Something strange happened on the way to the future. Building a company became easy. Knowing what to build it out of became hard.
There are more than thirty thousand AI tools you could run your startup on. Last year there were twenty thousand. Next year there will be more still, and most of this year's will be gone. Every week brings a new agent, a new model, a new "this changes everything." The Hacker News front page alone is enough to give you whiplash. Nobody knows. Everyone's building anyway.
Here is what we noticed, running AI-led ventures of our own: the tools kept changing. The jobs never did.
A company still needs to think. It still needs to build, to remember, to be seen, to sell, to answer its customers at 3 a.m., to keep its books, to prove it works, and to keep its founder sane. Those jobs are stable. They are, in a real sense, elemental. The vendors that do those jobs rotate like weather — but the jobs themselves have the permanence of chemistry.
So we did the obvious thing. We made the table.
Fifty-eight elements. Twelve groups. Twenty-eight essential. Everything your startup is made of, on one page.
Each element is a job, not a brand. Element 5 is the Coding Agent, not Cursor. Element 35 is the Support Agent, not whoever holds that title this quarter. When a tool is overtaken — and tools are always overtaken — the element stays exactly where it is, and the changelog records the succession. Your mental map survives the churn. That is the entire point.
The table is opinionated. We say which elements are essential and which are optional. We say which are stable and which are still gas. We name the five tools that hold each element right now, ranked, with the evidence and the date attached, and we say why, and we date the claim. No element's position can be bought. No tool pays to be listed. When something we're connected to appears on the table, it says so, plainly.
The table is versioned. Every quarter we issue a new edition, mark what changed, and retire what died. A reference that doesn't admit its age is a lie with a nice font. Ours wears its edition number on the front page.
And the table is bounded. It will not grow to thirty thousand entries, because the whole disease is thirty thousand entries. If a new element truly emerges — and a few are coming; you can see their dashed outlines in the table — something else must earn its place beside it. Scarcity is the curation. The meta-skill of the AI era is not collecting tools. It is knowing what to ignore.
One more thing. The most important elements on the table are not software. Judgment. Community. Taste. Your own health. In a company where agents do the execution, the human elements are load-bearing. We put them on the table because a stack that omits them fails in production, every time.
Most of a company has already been automated. The knowing-what-to-run part hasn't.
elems exists to change that.
— elems · Edition v2026.Q3 · Explore the table →