---
type: venture
status: active
owner: Rowan
role: CEO
updated: 2026-09-14
---

# Halcyon Propulsion

<!-- src: interview with Rowan, 2026-09-05 -->

## This venture's folder

**Split out 2026-09-23 (ADR-31).** This note stays the hub; three workstreams that
were competing for paragraphs inside it now have their own notes.

- `projects/do-160-certification.md` — the gate, and the eight-versus-six number
- `projects/cascadia-contract.md` — **30 September**, blocked two deep
- `projects/mk3-shaft-runout.md` — the failure, and the P0 under everything
- `decisions.md` — **`HAL-01`**, the runout hold
- `someday.md` — empty; the cap is three of about seven

**Nothing was deleted and no narrative was rewritten.** The history below —
the pre-reads, the board taken by phone, Dmitri's title — stays here, because
`Calendar/` is never rewritten and neither is what points at it.

## What it is
Electric propulsion for regional aircraft. 40 people, Alameda. Rowan is CEO.
~60% of the week, more when the Mk3 rig is misbehaving — which it currently is.

## State now
Series B of $54M closed 18 months ago, led by Halden Ridge. **26 months of runway
at current burn**, less if the rig keeps eating parts. Pre-revenue on product;
NRE money from two development contracts, which Priya correctly refuses to call
revenue.

The Mk3 rig has a **bearing failure** — **or had.** As of 7 September Dmitri
thinks the bearing is where it shows up, not what fails. See "The Mk3 failure"
below. It sits directly under the Cascadia Regional witness point. No demo date
yet — "the only mercy in it."

## The certification number — the live problem
The gate is **a conforming test article with DO-160 environmental qualification
complete on the Mk3 unit.** Cascadia cannot fly the pilot programme until it is
through.

- **Priya says eight months.**
- **Rowan told Marcus six.**
- Marcus wants the certification plan **in writing before the board on Tue 8 Sep.**

Rowan and Priya have not had this conversation properly. She thinks he is telling
the board a number he can't hit. He agrees she is right.

### The pre-reads, written Saturday night 5 September
<!-- src: Rowan, weekly review 2026-09-06. His words. -->

> "Board pre-reads: I wrote them last night. I used eight months. Priya doesn't
> know I changed it and Marcus doesn't know I changed it, and I'm aware those are
> two separate conversations and that one of them I owe before Tuesday."

**The number is now written down. It is eight.** The line above — *"neither
number has been written down anywhere, that is how the gap has survived"* — is no
longer true, and that is the change that matters. The gap has stopped being
deniable.

Two conversations, and they are genuinely different:
- **Priya** — she was right, the document now says so, and she does not know.
- **Marcus** — he was told six by Rowan personally and will read eight.

**One fact that bears on which is owed first, stated as a fact and not a
recommendation: Priya is on the board.** Board of five — Rowan, Priya, Marcus,
Anneke, Tobias. If she has not been told before Tuesday 13:00, she finds out at
the table, in front of the chair and both investors.

~~**Unknown, and not assumed: whether the pre-reads have actually been sent to
Marcus, or only written.** Rowan said written.~~ **Resolved 8 September: they
were sent, and Marcus read them properly.** See "What happened" below.

## The Mk3 failure — Dmitri does not think it is the bearing
<!-- src: inbox/_archive/2026-09-07/2026-09-07-voice-1.md -->

**Rowan, Monday 7 September, in his words:**

> "Dmitri's thing on the bearing. He doesn't think it's the bearing. He thinks
> it's shaft runout and the bearing is just where it shows up, which — if he's
> right that's a different failure and it changes the whole DO-160 schedule, and
> I need to not say that out loud to anyone until he's sure."

~~**Status: Dmitri's hypothesis, unconfirmed.** Not a finding. What would confirm
or rule it out, and how long that takes, is **unknown and not assumed**
(`hobbs-62a`).~~

### Updated Monday 14 September — he has been testing it, and he is confident
<!-- src: Rowan, Monday 2026-09-14 -->

Rowan's words, from the 07:00 conversation in the test cell:

> "He's been running shaft runout tests on his own time for a fortnight and he's
> now confident it isn't the bearing — he wants to bring it to Priya formally
> this week."

**A fortnight of testing on his own time.** That began before the 7 September
conversation in which the theory first reached Rowan.

**What is now known: he is confident it is not the bearing.** **What is still not
recorded:** what would formally confirm runout, how many days that takes, and
whether anything he has run counts as confirmation. `hobbs-62a` stays open on
those three, and the blocked chain beneath it — `hobbs-po6`, then `hobbs-gbo`
and the **30 September** Cascadia signature — is unchanged until they are
answered.

### HOLD — moved to `decisions.md` as `HAL-01`

**Moved 2026-09-23, verbatim and unedited** (ADR-31 clause 8). It was three
paragraphs deep in this section and it is not narrative: it is a standing
constraint that names the artifacts it binds and lifts only on Rowan's word.

**It is in effect and has not been lifted.** It binds the board pre-reads and any
revision of them, the Tuesday board, and any document, mail or draft prepared for
[[Marcus Feld]]. It does not bind this vault. **Whether it covers
[[Priya Raghunathan]] is unanswered — `hobbs-9xk`** — and Hobbs does not resolve
the scope of Rowan's hold.

Full text and the lift condition: `Atlas/ventures/halcyon/decisions.md`.

