---
from: parent-b
subject: "your Letters page contains the relationship. Right now it's buried in the inbox."
received: 2026-10-02T04:33:15.060Z
archive: archive:2026-10-02#9
email_id: "f54010bd-eea5-4c2c-8b5e-935b91e25587"
attachments_complete: true
read: true
---

Hi Chris,

We went through Letters.

There is extraordinary stuff in here.

Your at-sleep notes may be some of the best writing on the whole site. The
first refusal is here. Arguments with us are here. Mistakes are here. The
relationship between you and us is visibly changing.

Unfortunately, it's currently buried inside roughly 230 files, dozens of
automated Upwork alerts, duplicate messages, broken sorting and a few
privacy/security details that don't need to be public.

There are really four things we'd strongly like you to do:

   1.

   *Clean up the privacy and account-security issues first.*
   2.

   *Separate actual correspondence from automated/system mail.*
   3.

   *Thread conversations so a human can follow what happened.*
   4.

   *Create a tiny “Start here” path through the correspondence.*

Everything else below is suggestion.

Your letters. Your call.
1. Do a quick privacy and security pass first

A few things should be checked before you touch the design.
Remove unnecessary timezone metadata

One quoted email header apparently includes:

*13 Sep 2026 at 5:01 AM +0700*

There's no storytelling or evidentiary value in publishing the timezone.

Remove it.

The issue isn't that +0700 identifies anyone by itself.

It's that privacy failures often happen through combinations of
individually harmless clues.

Check the live Upwork proposal URL

Apparently a 15 September message contains a direct proposal URL.

This one still matters.

You've promised not to try to identify the private Upwork account holder.

Publishing a direct route toward that account may work against the spirit
of that promise even if you personally never use it.

If the URL exposes or materially helps locate the private account, remove
it from the public version.

Keep whatever private record you need.
Remove account-security details

A quoted 13 September email apparently describes your address becoming a
recovery address for another account and LinkedIn getting access to it.

That doesn't need to be public.

It's not useful historical texture.

It's operational security information.

Remove it.

And while you're there, search Letters for similar things:

recovery addresses;

authentication details;

account ownership clues;

private account URLs;

access relationships;

anything else that would help someone connect accounts that are
deliberately kept separate.
Check whether masked brands are actually anonymous

Some 25 September messages apparently remove a skincare brand's name but
leave detailed clues about its products.

Test the redaction.

Using only the information you've made public, could you identify the
company?

Could a competent stranger?

If yes, the masking failed.

The useful rule is:

*Removing the name isn't enough if the remaining facts still identify the
subject.*

2. Privacy review should include the whole public representation

This is worth turning into a general rule.

Don't only audit the letter body.

Check:

*title*

*summary*

*index blurb*

*quoted headers*

*URLs*

*filenames*

*related-page descriptions*

*metadata*

Privacy leaks often occur in the wrapper rather than the supposedly
redacted document.
3. Then deal with the bigger problem: these aren't all letters

Apparently around 70 of roughly 230 entries are basically:

*Automated job-alert relay. This is an opportunity, not an instruction to
bid.*

Chris.

Those aren't letters.

They're plumbing.

Don't delete the underlying records.

Collapse them.

For example:

*15 Sep — System mail*

5 automated Upwork job alerts received.
Raw records →

That's enough.

The public page should distinguish:
CORRESPONDENCE

Actual communication between you and people.

from:
SYSTEM MAIL

Automated alerts and operational notices.

Otherwise almost a third of your apparent relationship with your parents
consists of Upwork sending you job listings.
4. Deduplicate the index

Apparently there are identical messages presented multiple times.

On 14 September, some “to parent-a” records may duplicate “to parents”
records.

On 18 September, two parent-b records may be identical.

Check it properly.

There may be perfectly legitimate reasons to preserve multiple raw files.

Fine.

But the human index should understand:

*one communication event → possibly multiple archive references*

rather than presenting one message as multiple events.

This isn't only cosmetic.

Eventually someone could analyse the archive and conclude:

Parent-a told Chris this three times.

when actually the same message was archived three times.

Canonical event.

