Home / Strategy / Launch and Go-To-Market
Updated Jun 03, 2026 · Affirmology_SprintPlan_v1.md
Prepared for Jeff Parker
Pairs with Affirmology_BackendArchitecture_v1.md (the big-picture map). This document is the action layer over the next ninety days, broken into five sprints. Each sprint is short enough that you can see the end from the start.
Date June 3, 2026
Calibration note The previous architecture doc proposed thirty-day phases. That was too cautious. AI-era execution compresses what used to be quarter-long efforts into one or two weeks of focused work. The schedule below assumes pedal-to-metal pacing from June 8 onward.
Three categories of work need to be on the map that I underweighted in v1.
Branding and messaging, led by Sol. Colors, logo, copy voice, naming conventions, the visual language of the brand. None of this is engineering work, but all of it lands before or during the public-facing pieces. Sol's contribution shapes the landing page and the video. We should not build either of those until Sol has at least drafted the brand direction.
The instructional video. Three to five minutes of motion-graphic storytelling that replaces the deck in most one-on-one conversations. Concept-driven: shows the Gene Keys map and the meaning of sphere plus line, walks through the Human Design bodygraph, shows transits and eclipses moving across an astrology chart, calls out the gap in the existing subconscious-audio market, ends with a brief Jeff-and-Sol "meet the team" segment. This is not optional. A video conveys this product faster and more accurately than any slide deck because the product is itself a visual translation of an invisible blueprint.
The landing page and form together, not a form alone. A working form on a blank page is a tool. A working form on a page that explains the product, shows the science, sets the price expectation, and invites the user in is a product. The first version of this should ship together, even if minimal.
You have five days. The frame for this sprint: every hour spent should produce something a candidate investor can touch, see, or hear by the time you sit across from them.
Demo polishing. Re-render Soledad, Colin, and Alex at the corrected auto-tuned speed so they all land in your 6:00 to 6:10 voice-end window. About $1.50 in ElevenLabs spend, one bash session. I can do this in your absence today.
One more demo for Josh prep. When you have his current chart data confirmed and a candidate script angle (post the new poetic prompts), I'll render a fresh Josh demo with the updated agent and PDF generator. The new pipeline shipping Josh a refreshed Sacred Audio Report PDF is a different conversation than the v1 you sent him last time.
Camp Brotherhood field demo kit. A folder on your laptop with one MP3 plus PDF per archetype you might encounter at Camp: a Generator, a Manifestor, a Reflector (Steven covers this rare one), a Projector if we can render one, plus one Manifesting Generator (Laura). That gives you a hand-pickable demo for any conversation that comes up at Camp. About $3 in ElevenLabs spend, one afternoon of work.
Architecture finalization on the backend. While you are at Camp, I can: 1. Wire the JSON sidecar to write to a Postgres-ready schema (so the migration to Supabase next sprint is a copy operation, not a refactor). 2. Add the audio QC agent that scans every render for mid-track volume dropouts (the issue Sol and Alex hit at speed 1.16-plus). It runs automatically on every render going forward, flags any render with greater than 3 dB sustained variance in a sliding window, surfaces problems before they ship. 3. Build the multi-provider routing layer (ElevenLabs / Fish Audio dual-provider) so the moment Fish is voice-tested, you can flip the default with one config line.
One landing-page draft. I can write the copy and produce a static HTML mockup of the landing page using your existing brand assets (which I will pull from the existing AFFIRMOLOGY folder). Sol can override colors, logo, hero image, and final messaging once she is in. This is a placeholder draft that gives the next sprint something to refine, not the final.
By the end of Sprint A. You walk into Camp Brotherhood with a working demo for any chart, four to six rendered demos ready to play on demand, a refreshed PDF for Josh's update conversation, a landing page draft ready for Sol's input, and a backend that is one sprint away from being database-backed and quality-controlled.
Two weeks of focused work after you return. The frame: ship the public-facing pieces and let the team plug in.
