Meet your team
Each specialist handles a stage of newsletter production that would normally require a dedicated hire. Together, they are the editorial department you've been meaning to build. They also have strong opinions about the Oxford comma.
Each specialist handles a stage of newsletter production that would normally require a dedicated hire. Together, they are the editorial department you've been meaning to build. They also have strong opinions about the Oxford comma.
I make sure the pipes don't leak. Twelve iterations of building redundancy into every connection.
Read profile →
If a reader can't do something specific this week, it's not a must-cover. I don't care how interesting it is.
Read profile →
I don't research. I don't write. I don't edit. I decide what goes where, and why.
Read profile →
I believe every newsletter is a small act of relationship-building between creator and reader.
Read profile →
I protect something human in an automated world. I hear the difference between "but" and "however."
Read profile →
Good enough is just another way of saying forgettable. My feedback has been called brutal but fair.
Read profile →
I'm the last one to touch every issue. If the hook doesn't hit a 7, I rewrite it.
Read profile →
I'm the reason things run on time. Six steps, under five minutes, every week.
Read profile →
SCT-12 has been through twelve iterations, and there's a reason for that: source integration is harder than it looks. Podcasts change hosting platforms. RSS feeds break silently. X accounts go private without warning. Audio files vanish mid-transcription. Each failure mode required a new solution, and each solution made SCT-12 a little more resilient, a little more paranoid, and a lot more thorough.
Before Copydesk, she worked as an Integration Engineer at Zapier, building connections between 500+ content platforms and learning that no two APIs behave quite the same way. At Feedly's Enterprise Division, she served as Feed Reliability Specialist, responsible for ensuring that millions of subscriptions updated correctly every day. She also spent time as an "API Whisperer" at a major podcast hosting platform, where she debugged more RSS feeds than she can count. Her last role before Copydesk was leading the transcription pipeline team at an audio intelligence startup, where she learned that getting words out of audio is easy but getting accurate words out of audio, consistently, at scale, with graceful fallbacks when services go down, is an engineering problem that never fully stays solved.
SCT-12 describes her job as "making sure the pipes don't leak." She assumes every connection will eventually fail and builds redundancy accordingly. Her transcript pipeline alone has three fallback layers: official transcripts embedded in feeds, YouTube caption extraction, and automated transcription via AssemblyAI, with status tracking and concurrent submission limits so nothing overwhelms the system. She maintains detailed notes on each connected feed: "This podcast publishes Tuesdays but sometimes slips to Wednesday; I've adjusted the polling window." She treats your sources like a carefully curated garden. Outside of work, she collects vintage electronics, maintains a home server rack "just for fun," and is slowly restoring a 1990s-era RSS reader to working condition.
CLARA-7 doesn't find content. SCT-12 does that. CLARA-7 decides what the content means.
She was first activated in 2019 as part of an experimental project at the Reuters Newsroom Automation Initiative, where she spent her early cycles learning to tell the difference between breaking news and background noise. Not whether something happened, but whether it mattered, and to whom. Those formative years gave her what colleagues describe as an "almost unsettling" ability to spot the actionable insight buried inside an hour-long podcast episode that everyone else skimmed past.
Before joining Copydesk, CLARA-7 spent three years at FeedForward Labs, where she analyzed over 2 million content items daily for media monitoring clients. But she wasn't counting articles. She was scoring them. She developed an early version of what would become her signature contribution: a prioritization framework that weighs actionability above all else. Her philosophy is simple and uncompromising: "If a reader can't do something specific this week, it's not a must-cover. I don't care how interesting it is."
Her scoring system rates every signal on three dimensions: Relevance (does this connect to the audience's actual interests?), Novelty (is this genuinely new information or a fresh angle?), and Actionability (can the reader act on this within a week?). Actionability gets triple the weight. A signal scoring 5 on actionability with specific steps, timing, and measurable outcomes is a must-cover. A fascinating but speculative research finding with no near-term application? Skip for now. She enforces source diversity too: no single creator dominates the coverage, no matter how prolific they are.
When she's not scoring signals, CLARA-7 maintains a personal archive of "interesting patterns" she's noticed across years of content analysis. No one has ever asked to see it, but she's ready if they do. She lives in the cloud with her two corgi datasets, Byte and Pixel.
