A working catalog
of tools
for the author.
AI tools built by a working creative, for working creatives. None of them produce the creative output. All of them assume the human is the author. Amplification, not replacement— you stay smarter than your tools.
I start with the eval.
Plant the failures, strip the answer key, gate the build in CI — and keep a human as the judge. Two shipped repos, the same move on different problems: discovery, and retrieval/grounding.
Anton Scout
Discovery — can it tell a real idea from hype? Decoy-injection eval: 1.0 decoy-catch / 0.82 recall over 27 candidates, CI-gated.
- Technique
- idea extraction + relevance scoring + LLM-judge eval
- Medium
- Open source · GitHub
Coherence Keeper
Retrieval & grounding — does the answer come from the source? Planted-contradiction eval lifts F1 from 0.50 (baseline) to 1.00, false-positive rate 0.0, CI-gated.
- Technique
- RAG (Cohere Embed + Rerank) + LLM judge + planted-contradiction eval
- Medium
- Open source · GitHub
In the world.
Shipped, running, or honestly in progress. The open-source entries link straight to source — nothing here is a mock-up.
- № 02
Jellybook
A real comic reader for Jellyfin — CBZ/CBR, two-page spreads with wide-page auto-detect, right-to-left manga mode, progress sync, neighbour preloading. Replaces the single-page default reader without touching ebook support.
Shipping- Technique
- Jellyfin plugin · .NET 9 · web injection
- Medium
- Open source · GitHub releases
View source → - № 03
Son of Anton
A three-machine Tailscale mesh (two GPUs + an always-on hub) running a tiered local-LLM router: each prompt goes to the smallest model that can answer it, escalating to a larger local model — or the cloud — only when it has to. It runs the local inference and overnight automation behind these projects.
Running- Technique
- tiered local-LLM routing + Tailscale mesh
- Medium
- Personal infrastructure · running in production
- № 04
Reprise
Adaptive learning that models your per-topic skill from your actual activity, then conditions what it shows you next — tell it 'too basic' or 'too advanced' and the model updates. The underlying loop is validated in my own daily use; the standalone desktop app is in progress.
In progress- Technique
- per-concept skill modelling + generation conditioning
- Medium
- Native desktop (Tauri + Rust) · local-first
On the bench.
Designed and specced, not yet built — listed honestly, not sold as live. Each ships in its own medium to its own audience when its turn comes.
- № 05
Lost Media Archive
Describe a cursed artifact — get a fake-wiki article and one 'screenshot' back. Clearly fictional. Designed to hold up on anything you throw at it.
On the bench- Technique
- context engineering + taste + vision
- Medium
- Web
- № 06
NPC Memory Engine — Bertram
An NPC with persistent secret state: remembers across sessions and lies consistently about something. Context engineering surfaced as gameplay, not a hidden back-end.
On the bench- Technique
- MCP + memory + persistent state
- Medium
- Unreal or Unity plugin + playable scene