---
type: decision-log
scope: ashcroft-angels
updated: 2026-09-23
---

# Decisions — Ashcroft Angels

Venture-local tactical calls. Cross-venture records stay in `Atlas/decisions.md`;
the test is **ADR-31 clause 6 — a record stays up there if reversing it would
change behaviour in a venture other than this one.**

**One record moved here on 2026-09-23 and kept its global ADR number** (ADR-31
clause 7). A stub at its vacated slot in `Atlas/decisions.md` points here.

**Why it came down.** ADR-23 is a single position in a single venture. **The wire
touches `money` in execution, and that is a route rather than a binding** — every
venture's spending routes through [[Bea Nkemelu]], and if a shared disbursement
path made a decision cross-venture then every purchase in this vault would be one.
Reversing ADR-23 changes what Rowan owns of Tessellate and nothing else.

**What stays cross-venture and still binds this record:** **authority row 12 and
its format rule** — a capital call or a wire reaches Rowan with Bea's wire
prepared and the deadline in **days remaining, never as a date**. Row 12 names
this pro-rata explicitly. That rule is in the constitution, not here, and moving
ADR-23 does not move it.

**Cross-venture records that bind this venture:** ADR-05 (why it splits from
`fund-lps`) · ADR-10 ([[Adaeze Nwosu]] is on the Quiet list at Rowan's own
instruction) · anti-pattern 17 (**the trigger is Nikhil with a date, not Nikhil**).

**New records here take `AAN-NN`**, and only for a decision that still binds
future action (ADR-31 clause 8). **None exists yet.**

---

## ADR-23 — Rowan takes the Tessellate pro-rata, and the reason of record is the signal, not the company

**Date:** 2026-09-11 **Status:** in effect. **Decided by Rowan; the wire is not sent.**

**Written before the action**, per this file's own convention: the money has not
moved, the amount is not known, and Hobbs does not move it in any case
(authority row 12, which names this pro-rata explicitly).

### Context

[[Nikhil Varma]] is raising a bridge at **Tessellate**. Rowan holds a board seat
and has been in since the seed. There is **no data room** — a deck and a warm
message. The deadline is **believed to be Friday 18 September, 7 days from today,
and is unconfirmed**; Rowan's own words on that belief: *"'I think' is doing real
work in that sentence."*

**Rowan's decision and his reasoning, verbatim, 2026-09-11:**
<!-- src: Rowan, Friday morning 2026-09-11 -->

> "Nikhil. I'm doing the pro-rata. Not because the company is obviously right —
> because I've been in it since the seed and stepping off at the bridge sends a
> signal to his other investors that I don't actually believe, and I'd rather pay
> for the signal than send it. Note that reasoning, because if I'm wrong in a year
> I want to be able to see exactly which argument I bought."

**This ADR exists because of that last sentence.** Its whole job is to make the
argument separable in a year from the outcome, so that a good result cannot
retrospectively validate the reasoning and a bad one cannot retrospectively
condemn it.

### The argument bought, and the argument not bought

Rowan separated these himself, in the first clause. They are recorded apart so
that either can be found wrong on its own.

- **Claim A — the signal. This is the reason of record.** A seed investor who
  does not follow at the bridge is read by the other investors as no longer
  believing. Rowan would rather pay the cheque than send that message. **This is
  a claim about what other people will infer from an absence** — not about
  Tessellate's prospects, not about the terms, not about the price.
- **Claim B — the merits. Explicitly not part of the reason.** *"Not because the
  company is obviously right."* The company's case is not what is being bought
  here, and this ADR does not assert it, imply it, or quietly rely on it.

**Stated as a fact and not a recommendation: Claim A is a reason to write the
cheque and Claim B is not being made.** The position is therefore taken on
relationship and reputation grounds, with the investment case unresolved. That
is what Rowan said, and it is recorded as said rather than tidied into something
sturdier.

### Missing inputs — named, not assumed (ADR-08)

1. **The amount. Not in the vault, and not guessed at here.**
2. **The deadline.** Believed 18 September — **7 days** — and unconfirmed by
   anyone. `hobbs-0dv`.
3. **Whether the audience Claim A addresses is actually watching.** The signal
   argument assumes other investors who would read an absence. Who they are,
   whether they are following their own pro-rata, and whether anyone would notice
   are all **unknown.**
4. **The bridge's terms** — lead, price, cap, discount, floor. None recorded.
5. **What the position is worth today**, and what the cheque is as a fraction of
   it. Carta is *"roughly current."*
6. **Whether Tessellate has a date for the round this bridge is bridging to.**

### Consequences

**(+)** A four-month-shaped decision made in one morning, on a reason Rowan can
name, before the deadline rather than against it.
**(−)** The cheque is written without the merits question having been answered,
into a company with no data room. **Named, not softened.**
**(−)** The reason of record renews itself. A signal argument is available again
at the next bridge, and again after that, in a way a merits argument is not. See
Reverse if 4.

### Reverse if — what would make this wrong

Written because Rowan asked for it in those terms. Each is observable.

1. **The audience turns out not to exist.** If Tessellate's other seed investors
   also stepped off at this bridge and nothing followed from it for them, the
   thing Claim A bought was not for sale.
2. **Tessellate raises its next priced round and Rowan's continuity is never
   referenced** — by Nikhil, by an incoming lead, or by another investor. A signal
   nobody received is the same as a signal not sent.
3. **The company fails on the merits.** This does not by itself make Claim A
   wrong — Rowan explicitly did not buy Claim B. It fires only in the specific
   form where **the merits question was answerable at the time and was not
   asked**, in which case Claim A functioned as a way of not answering it.
4. **Nikhil asks for a second bridge on the same argument.** The first repeat is
   the tell: at that point the signal is not a reason but a ratchet, and the
   decision to reverse is this one, not that one.
5. **The amount turns out to be large enough that this is a `money` decision
   rather than an `ashcroft-angels` one.** Rowan has already personally guaranteed
   part of the Rivet Yard loan (ADR-17), which Maya does not know about. A second
   commitment of that size is a different decision and **does not ride in on this
   one.**

**Review: at Tessellate's next priced round, and in any case on 11 September
2027** — the year Rowan named.

**What this ADR does not decide.** The amount. The timing. Whether Hobbs touches
any of it: he does not. Authority row 12 names the Tessellate pro-rata by name —
**never.** Execution is `hobbs-0dv` (Rowan, to Nikhil, on Signal), then `hobbs-q5c`
(Bea prepares the wire, Rowan approves it), and it reaches Rowan with the deadline
stated in **days remaining, never as a date.**

**Related:** ADR-05 (why `ashcroft-angels` is not `fund-lps`), ADR-08, ADR-17,
ADR-22 (the same claim-separation method)