The second world
Pono — te reo Māori for truth — takes the contract Kano’s planets honour — coordinates, time, a causal generator, an honest oracle — and points it at the most causally recorded artifact most teams own: their repository. The result is a world an agent can stand on while it works on your code.
One thesis, two worlds
Kano proved the idea on rock. Pono proves it isn’t about rock.
Kano A seed grows the planet.
Pono A repository’s history grows the world — every commit already a cause.
Kano Point at a ridge and it explains the forces that raised it.
Pono Point at a file and it explains the decisions that shaped it.
Kano Rewind time: “was this valley wet three months ago?” has one true answer.
Pono Rewind history: “was this boundary load-bearing last month?” has one true answer.
Kano Every answer traces back to the world, never to a guess.
Pono Every answer traces back to a commit, never to a vibe.
What Pono is
Pono is a deterministic projection of a codebase’s own record. The derivation pipeline contains no AI: run it twice on the same history and you get the same world, bit for bit.
Files move, split, and get renamed. Pono keeps the thread — the thing you’re asking about stays the same thing across its whole life, so “why does this exist?” reaches all the way back to the commit that bore it.
Not semantic search, not a call graph, not “AI memory.” A queryable world where recent change, current structure, and the provenance of any part are facts an agent can look up — and cite.
Answers carry their anchors. When an agent says a crate is load-bearing or a decision is settled, the claim points at evidence you can open — or contradict.
We test it in public
We don’t ask you to take grounding on faith. We charge Pono rent — in writing, in advance, with the failure condition pinned before the data exists.
Most recently: identical twin agent sessions on our own repositories — same model, same questions, same commit — one grounded in Pono’s derived history, one not. The ungrounded twins invented facts, confidently. The grounded twins checked first. And the one false claim a grounded session did make is published at the same volume, because an instrument that only finds what you hoped for isn’t an instrument.
Where it lives
Pono grounds our own working sessions, every day.
Every agent session on our estate opens standing on Pono’s ground — and the readout above is how we hold that habit to account. For teams, Pono surfaces through Kete, our assurance platform for AI-assisted development (closed alpha), alongside Ārai, our open-source guardrail layer. The interrogation is the demo; the world-building is the product.
Pono is in daily internal use inside Taniwha and reaches teams through the Kete closed alpha. If your agents make confident claims about your codebase and you’d like those claims anchored to its history instead — we’d like to talk.