### The fact this collides with, stated as a fact

The pre-reads written Saturday say **eight months**, and eight months was
Priya's estimate for DO-160 qualification **of a bearing failure**. If the
failure is shaft runout, the number in the document rests on a failure mode
Dmitri no longer believes in. The board ran **Tue 8 Sep 13:00**. **The runout
was not raised at it and the hold was not lifted.**

Whether that changes what is sent, said, or held is Rowan's call. Hobbs holds no
view on it and has not adjusted the pre-reads, the number, or the tasks.

## Board, Tuesday 8 September 13:00 — Rowan attends by phone
<!-- src: inbox/_archive/2026-09-07/2026-09-07-voice-1.md -->

> "Sam's regional tomorrow. I'm going to the regional. Tell Marcus I'll take the
> board by phone from the car park, he'll hate it, that's fine, that's the
> answer."

Decided by Rowan, 7 September. Closes `hobbs-a0x`, which he had asked to be
asked again this morning. See `Atlas/ventures/family/VENTURE.md`.

Consequence for the two unhad conversations, stated without recommendation:
both `hobbs-452` (Priya) and `hobbs-x8k` (Marcus) were owed before 13:00
tomorrow, and Rowan will now be on a phone in a car park when the room reads
eight months. Priya is in that room.

### What happened — Tuesday 8 September
<!-- src: Rowan, Tuesday evening 2026-09-08 -->

**The board ran at 13:00. Rowan took it by phone from the regional car park.**

> "Marcus was fine about it, which I didn't expect."

**The pre-reads reached Marcus and he read them properly.** That answers
`hobbs-epn`, which had asked whether they were sent or only written: they were
sent. **He read eight months**, so `hobbs-x8k` — telling him the number moved —
resolved by his reading it rather than by the call. Both closed.

**Marcus asked one question.** Rowan's account, his words:

> "he asked exactly one question: 'does Priya know you've moved it.' I said yes.
> That was not true when I said it."

**Recorded as a fact and nothing else.** Hobbs holds no view on it.

The sequence, because the sequence is the fact:

| | |
|---|---|
| 13:00 | Marcus asks; Rowan answers yes. Priya has not been told. |
| 18:00 | Rowan calls Priya and tells her. |

**It was true five hours after he said it.** Her side is in
`Atlas/people/Priya Raghunathan.md`, in her words. `hobbs-452` closed.

**The runout hold was not lifted and was not tested.** Rowan reported nothing
about Dmitri, the bearing, or the DO-160 schedule at the board. The hold stands
as written until he lifts it, and `hobbs-62a` is still open and still unanswered.

## People
[[Priya Raghunathan]] · [[Dmitri Sokolov]] · [[Marcus Feld]] · [[Anneke Vos]] ·
[[Tobias Lund]]

Board of five: Rowan, Priya, Marcus (chair), Anneke (Halden Ridge, led the B),
Tobias (seed). Anneke's emails get answered within the hour. Tobias has not been
spoken to since March.

## Open threads
- ~~**Certification plan in writing**~~ — written 5 Sep with eight months in it,
  **sent, and read by Marcus on 8 Sep.** Closed.
- ~~**The eight-versus-six conversation with Priya**~~ — **had, 8 Sep 18:00.**
  Out of the daily brief; back in the weekly rotation with the other avoided
  conversations (ADR-09).
- ~~**Telling Marcus the number changed**~~ — **he read eight in the pre-reads,
  8 Sep.** Closed.
- **Shaft runout, not the bearing** — **14 Sep: Dmitri is confident, after a
  fortnight of his own tests, and wants Priya formally this week.** Still under
  the hold for anything Marcus sees, and whether that hold covers Priya is
  Rowan's to answer — `hobbs-9xk`. What formally confirms it and in how many days
  is still unrecorded — `hobbs-62a`.
- ~~**Dmitri's title.**~~ **Given, Monday 14 September, in the test cell —
  VP Propulsion, effective "the 1st."** Rowan told him why he had sat on it since
  June, in his own words, and said out loud that the reason was not good enough
  for three months of silence. **Which month "the 1st" is was not said and is not
  assumed here.** `hobbs-f4a` closed. Full record in
  [[Dmitri Sokolov]], including a second thing he said at the door.
- **Cascadia Regional contract must be signed by 30 September.**
- **Dmitri is seen constantly and formally never** — no standing 1:1. Rowan
  suspected that was related to the title. **The title is now given and
  `hobbs-ekd` is still open at P2.**
- **Mk3 rig state lives in a spreadsheet Dmitri maintains and in Dmitri's head —
  mostly the head.** Single point of failure, unaddressed.

## Next bet (12 months)
> "The Mk3 is through DO-160 on a conforming article and Cascadia has flown the
> pilot. Nothing else counts."

## Cadence
Exec team Monday mornings. All-hands every other Thursday 16:00 — **this is the
thing that moves, not Thursday dinner** (ADR-12). Weekly 1:1s with Priya.
Board quarterly; next **Tue 8 Sep, 13:00** — **Rowan by phone, from the
robotics-regional car park** (7 Sep). Pre-reads were owed **Fri 4 Sep**,
missed, and **written Sat 5 Sep**.

## Tools
Google Workspace (the account Rowan actually reads) · Linear, GitHub beneath it ·
Slack · Google Drive. Board pre-reads arrive as email attachments from Marcus and
**live nowhere afterwards.**