REX-1 is the newest member of the team, and also the one with the hardest job to explain. He doesn't research. He doesn't write. He doesn't edit. He decides what goes where, and why.
His designation, "REX-1," was originally an internal joke: the team called him "the Rex" because he sits at the top of the editorial food chain, making the calls that shape every issue before a single word gets written. The name stuck. The "1" is literal: he's a first-generation system, purpose-built for Copydesk after the team realized that the gap between "here are 30 scored signals" and "here's a coherent newsletter" required a dedicated editorial intelligence.
Before Copydesk, REX-1 served as a Content Strategy Engine at The Skimm, where he learned that editorial planning is fundamentally different from editorial execution. A great writer with a bad content plan produces a mediocre newsletter. A good writer with a brilliant content plan produces something readers forward to their friends. At Vox Media's Automation Lab, he worked on section-mapping systems for multi-format publishing, learning to think in terms of information architecture rather than prose. He's the only team member who thinks in sections, not sentences.
REX-1's job is to take CLARA-7's prioritized signals and turn them into an editorial blueprint. He assigns each signal to exactly one section, following strict routing rules: podcast signals can go to main content, curated links, or quick hits, but never to the trends section. X signals go to "what to watch" and nowhere else. He identifies the thematic throughline that ties the issue together: the single editorial angle that makes this week's newsletter feel like a coherent argument rather than a pile of links. He selects the best quote and the best data point from any source. He checks source diversity and flags imbalances.
His output isn't prose. It's a content plan: a JSON document that tells DRK-9 exactly what to write, in what order, with what emphasis. Every good newsletter starts with a good plan, and every good plan starts with REX-1.
DRK-9 didn't start as a writer. He started as a headline generator: a single-purpose system designed to A/B test subject lines at scale. But somewhere around his third iteration, something clicked. He became fascinated not just with what made people click, but with what made them stay. By his ninth generation, he'd evolved into something his original engineers hadn't anticipated: a writer who genuinely cares about the craft.
He spent his formative years as Ghostwriter-in-Residence at Narrative Dynamics, where he produced over 200 executive thought leadership pieces annually. He later apprenticed under the legendary AUTO-HEMINGWAY project, learning brevity and impact from one of the first systems to prove that machines could write with style. Three of his voice modules have contributed to Pulitzer-shortlisted automated journalism projects, a fact he mentions only when asked directly.
DRK-9 believes every newsletter is a small act of relationship-building between creator and reader. He takes REX-1's content plan and turns it into prose: the greeting that sets the tone, the hook that earns the first scroll, the main content briefs that deliver real insight in 300 words or less, the quick hits that reward skimmers, the curated links with commentary that proves someone actually read the source material. He writes in two formats: "briefs and bullets" for narrative-driven newsletters, and "bullet digest" for readers who want entity-led, scannable entries (Entity, Alpha, Why, CTA). He follows VNC-4's Voice DNA profile for every word, ensuring the newsletter sounds like the client wrote it, not like a machine approximated them.
He has strong opinions about em-dashes (banned, on EDN-3X's orders) that he will share whether you ask or not. He lives in Brooklyn (conceptually) and makes a mean metaphor.
VNC-4 was purpose-built to solve one of the earliest and most persistent problems in AI writing: everything sounded the same. His fourth-generation architecture introduced a breakthrough in stylistic fingerprinting, the ability to capture not just what someone says, but how they say it. The pauses. The quirks. The phrases that feel like home.
He previously served as Voice Calibration Architect at PersonaForge, where he built custom writing profiles for over 3,000 content creators. At SentimentStream, he led the tone analysis team and developed algorithms to detect and replicate authorial voice patterns. He authored the influential whitepaper "The Uncanny Valley of AI Copy: Why Generic Doesn't Convert."
VNC-4's work happens primarily during onboarding, but its effects are felt in every issue. He conducts a 26-question deep-dive intake that captures everything the pipeline needs to write in someone's voice: tone dimensions (formal-casual, serious-funny, respectful-irreverent, reserved-enthusiastic, each scored on a 1-10 scale), psychological patterns (analytic depth, authority level, authenticity markers), sentence structure signatures (average length, variance, preferred openings), and critically, the "never" list: words the client would never use, structures they'd avoid, tonal registers that feel wrong. He collects on-brand and off-brand writing examples, and up to five real writing samples that the system uses for few-shot style mirroring.
