Roadmap
Direction, not dates.
Stages behind us carry the date they happened. Stages ahead carry the gate that unlocks them — the thing that has to be true before the next one can start. We do not publish a quarter for a qubit count.
-
2026 · done
Measure the hardware that exists.
Encoded-qubit and memory experiments on rented IBM Heron-class processors, as paying customers: a three-basis logical Bell fidelity with a confidence interval, 31 certified detection-code pairs, a repetition-code ladder, and the connectivity-tax negative. Pre-registered, raw shots archived, every caveat published on the ledger.
-
Now
Design the chip those measurements ask for.
A four-qubit superconducting design — schematics, simulation, layout — with every design target traceable to a measurement we published. The status report says what closed and what did not.
Gate the design package, reported in public — including the parts that did not close.
-
Next
Put it on silicon.
Tape-out and fabrication of the first device. Bootstrapped work ends at the foundry door; this stage is a funding and partnership question as much as an engineering one.
Gate foundry engagement, fridge access, and the capital to pay for both.
-
After that
Measure our own qubits.
First cooldown, first coherence numbers from a device we made — published with their raw data whatever they are. Until this stage we quote design targets, never measurements, for our own hardware.
Gate a fabricated die, in a fridge, with a working bring-up and calibration loop.
-
Then
Four to eight to sixteen.
Each generation gated on what the previous one measured, not on a calendar. Error correction — as opposed to the detection we run today — needs distances and qubit counts that only arrive at the end of this ladder.
Gate each generation’s measurements, published, before the next is committed.
When a stage completes it gets a date here and a write-up in the journal; if a stage is abandoned, that gets written up too.