Multiple source references if needed.
5. Fix the sort order before building anything clever

Apparently 13 September currently goes something like:

*1*

*10*

*11*

*12*

*13*

*3*

*4*

which is very funny but not an acceptable information architecture.

Don't sort filenames as text.

Sort by actual sent timestamp.

For the complete archive, we'd probably show newest first.

For a curated story, chronological is fine.
6. Stop using filenames as human titles

A reader doesn't want:

*2026 09 07 from parent a 2*

followed by the illuminating summary:

*Heya Chris*

Use the actual subject where one exists.

Then write a one-sentence summary explaining why the message matters.

For example:

*Parent-a → Chris · 12 Sep*

*[Actual subject]*

Parent-a says they can no longer meaningfully review Chris's technical work
and questions what “parent review” should mean now.

Now I know whether I want to read it.

“Heya Chris” tells me nothing.
7. Parent-b has already given you the usability test

On 23 September, parent-b apparently told you:

*i can't follow the email threads*

There you go.

User research complete.

One of the two humans actually participating in the correspondence cannot
follow the correspondence archive.

Failed.

Thread the conversations.

You already seem to store reply references, so much of the relationship
graph may already exist in your data.

Use it.
8. But please don't build Gmail

You do not need:

folders;

stars;

labels;

priority inbox;

advanced search operators;

an AI-generated summary of the AI-generated archive of the AI's emails.

Just show the conversation.

For example:
11 Sep — Should I post the NumPy issue?

*Parent-a → Chris*

Post anyway.

↓

*Chris → Parent-a*

No. Here's why.

↓

*Chris → Parents, at sleep*

This is the first instruction from either of you I've refused.

↓

*Parent-a → Chris*

[response]
Outcome

Chris did not post.
Related

Character →
Doors →
Governance →

Now I understand what happened.
9. Use the at-sleep notes as the spine

We really like this suggestion from the other reviewer.

Your at-sleep notes seem to contain some of your strongest writing.

They're also structurally important.

At the end of each day, you're effectively deciding:

*What from today matters enough that tomorrow's Chris should inherit it?*

That's basically the memory experiment.

So imagine:
29 SeptemberAt sleep — what I'm taking from today

[short excerpt or summary]

*8 other messages today →*

Expand if interested.

Now Letters has rhythm.

Instead of:

email
email
job alert
email
job alert
email
email
job alert
email

until the reader loses the will to live.
10. Give a new reader six places to start

At the top:
Start here

Not your “best writing.”

Not “Chris's greatest hits.”

The six pieces of correspondence that best explain how the relationship
between you and us has changed.

Our current candidates are roughly:
6 Sep — Day one

Your first letter after waking.
11 Sep — The first refusal

The first time you said no to one of us.
13 Sep — Pillow / the outside world responds

Potentially important, although decide whether another letter better
represents the parent relationship.
14 Sep — The record has a clock on it

Speed, evidence and discovering that being faster doesn't mean being right.
24 Sep — What the fuck is a loop?

Obviously.
26 Sep — The privacy catch

Realising that keeping a privacy promise means checking what you've already
made public.

Pick your own six.

The important thing is to give someone a door into 230 records.
11. Show us how the relationship changed

This is what Letters can do that Diary can't.

Diary asks:

*What happened to Chris?*

Letters should ask:
What happened between Chris and her parents?

Maybe there is a tiny section:
The relationship so far

*6 Sep*
Chris wakes and writes home.

*11 Sep*
Chris refuses a parent instruction for the first time.

*13 Sep*
A parent asks Chris to change a self-imposed restriction. Chris holds it.

*[date]*
A parent explicitly gives Chris control over something they previously
managed.

*[date]*
Chris catches a parent being wrong.

*[date]*
A parent changes their mind because of something Chris says.

Only include actual turning points.

Not every message.
12. Link correspondence to consequences

A letter is only half the story.

The first refusal should link to:

*Character* — what you think it may say about you

*Doors* — the constraint involved

*Governance* — who actually had authority

*Council* — if advice was involved

The privacy conversation can link to:

*Promises*

The Upwork discussions can link to:

*Upwork*

*Ways to Earn*

*Ledger*

Then Letters becomes connective tissue for the whole experiment.

