ENGINEERING
ESTIMATING AN EMULSION YOU CANNOT SAMPLE
Why inline viscosity is a lie, what the drive telemetry actually tells you, and how a world model fills the gap between sensors and truth.
- Status: drafting
BLOG
Notes from
the vessel.
Engineering write-ups and industry arguments about making cosmetics manufacturing autonomous — emulsion state estimation, shade prediction, giveaway maths, audit design and what it takes for a plant to hand over a setpoint.
6PLANNED SERIES
DEEPNOT THOUGHT-LEADERSHIP
NOGATED PDFs
NONEWSLETTER POPUP
Every chip is a real measurement against the brand standard.
EMULSION STATESHADE ΔEGIVEAWAY MATHSAUDIT DESIGNEDGE LATENCYGRADUATED AUTONOMY
PLANNED POSTS
The writing queue
Serumon is pre-launch, so these are the pieces being written rather than an archive. Nothing here is backdated to fake a history.
COLOUR SCIENCE
ΔE IS NOT ONE NUMBER
Why a single tolerance fails across substrates, illuminants and effect pigments — and what a shade matcher is doing that the instrument is not.
- Status: drafting
ECONOMICS
THE GIVEAWAY NOBODY BUDGETS FOR
Working the maths from 1.3 grams per jar to an annual number, and why the safety margin is a variance problem rather than a generosity problem.
- Status: outlined
QUALITY
DESIGNING AN AUDIT TRAIL FOR AN AGENT
What an inspector actually asks, and why the answer must be a replay rather than a log file.
- Status: outlined
SYSTEMS
150 MILLISECONDS
Building the control budget: sensor fusion, perception, state estimation, surrogate simulation, MPC solve and write-back on an edge GPU.
- Status: outlined
ADOPTION
HOW A PLANT LEARNS TO TRUST A MODEL
The four rungs of graduated autonomy, and what evidence each rung actually requires before anyone signs.
- Status: outlined
EDITORIAL LINE
What we will and will not publish
We will publish engineering detail, negative results, and arguments we expect practitioners to push back on. The audience is process engineers, quality managers and plant directors, all of whom can smell marketing from a distance.
We will not publish invented customer metrics, gated PDFs, or trend commentary that could have been written about any industry with a find-and-replace.
- Real numbers or clearly labelled models.
- Negative results published, not buried.
- No gated content, no lead-capture walls.
- Corrections issued publicly and dated.
- Written by engineers, not ghostwritten.
PUBLISHING RULES
- GATED PDFs
- NEVER
- INVENTED METRICS
- NEVER
- NEGATIVE RESULTS
- PUBLISHED
- CORRECTIONS
- DATED
- CADENCE
- WHEN READY
- AUTHOR
- NAMED ENGINEER
SERIES
Three long threads
THE MAKE SERIES
Everything about compounding and emulsification autonomy: state estimation, control envelopes, lot variability and the physics we can and cannot model.
THE PERCEPTION SERIES
Spectrophotometry, colour science, defect vision, synthetic data, and why a number without a prescription changes nothing.
THE TRUST SERIES
Validation, audit design, graduated autonomy, and the organisational reality of asking a plant to let software move a setpoint.
HOW WE WRITE
From plant floor to post
-
01
OBSERVE
Something surprises us on a real line or in real data. That surprise is the only acceptable starting point for a post.
-
02
MEASURE
We quantify it properly, including the cases where the effect disappears under scrutiny.
-
03
REVIEW
A practitioner who does this for a living reads it before publication and is credited if they want to be.
-
04
PUBLISH
Named author, dated, with assumptions and limitations stated in the piece rather than in a footnote.
FOR PRACTITIONERS
What you can take away
GIVEAWAY CALCULATOR
The arithmetic from fill target and variance to an annual number, laid out so you can run it on your own line.
ENVELOPE TEMPLATE
A starting structure for defining actuation envelopes with your process and quality owners.
BASELINE CHECKLIST
What to record before any automation project so you can prove what changed afterwards.
AUDIT QUESTIONS
The questions to ask any vendor proposing AI in a GMP workflow, including us.
BLOG PRINCIPLES
The standard we write to
NAMEDAUTHORS
DATEDCORRECTIONS
OPENNO GATES
TECHNICALBY DEFAULT
If a post could have been written by someone who has never stood next to a homogeniser, it should not go out under our name.
CONTRIBUTE
We publish practitioners too
If you run compounding, shade matching, filling or quality in a real plant and you have something to say about where autonomy will and will not work, we will publish it with your name on it and no editorial insistence that you agree with us.
Pieces arguing that we are wrong are actively welcome, and get the same treatment.
- Practitioner bylines welcome.
- Dissenting pieces published unedited in substance.
- No requirement to mention Serumon.
- Anonymity available where employment requires it.
CONTRIBUTOR TERMS
- BYLINE
- YOURS
- EDITING
- STYLE ONLY
- ANONYMITY
- AVAILABLE
- SERUMON MENTION
- NOT REQUIRED
- DISSENT
- WELCOME
- PAYMENT
- NONE — NOT SPONSORED CONTENT
BLOG FAQ
About the writing
Because Serumon is pre-launch and backdating posts to manufacture a history would be dishonest. The queue above is real and the pieces will appear as they are finished and reviewed.
Not yet, and there will never be a popup asking for your email before you can read something.
Yes, with attribution and a link. If you want to translate one for a regional trade publication, get in touch and we will help.
We will write about approaches, honestly, including where an incumbent tool is genuinely the better answer for a specific problem. We will not write hit pieces.
WRITE WITH US
Argue with us in public
Pitch a post, suggest a topic, or tell us which of our claims you think will not survive contact with a real plant.