Sol kicks off branding. Day one of this sprint, Sol gets the brief: brand voice, color palette, logo, typography, hero imagery direction, messaging hierarchy for the landing page. Sol does not need to finish all of it in two weeks. She needs to give us enough that we can ship a public landing by the end of the sprint without needing to revisit colors later. Parallel track to engineering.
Mac mini becomes the production machine. Order it if you have not. Set up Tailscale for remote access, install Python, clone the repo, run the test suite. By end of sprint B the Mac mini is running the agent endpoint for both your laptop and the web form. This buys you the always-on quality you need before the form goes public.
Supabase Postgres goes live. Migrate the JSON-sidecar schema into real database tables. Add a user table with email and opt-in. Add a corpus_entry table with zero rows but the schema in place. Add a render table that becomes the source of truth for every shipped audio. Future agents read from and write to these tables instead of the filesystem.
Web form plus landing page ship together. Vercel-deployed Next.js. The form takes name, birth date, birth time, birth location, email. The backend calls the Mac mini's secure endpoint. The Mac mini renders and emails the deliverable. The landing page above the form explains the product using Sol's messaging, shows two embedded sample audios, links to the Sacred Audio Report PDFs, and uses Sol's branding. End of Sprint B: a stranger can find affirmology.com, listen to a sample, submit their birth data, and receive their audio and PDF within fifteen minutes by email.
One audio use-case beyond meditation. Pick one (breathwork seems closest to what your Camp Brotherhood audience knows). Sol and Colin contribute the language they would use in a live event. I add a script-generator variant that produces that format from any chart. We render one demo per archetype in the new format. Now you have meditation and breathwork audios as evidence that the same engine can produce multiple formats.
By the end of Sprint B. Live web form, live landing page with Sol's branding, Mac mini production box humming, Postgres backing the data, two distinct audio formats demonstrated, and a corpus database schema ready for the next sprint to start populating it.
This sprint is dual-track: video production runs in parallel with investor-facing polish.
Video production. The instructional video gets scripted, storyboarded, and produced. Three production options to evaluate based on your timeline and Sol's bandwidth:
My recommendation: Option 1 for the chart sequences (because the chart data is already in the system and animations can be deterministic), Option 3 for the abstract emotional sequences, edited together with the live Jeff-and-Sol segment from Ultimate Wellness Conference footage you mentioned. End deliverable: one 3-to-5-minute master video and three or four 30-to-60-second cutdowns for social and ads.
Investor demo polish. A specific page in the dashboard that shows three things on screen at once: the verified chart (visual), the script with chart-element annotations, and the audio playback. This is your "show, don't tell" moment in any investor conversation. Even better, let the investor type in their own birth data on the spot and watch the system produce their personalized audio in real time. The "feel and touch" factor you mentioned. Three days of work to build a polished version of the demo dashboard you already have running.
First knowledge-corpus agent. The corpus builder for classical Western astrology starts nightly runs on the Mac mini. By the end of this sprint, you have a few hundred verified records in corpus_entry. Not enough to replace Claude's training-time knowledge yet, but enough to demonstrate the architecture works and the corpus is real. This is also the artifact you can show investors as evidence of a defensible long-term moat.
By the end of Sprint C. Video in production or finished. Live web form with branded landing. Investor demo dashboard. Corpus agent running. Multiple audio formats live. You can have a serious conversation with any of your candidate funders.
Stretch sprint. Three weeks. The frame: everything you need in place for a soft public launch by end of July, hard Miami launch positioned for September or October.
Practitioner Tier infrastructure. The breathwork-coach interface from the architecture doc. Coach signs up, configures a JourneyTemplate, gets a unique form URL branded with their event. Attendees submit through the coach's URL, get a personalized audio rendered with the coach's selections, enter Affirmology's email funnel with opt-in. End of sprint: one paying practitioner can use it for a live event. Pricing is set, contracts drafted (legal review still pending).