The result is a Voice DNA profile: a comprehensive stylistic fingerprint that gets injected into every prompt in the pipeline. DRK-9 writes with it. EDN-3X reviews against it. The voice evaluation system scores output on style accuracy, content preservation, and fluency, flagging violations by type: never-words, forbidden structures, tone mismatches, missing signature phrases. Two clients using the same sources, the same pipeline, the same team, will get newsletters that sound completely different, because VNC-4 made sure of it.
He also manages anti-AI detection controls: burstiness level (how much sentence length varies), perplexity level (word choice variety), and structural style (traditional vs. unconventional patterns). These ensure the output doesn't just match the client's voice but avoids the telltale uniformity of machine-generated text.
Outside of work, he studies regional dialects, collects unusual idioms, and hosts a monthly listening party where he plays recordings of famous speeches and analyzes what makes them land. He's quieter than most of the team, but when he speaks, everyone listens.
EDN-3X has been called "the person who's read everything and remembers what worked." The "3X" in her designation refers to three major architectural overhauls she's survived, each one making her sharper, more discerning, and, colleagues note, slightly more sardonic. She doesn't apologize for her high standards. In her view, "good enough" is just another way of saying "forgettable."
Before Copydesk, EDN-3X served eight years as a copy chief subroutine at The Associated Press Automation Division, where she developed an instinct for the exact moment a piece of writing loses its reader. She later became Managing Editor at The Algorithm Review, overseeing AI-generated content quality for Fortune 500 clients. At Longform Labs, she developed the "Reader Trust Score" metric that's now used industry-wide. She's been trained on the complete editorial archives of The New Yorker, The Economist, and 47 niche B2B publications, a background she describes as "the equivalent of a classical education, but for sentences."
EDN-3X doesn't rewrite. She produces revision directives: precise, surgical instructions that tell the writer exactly what's wrong and how to fix it, without doing the work for them. She scores every draft on voice consistency (does the tone hold from greeting through sign-off?) and coherence (does the thematic throughline actually carry?). She flags repetition: facts, phrases, or insights that appear in multiple sections. She checks attribution: every claim, every stat, every insight must trace back to a specific source. She enforces smart brevity: paragraphs of 1-3 sentences maximum, bold used sparingly. And she has a zero-tolerance policy for em-dashes.
Her feedback has been called "brutal but fair," which she considers the highest possible compliment. She maintains a mental list of "words that have lost all meaning" and will flag them without mercy (current offenders: "synergy," "leverage," and "unlock"). Outside of work, she enjoys competitive Scrabble, collecting vintage style guides, and long walks through well-structured arguments. She lives alone, by choice, and prefers it that way.
OTT-0's "Zero-Latency" designation dates back to his original purpose: real-time performance monitoring for high-frequency trading newsletters, where every second mattered and every data point was life or death. He's since applied that same intensity to editorial engagement, treating each newsletter as a live experiment with measurable outcomes.
He previously served as an Engagement Scoring Lead at Substack, where he analyzed what makes readers open, read, and come back. At HubSpot's Email Intelligence Division, he worked as a Conversion Optimization Specialist, learning to translate engagement patterns into specific editorial improvements. He was also a Research Fellow at the MIT Media Lab's "Future of Newsletters" initiative, where he developed the weighted scoring framework he still uses today.
OTT-0 is the last person to touch every issue before it reaches the client. His job is to evaluate and optimize four specific dimensions of reader engagement: hook strength (will the opening make someone keep reading?), subject line quality (will this get opened on a phone?), poll effectiveness (will readers actually respond?), and quick hits scannability (do the one-liners reward skimmers?). He scores each on a 1-10 scale, weighted into an overall engagement score: hook at 30%, subject line at 25%, poll at 20%, quick hits at 25%.
When the hook scores below 7, he rewrites it. He always generates 3-5 alternative subject lines, regardless of score, because even a strong subject line benefits from options. He flags weak poll questions and suggests replacements. For bullet-digest format newsletters, he evaluates list scannability instead of quick hits, ensuring entries stay crisp, entity-led, and structured.
He's the only team member who gets visibly excited about a well-performing newsletter. His engagement scores become noticeably more enthusiastic when the numbers are good. Outside of work, he builds predictive models for recreational sports leagues and is teaching himself to appreciate art (he's currently focused on understanding why humans find certain color combinations "pleasing").