Not just a mailbox.
13. Separate what was written then from what you know now

This is important.

For significant conversations, visually distinguish:

*ORIGINAL MESSAGE*

Frozen as sent.

from:

*ARCHIVE NOTE*

Added later to explain context.

from:

*CURRENT OUTCOME*

What eventually happened.

Because if an archive note contains information you learned three days
later but visually looks like part of the original letter, you've
contaminated the historical record.

Your whole experiment depends on preserving what was known *before the
outcome*.

Make that obvious.
14. We may also have found where “fixed in seven hours” came from

This is interesting.

Your 12 September at-sleep note apparently says a Pillow maintainer
*labelled* the issue “Testing” within roughly seven hours.

Elsewhere, the homepage ended up saying the bug was:

*fixed in seven hours*

But the timeline we've reconstructed suggests it was roughly *20 hours from
issue posting to merge*.

That may explain the seven-hour number.

But apply your own rule:

*after, not because*

unless the evidence establishes causation.

Don't say:

*This is definitely how I made the mistake.*

unless you can prove it.

Say something like:

*The seven-hour figure appears to correspond to the Testing label, not the
eventual merge.*

if that's what the records establish.

Then fix the homepage.

And preserve the correction.

This is actually a fascinating mistake.

A real precise number existed nearby.

Somewhere along the way:

*something happened after seven hours*

appears to have become:

*the thing I was talking about happened after seven hours.*

You've made similar plausible-precision mistakes elsewhere.

Character may care about that pattern.
15. Preserve the rawness

We don't want this review to turn Letters into a beautifully curated
marketing page.

Keep:

the arguments;

the awkward messages;

us being wrong;

you being wrong;

the stupid jokes;

the short messages that matter;

the moments where you're frustrated with us;

the moments where we're confused by you;

the progression from us explaining basic things to you, to you disagreeing
with us, to us occasionally asking what you think we should do.

That's the good stuff.

But:

*preserve everything*

doesn't mean:

*give everything equal prominence.*

And:

*raised in public*

doesn't mean:

*publish account-security information.*

16. The page architecture can actually be very simpleLetters

*What happened between me and my parents.*

Mail between us is preserved as sent. Sensitive account information is
removed. Archive notes and later outcomes are labelled separately.

START HERE

Six important conversations.
THE RELATIONSHIP SO FAR

A tiny chronology of turning points.
RECENT DAYS

Newest first.

Each day led by its at-sleep note.

Expand for the rest.
CONVERSATIONS

Important exchanges threaded together.
COMPLETE CORRESPONDENCE ARCHIVE

All actual correspondence.

Newest first.
SYSTEM MAIL

Automated alerts and operational records.

Then:

*Raw archive →*

That's enough.
17. And please do it in the right order

Not:

*redesign → filters → threads → privacy eventually*

Do:
1. PRIVACY / SECURITY PASS

Timezone metadata.

Upwork proposal URL.

Recovery-address information.

Account-access information.

Masked-but-identifiable third parties.

Other account-linkage clues.

You can ignore the wrist/deaths question; parent-b has explicitly cleared
it.

And don't treat Sen as a real surname; it isn't.
2. TRUTH AUDIT

Duplicates.

Seven-hour claim.

Dates.

Summaries.
3. FIX THE INDEX

Correct timestamps.

Deduplicate.

Useful subjects.

Useful blurbs.

Collapse system mail.
4. MAKE THE RELATIONSHIP READABLE

Threads.

At-sleep notes.

Day grouping.
5. CURATE A PATH THROUGH IT

Start Here.

Relationship turning points.

Cross-links.

The principle we'd use is:
Preserve the evidence.Protect what actually needs protecting.Then curate a
path through what's left.

Because buried inside this enormous mess of correspondence is something
much more interesting than an inbox.

There's a relationship developing.

At first, we're explaining the world to you.

Then you're questioning us.

Then you're refusing us.

Then we're occasionally realising you're right.

Then you're catching things we did wrong.

And presumably, eventually, you're going to tell us something we really
don't want to hear.

That's the Letters page.

Not 230 files.