Message Mommy
Protocol translator, not tone police.
Most communication failures are not intelligence failures. They are protocol mismatches: what one person means, what another person hears, and what either side is socially equipped to process are often three different things. The internet keeps treating these collisions as moral failures or personality flaws when most of them are actually translation errors between incompatible processing styles, tone expectations, and social threat detectors. Message Mommy exists because that mismatch is tractable — not solvable, but legible, if you build the right lens.
The Trigger
"im deeply troubled by the state of the internet, it feels like everyone is so polarized, and proposing anything that will bring light to issues also means that they will likely have to check their own attitudes, which most people don't have the courage to do"
This didn't start as a product pitch. It started as frustration — the kind that accumulates from too many polarized exchanges, too many status performances disguised as wisdom, too many conversations where trolls on one side and cynical gatekeepers on the other burn through whatever substance might have existed. The deeper pain was the sense that people resist any tool that forces self-confrontation. They'll adopt anything that helps them decode others, but recoil from anything that shows them how they land.
"these people are everywhere, and it's part of the problem. just as the fucking internet trolls are. its like culture and counter culture or something"
The product clicked when that frustration collapsed into one concrete UX idea.
The Shape
"as a basica ass idea for this. I want a chrome plugin where you can click on a message, and it translates it based on how you process inforamtion, and you put your reploy, and it says 'the person may see this as xyz'"
That's the entire product in one sentence. Click a message, get it translated into the way you process information. Write your reply, get a prediction of how the other person will read it. The tool is bidirectional because the problem is cyclical — interpretation changes reply, reply changes reception, reception changes escalation. Most tools in this space handle one direction. They either help you decode other people or help you polish your own words. Message Mommy insists on both because the dysfunction lives in the gap between the two.
"my sentiment is that it's part of the cure. it's less a buffer, and more a way to track sentiment, prioritize, and also provide pushback to abusers."
Three Pipelines
The product is three operations, each with a distinct purpose and a carefully constrained prompt contract:
Analyze is inbound protocol translation. Take a message you received, surface the likely intent, emotional subtext, communication style, and implicit expectations. Not summarization — interpretation with uncertainty attached. Fifteen intent labels, eleven reception labels, confidence thresholds, and explicit anti-pathologizing rules.
Audit is outbound reception prediction. Take a message you're about to send, predict how the recipient will likely read it across dimensions like tone, status signaling, ambiguity, and power dynamics. Not grammar checking — social pattern recognition.
Rewrite is signal adjustment with meaning preserved. Six modes — from softening to clarifying to professionalizing — but the constraint that matters is this: the author's meaning and viewpoint are non-negotiable. The tool adjusts how the signal lands without replacing the person behind it.
That level of specificity in the prompts exists because the default model behavior in this space is sloppy. Without explicit constraints, models moralize, over-diagnose, flatten ambiguity, mistake style for character, and turn every rewrite into personality laundering. The prompt contracts — roughly 350 lines of interpretive framework — are there to stop the model from becoming a smug therapist or a generic writing assistant.
The Ethical Position
The core design principle is in the name of the first pipeline: analyze, not diagnose. The language choices throughout the system are deliberate. "Likely" instead of "is." "Signal pattern" instead of "diagnosis." "Preserve meaning" instead of "make it nicer." "Reception" instead of "correctness."
"Never pathologize. 'Passive-aggression' is a signal pattern, not a character diagnosis."
"You are adjusting signal, not replacing the author."
The product is strongest when it feels like a mirror with uncertainty, not a judge with authority. It does not treat social meaning as mystical — it treats it as pattern recognition with confidence levels. That's a much sharper and less crowded claim than "AI helps you write better."
Chat Daddy's Sibling
Message Mommy lives in the same lore universe as Chat Daddy, but they solve different sides of communication. Chat Daddy is the transcript companion — a minimal viewer for AI conversations, born from the frustration of losing chats behind unstable GUIs. It says: here are all your conversations, unhidden. Message Mommy is the protocol companion. It says: here is what this conversation is likely doing.
The names are part of the point. They are internet-native, memorable, a little provocative, and deliberately unlike sterile SaaS naming. The more boring enterprise wrapper can always be spun up later under a different brand with the same engine underneath.
"i want to go the sharp internet native tool. i think we can create a sister company to be the more corpo boring shit. i think internet native will have a wider reach and impact at the start"
What Shipped
| metric | value |
|---|---|
| Packages | 3 |
| Intent labels | 15 |
| Reception labels | 11 |
| Rewrite modes | 6 |
| Integration adapters | 4 |
| Core runtime deps | 0 |
| Prompt contract | ~350 lines |
@message-mommy/core— the analysis engine. Provider-agnostic (Anthropic, OpenAI, or bring your own). Zero runtime dependencies beyond fetch.@message-mommy/integrations— pre-built handlers for GitHub, Linear, Discord, and generic webhooks@message-mommy/extension— Chrome extension (Manifest V3) with a dark terminal-style sidepanel UI- Processing-style config — directness tolerance, ambiguity tolerance, preferred tone, explanation depth. The tool adapts to how you process, not to a universal standard.
TypeScript strict, npm workspaces monorepo, Vite for the extension, Preact for the sidepanel, tsup for the SDK builds. The initial repo landed the same day the idea crystallized — concept to shipped monorepo in a single session.
The Partial Solution Thesis
"my goal is to say 'I am providing solutions to part of the problem - somthing we ALL need to do, then we will have a whole solution and not just from one person'"
The project does not claim to fix online discourse. It claims that a lot of online dysfunction persists because behavior stays illegible until it explodes. Message Mommy is an attempt to make that behavior legible earlier without pretending to possess certainty. Interpretation over judgment. Preservation over polishing. Partial tractable interventions over fake grand solutions.
The browser is where the communication loop happens in real time, so the Chrome extension is the right surface for the first version. Click a message. Understand what it's likely doing. Draft your reply. See how it's likely to land. That's the whole cycle, running in the sidebar while the conversation happens live.
Technically yours,
Ana Iliovic
Message Mommy is on GitHub. Core SDK, Chrome extension, and integration adapters. Protocol translator, not tone police.