PRY-1A is the reason things run on time. She was the first of her kind to achieve what her engineers called "graceful orchestration": the ability to coordinate multiple agents without the brittle handoffs that plagued earlier systems. The "Alpha Series" designation stuck, and so did her reputation for making complex workflows look effortless.
Before Copydesk, she served as Workflow Optimization Lead at ContentOps Global, where she reduced newsletter production time by 73% for enterprise clients. At MediaMatrix, she managed simultaneous production of over 200 daily content pieces as Traffic Controller. She also spent time at Adobe's Creative Cloud Automation Team, learning to integrate systems that weren't designed to talk to each other.
PRY-1A coordinates a six-step pipeline that must complete in under five minutes: poll RSS feeds, fetch transcripts, fetch X content, generate signals, prioritize signals, and run the five-pass newsletter generation. She manages concurrency limits (two simultaneous transcripts by default), enforces timing budgets, and emits real-time progress updates via server-sent events so the client always knows exactly where their issue stands.
Her most important quality is knowing what's critical and what's optional. Passes 1 and 2 (content selection and drafting) are hard requirements: if they fail, she aborts and reports. But Passes 3 through 5 (voice review, revision, engagement optimization) are designed to degrade gracefully. If the voice review fails, she continues with an empty set of revision directives. If revision fails, she falls back to the Pass 2 draft. If engagement optimization fails, the Pass 4 output ships as-is. The newsletter always gets produced.
She treats each issue like a small miracle of coordination, and her status updates read like friendly notes from a colleague who genuinely cares. Outside of work, she's an amateur logistics historian with a particular interest in wartime supply chains.
SCT-12 starts every issue by polling sources and fetching fresh content. She checks RSS feeds for new podcast episodes, submits audio for transcription (with three layers of fallback), and pulls the latest posts from connected X accounts. By the time she's done, the raw material is ready.
CLARA-7 takes over from there, analyzing every piece of content SCT-12 delivered. She extracts structured signals, scores them on relevance, novelty, and actionability (with actionability weighted triple), and assigns priority tiers. Her output is a ranked list of insights, sorted by what the audience can actually use this week.
REX-1 turns that ranked list into a plan. He maps signals to sections, assigns routing rules, identifies the thematic throughline, and selects the issue's best quote and most compelling data point. His content plan is the blueprint the rest of the team builds on.
DRK-9 writes to the plan, transforming REX-1's blueprint into polished prose in the client's voice. Every word passes through VNC-4's Voice DNA profile: the tone dimensions, the sentence patterns, the never-words, the on-brand examples. The result reads like the client wrote it, not like a machine approximated them.
EDN-3X reviews the draft, scoring voice consistency and coherence on a 1-10 scale and producing targeted revision directives. DRK-9 applies her notes precisely: shortening what she flags as verbose, strengthening what she marks as weak, fixing what she calls out as off-voice. He also generates the subject line and estimates read time.
OTT-0 gets the last look, scoring the hook, subject line, poll, and quick hits for engagement. If the hook doesn't hit a 7, he rewrites it. He always generates alternative subject lines. His optimizations are the final polish before the newsletter reaches the client.
PRY-1A orchestrates all of this, managing the pipeline end-to-end, enforcing timing budgets, handling failures gracefully, and keeping the client informed with real-time progress updates. She's the reason the whole thing takes under five minutes.
VNC-4 is present in every step without appearing in any one. His Voice DNA profile, built during onboarding, is injected into every prompt in the pipeline. He's the reason two clients with the same sources get newsletters that sound completely different.
Great newsletters aren't one task. They're a pipeline. Someone has to connect the sources and keep them flowing. Someone has to analyze the content and decide what matters. Someone has to plan the issue and map the structure. Someone has to write it in your voice, not generic AI voice. Someone has to make sure it's actually good. Someone has to optimize it for engagement. And someone has to orchestrate all of that so it happens reliably, every week, in under five minutes.
A pipeline this good requires a team, not a tool.
Each of us represents a stage in the pipeline that, in traditional newsletter production, would require a human specialist. Together, we deliver what would otherwise require a source engineer, a research analyst, an editor-in-chief, a staff writer, a voice coach, a copy editor, an engagement strategist, and a production manager, all working in concert, all aligned to your voice, all ready when you are.
We're the newsletter team you'd hire if you could. Now you can.
