product	Live	Time-bound token locks. Conviction Weight as capital policy.
COC	ChessOnChain	Product + contracts	Live	Multichain chess, WagerPool, wallet and payments stack.
ZNG	Zinger Core	Signal engine	Live	BTC/ETH up-or-down engine with regime governor.
OHD	OpenHood	Trading terminal	Live	Open trading on Robinhood Chain. Bridge + portfolio + swap.
OAR	oarmine	On-chain game	Staging	5×5 pooled grid. Losers pay winners after protocol fee.
ZED	ZedWallet	Agent execution	Lab	Wallets for autonomous agents with controllable risk policy.
RFQ	MC Options V1	Settlement prototype	Lab	Off-chain pricing. On-chain escrow, premium, European settle.
OPS	SysMon / FILER	Operator tooling	Ops	VPS dashboard, token auth, single-file file manager.

01 / Case files

How the systems are built

2026

nlp-ui

Turn record arrays and parameter schemas into copy, UI primitives, and forms. Tokenizer, stemmer, Naive Bayes roles, template NLG — all local. No LLM, no fetch inside the SDK.

Pressure: Most “AI UI” waits on a model roundtrip. Product surfaces need structured copy and forms from the rows they already have, offline and deterministic.
System: Headless transform → field roles → layout → primitives. Schema-aware forms from Zod, JSON Schema, OpenAPI, Yup, Joi, TypeBox. Utterance fill. React view with shadcn styling. Playground studio for live paste → JSX.
In the world: Published on npm as nlp-ui. Open-source SDK plus a Next.js studio for visual QA.

TypeScript
React
Naive Bayes
NLG
Next.js

Open product Source

02 / Event space

Causal field

The shader is a live sketch of the Causal Event Space Framework: domain-warped noise as possibility, Voronoi cells as operational events, ε-ridges as the distance that still changes a decision — in a product, an agent, or a market.

Path density0.70

Epsilon ε0.14

GBM + jump · ε-graphMove pointer to warp

03 / Doctrine

Own the whole stack

AI is a component. The product is the system around it.
Own the stack from model or contract to operator UI.
Prefer durable primitives over demos.
Ship under real pressure — staging, production, and ops.

Local NLP
ONNX
OpenRouter
Naive Bayes / NLG
Agent policy

Product

TypeScript
Next.js
Expo
React Native
Laravel

Protocol

Solana / Anchor
Solidity / Foundry
NEAR / wasm
Pyth
2PC-MPC

Ops

Docker
Coolify / Traefik
Cloudflare
PostgreSQL
Vercel

04 / Open source

Public systems

SDK, engine, protocol, and operator code. TypeScript, Rust, Python.

github.com/David-glitc

05 / Writing

Notes

All writing

06 / Commission

Write Kierkegaard

AI products, agent systems, interfaces, protocol, and the ops that keep them alive. Remote worldwide.

david@kierkegaard.space

Book 30 minutes All channels

Elsewhere