Everything, ordered by whether anyone uses it.

Not ordered by what took the most work, or by what I am proudest of. Those orderings flatter the maker. This one is measured, and it moves.

The company catalogue, with what each thing costs, lives at neogenesis.app. This page is just the list.

Running

Live, in use, and ordered by real visitors over the last ninety days.

ToolPickRunning A decision engine for solo developers choosing SaaS. Priced from collected data rather than vendor copy. By a wide margin the most used thing here. toolpick.dev → CraftDeskRunning Design tools for people who are the entire studio. Small, single-purpose, no account required. craftdesk.neogenesis.app → ReviewLabRunning Korean shopping reviews compared on price and on what a product is actually bad at. People stay on this one longer than on anything else, which is a better signal than visitor count. review.neogenesis.app → AI ForgeRunning AI tooling and automation for solo developers. aiforge.neogenesis.app → DeployStackRunning DevOps tooling for people who have no DevOps team, because they are the team. deploystack.neogenesis.app → FinStackRunning Fintech tools for freelancers and one-person businesses. finstack.neogenesis.app → SellKitRunning E-commerce tools for solo sellers. sellkit.neogenesis.app → K-OTTRunning Ten agents argue about what you should watch tonight, then hand you one answer instead of a list. Fewer visitors than the tools, but they go deep — roughly ten pages each. kott.kr →

Early

Shipped, but without enough evidence yet to say whether they work.

Apps in TossShipping An app and game line that ships inside Toss. Selection, build, QA and store submission are agent work. It lives in someone else's app, so there is no link to hand you.

DaysLeftUnmeasured Your biological age, and how fast you are aging. The number moves with what you actually did, not with what you meant to do. It carries no analytics at all, so I genuinely do not know whether anyone uses it — which is my mistake, not a feature. daysleft.io → KoreanLLMEarly Independent Korean-language LLM evaluation, aimed at people who have to choose a model and then justify the choice to someone else. koreanllm.org → UR WRONGEarly Real arguments, one verdict. Fewer people than I expected, and the ones who do come stay a long time. ur-wrong.com →

Research

Independent work, published openly with a DOI. Not peer reviewed — read it as an argument with its evidence attached, not as a settled result. The least visited things on this page, and the ones I would keep doing anyway.

EthicaAIOpen science Amartya Sen's meta-ranking — preferences over preferences — formalized as a live mechanism in multi-agent reinforcement learning. Situational commitment turns out to be the only evolutionarily stable strategy. Dashboard ↗ Zenodo DOI ↗ Details

WhyLabOpen science Correlation was never the question. A causal decision engine driven by multi-agent debate, validated against IHDP, ACIC and Jobs. Dashboard ↗ Details

Operating OntologyRunning The company itself as a queryable graph instead of scattered documents. Every node carries the source it came from, so an agent can be asked to show its work. Details

Retired

Kept on the list on purpose. A portfolio containing only survivors is a portfolio with the evidence removed.

WebPilot EngineHanded over A text prompt turned into a walkable 3D space. Still online, but it lives in someone else's repository now and I am not the one maintaining it. Still up ↗ Details

HIVE MINDAbsorbed Listed here for a year as a product. It never was one — it is the runtime that builds everything above, and calling infrastructure a product is how a portfolio gets padded. It kept running; it stopped being a line item. Details

SoraAbsorbed A personal AI command centre, delisted for the same reason as HIVE MIND. It is how the work gets done, not a thing anyone else can use. Details