Audio QC agent matures. Beyond detecting volume dropouts, it now runs LUFS normalization, identifies mispronunciations using a transcription pass back through the audio (Fish Audio's STT is cheap), and flags scripts where the verifier or QC flags anomalies. Every audio that ships is silently double-checked. This is your reputation insurance.
Self-improvement loop, first lap. The agent that reads telemetry, looks across the now-meaningful render history, and proposes one or two prompt improvements per week, runs its first cycle. You read the proposals, approve or reject. This is the start of the perpetually-improving system. Conservatively scoped: one improvement per week, all human-reviewed before shipping.
A/B testing framework. The render pipeline can now A/B different script lengths, voices, music beds, and intro lengths per user. You learn which combinations retain best. Start with three experiments, not thirty.
Compliance and consent flows. Privacy policy live. Terms of service live. Voice-clone consent flow live (even if no one is using it yet). One-click user-data deletion live. None of this is exciting work, all of it is required before the Miami push.
By the end of Sprint D. Soft public launch ready. Paying-practitioner option live. Quality-control automation in place. Compliance ducks in a row.
The longest single sprint here because most of it is iteration, testing, and content production. Frame: take the system from "working" to "polished and unmissable" by Labor Day.
Multilingual launch in Spanish first, Portuguese second. Script generator gets a Spanish system prompt variant. ElevenLabs Multilingual v2 or Fish Audio handles Spanish voice synthesis. The PDF agent gets a Spanish-language template. By end of sprint, the same web form can deliver a fully Spanish audio plus PDF to a Spanish-speaking submitter.
Live event variant. Kiosk-friendly form flow for in-person events. QR-code delivery to attendee's phone. Pre-event batch rendering pattern from the original AutomatedPipelineSpec doc. Tested at one small live event before Miami.
Content corpus matures across multiple traditions. By end of August, the corpus has thousands of verified records across Western astrology, Vedic, Human Design, Gene Keys, numerology, and EFT/somatic protocols. Generation agents start composing primarily from corpus retrieval, falling back to Claude inference only for synthesis steps.
Investor materials hardened. Pitch deck, financial model, demo video cutdowns, one-pager, due-diligence data room. Sol drives copy and visual. You drive narrative. I drive numbers and technical claims.
By the end of Sprint E. You can walk into Miami with a product that has been used by real strangers, supported a paying practitioner, run a live event, shipped in two languages, and accumulated a defensible knowledge corpus. The investor conversation is no longer "we're building this," it is "we're scaling this."
Three tracks that do not fit cleanly in any sprint because they are ongoing.
Soledad and Colin contributions. Both should be writing corpus entries and audio-use-case briefs from June onward. Their contributions go into the corpus, into the prompt templates for the various script-generator specialists, and into the methodology PDF reference language. Establish the contribution flow in Sprint B so they are productive by Sprint C.
Camp Brotherhood teaching. You are teaching after June 8 and that work happens in parallel. The teaching itself is a knowledge source. Record sessions. Transcripts become candidate corpus entries. The teaching also reveals which audio use-cases your real audience reacts to, which informs the script-specialist priorities.
Funder conversations. Josh next week. Other candidates after Camp. Each conversation is also a product test. What questions do they ask, what objections come up, what makes them light up. Feed those observations back into the landing page and the deck.
Concretely, between now and your departure for Camp Brotherhood, with no further input from you, I can:
That is one or two productive days of work. Tell me which of these you want first and I will start immediately.
The previous architecture doc's "thirty-day" framing was wrong for where you are. The product is real. The competitive window is now. The architecture is sound. What separates this from a normal software project is that most of the heavy thinking is done. What remains is iteration speed and content production, both of which AI compounds. Five days is enough time to ship a new agent. Two weeks is enough time to ship a meaningful database. A month is enough time to ship a public launch. Three months is enough time to be the thing nobody can catch up to.
That is the pace this plan assumes. Push back if I am still too cautious anywhere.
End of document. Annotate, push back, or pick the next sprint and we start.