IM8 · Internal Memo Vol 01 · LinkedIn Launch · 27 May 2026

For Marco · Review & Reply

Less AI.
More operator.

Two posts, five days apart. You and I each get our own moment. We stop bragging about Claude and start telling the truth about what running our ops was like before, and what changed after.

Posts 2 distinct voices
Cadence 5-day gap, Raymond first
Tone Operator first, AI in passing
Hashtags #SupplyChain · #Ops · #DTC
§ 01 · Post plan

The two posts

Same product, two different vantage points. Raymond writes about the decision to stop waiting. Marco writes about the daily reality of using the thing. Both posts open with a scene, not a stat.

Post 01 · Day 0 Raymond

Raymond

Why we stopped waiting for engineering.

Above the fold Six months ago I was on a Sunday call trying to figure out why our reorder quantity for one SKU was off by 40%. The answer was in a spreadsheet someone had renamed.

I'd had enough.

Marco and I started building. Not a side project — the actual system we needed to run IM8. Inventory, BOMs, carrier invoices, forecasts, gifting, the WhatsApp ops bot we call Jarvis. Twelve modules now. We use Claude alongside us most days, which is what made it possible for two non-engineers to ship something this big without it collapsing under its own weight.

It's not pretty in places. Some of the early code makes me wince. But last Tuesday I checked our 14-day demand forecast on my phone between meetings and it was just… right. No spreadsheet archaeology. No Slack thread asking Marco which file was current.

That's the part I didn't expect — how much mental space you get back when the system you rely on actually tells you the truth.

Marco's going to share his side of this next week. His view is more interesting than mine.

Post 02 · Day +5 Marco

Marco

What changed when the system stopped lying.

Above the fold The thing nobody tells you about running ops for a fast-moving brand is how much of your week is spent answering “wait, which number is right?”

I used to keep five 3PL portals open at once. Atlanta, Reno, Hong Kong, Amsterdam, plus the gifting one. Each with its own login, its own export quirks, its own definition of “shipped.” Reconciling carrier invoices took me two full days a month. I built workarounds on top of workarounds.

Raymond mentioned the HUB last week. From the user side — meaning me, the person who actually lives in it at 8am every morning — the change isn't the dashboards. It's that I stopped being the human cache.

People used to ping me to ask what the HK on-hand was. Now they just look. Last month I closed our carrier reconciliation in under an hour. First time that's ever happened.

I still find bugs. I file them in a channel and they usually get fixed the same day, sometimes by Raymond, sometimes by Claude with him watching. Weird workflow. Works better than anything I've used at bigger companies.

§ 02 · Timing

Cadence

A five-day gap is the sweet spot. Wide enough that the second post doesn't fight the first for reach. Tight enough that Raymond's audience still remembers when Marco shows up.

Step 01

Day 0 · Tue/Wed AM HKT

Raymond posts. 9–10am HKT captures APAC engagement plus the US morning catch-up.

Step 02

Day 0–3 · Comment activity

Marco drops into Raymond's comments to add color. Algorithm rewards real conversation.

Step 03

Day +5 · Same window

Marco posts. Opens with “Raymond mentioned the HUB last week” — natural continuity, no quote-link or first-line tag.

Step 04

Day +5–8 · Raymond reciprocates

Raymond comments on Marco's post. We let it breathe. No Part 3. We see what the audience does with it.

§ 03 · Editorial decisions

What we cut, what we kept

The original draft leaned on every LinkedIn-flex tic in the book. Here's what we pulled out, and the one thing we deliberately kept against the instinct to bury it.

Cut

  • “$200K enterprise platform”Reads as a procurement humblebrag. Invites “how'd you calculate that” cynicism.
  • “Zero engineering hires”Alienates engineers in our network. Not fully honest — Raymond was the engineer, Claude was the pair.
  • The 12-modules-in-5-categories listFeature inventory reads like a product page. Operators don't reshare taxonomies.
  • “Could your team ship internal tools with Claude?”Generic CTA. Performative. Cut entirely.
  • The #AI #Claude hashtag wallPattern-matches to AI-influencer slop. Operators scroll past it.

Kept

  • Claude as enabler, not headlineMentioned once in each post, mid-paragraph, in passing. Quiet but present.
  • The AI thesis itselfIn 2026 every ops lead claims AI; almost none have actually shipped a running platform with it. That's the moat.
  • Jarvis by nameNaming the WhatsApp bot gives the post a real character without becoming a feature pitch.
  • Specific numbers40% off on one SKU. Two days a month. Under an hour. Specifics beat round numbers.
  • Self-deprecating notes“Some of the early code makes me wince.” The brag is what kills credibility — the wince builds it.
§ 04 · Over to you

Open questions for Marco

Things I need your read on before we publish. Honest takes preferred — if any of these don't ring true, the whole thing falls flat.

  1. Is the “40% off reorder quantity” anecdote close enough to a real moment? Or do we have a sharper one from our actual history?

  2. Is “two full days a month on carrier reconciliation” the right number for you, or do you want to adjust it up or down?

  3. Do we name Jarvis in Post 1, save him for Post 2, or hold him for a future post entirely?

  4. Anyone on the team we should tag — operators, 3PL contacts, vendors — or do we keep it ungarnished?

  5. Should either post carry an image? My instinct: one screenshot of the runway/forecast view would be powerful. A five-tile mood board would undercut the “operator typed this on his phone” voice.

§ 05 · The reviewers disagreed

What the agents argued about

Before I wrote this, I ran the original draft past four reviewers with different lenses. The strongest disagreement was over the AI tone itself. Worth sharing.

The contrarian's pushback

“Raymond is wrong to strip the AI angle. Less AI tone is not less AI story. In 2026 every ops leader claims AI; almost none have shipped a $-running internal platform with it. That's the moat. Remove the slop tone — the em-dashes, the ‘Here's what we learned:’ lead-ins, the emoji bullets — but keep the AI thesis loud.”

They also pushed against splitting at all: “LinkedIn's algorithm penalizes the second post in a sequence. Two posts to manufacture two share-moments is the inauthenticity you're otherwise trying to avoid.”

How we resolved it

On tone: the operator-voice drafts thread the needle. Claude is in the story — named in both posts — just not the story. That's “less AI tone” without burying the differentiator.

On splitting: the contrarian assumed coordinated same-day posting. A five-day gap, with Marco's distinct user-not-builder angle, reads as two operators independently reflecting — not a coordinated campaign.

§ 06 · Guardrails

What we are not doing

A short list of temptations, named so we don't drift back into them as we revise.