Home / Ops / Housekeeping and Process

Affirmology, New Information and Decisions Log

Updated Jun 18, 2026 · Affirmology_NewInfo_Log_v1.md

Affirmology, New Information and Decisions Log

A running record of new information and decisions as they come. Newest at the top. The investor master-doc PDF Colin has is a frozen snapshot (the 15-point manifesto version); everything captured here came after it and should be folded into the master doc, the Executive Summary, and a future PDF refresh when wanted.

2026-06-17

• CORPUS QUALITY-GATE METHODOLOGY FIX (decision, must hold): the Nemotron-vs-Haiku judge initially scored each structured element against only a ~1,500-char source snippet, which often did not contain the element being judged, unfairly penalizing BOTH backends and making absolute scores meaningless. Fix: the judge must see the SAME source context the structurer saw (ideally the full doc / relevant section), not a bigger arbitrary snippet. Consequence: the PASS/HOLD verdict hinges on the RELATIVE A/B (Nemotron vs Haiku on identical fair context), not absolute numbers. Bar: Nemotron wins the majority of comparisons AND mean quality within 0.5 of Haiku. ALSO required before trusting a PASS: read 5-10 real side-by-side Nemotron-vs-Haiku outputs by eye (the LLM judge is itself fallible; our rule is to read the real output). On HOLD: run the one-line cleanup to purge rejected extracted_by='nemotron' records so the corpus is not polluted. Overnight backlog stays held until the verdict is seen (corpus quality touches audio content, so decided with Jeff). Pilot (n=3) leaned Haiku-better but that is noise; the 75-doc run is the real read.

• INFRA ACTIVATION state (cloud move in progress): Cloudflare account ID e9740eb6decf2d83d73a036abf9a89ad; R2 bucket affirmology-media created. Studio repo now PRIVATE on GitHub at jeffparkerlove22/affirmology-studio (render.yaml + api/Dockerfile included, secrets gitignored, gh authed). R2 access uses an ACCOUNT API token (Object Read & Write, scoped to the bucket), not a user token. Render env var names (from config.py): STUDIO_STORAGE=r2, STUDIO_REQUIRE_AUTH=true, STUDIO_AUTH_TOKEN (strong secret, worker must match on repoint), R2_ACCESS_KEY_ID, R2_SECRET_ACCESS_KEY, R2_BUCKET=affirmology-media, R2_ACCOUNT_ID (no separate R2_ENDPOINT, code derives it), R2_PUBLIC_BASE_URL (effectively required or download links break, set to r2.dev public URL or a media.affirmology.ai custom domain). Caveat: R2Storage needs boto3 in the deploy deps. After Render is live, repoint studio.affirmology.ai off the laptop Tailscale funnel to Render via the Cloudflare connector, and sync the corpus DB up so cloud renders stay chart-grounded.

• CORPUS GROWTH FINDING (course-correct): the "707-doc backlog" turned out to be non-structurable junk (failed extractions, off-topic books, raw HTML) and produced 0 records for ~$2.88 before the cost cap stopped it. Corpus is unchanged and intact at 12,905 records. DECISION: do not reprocess the residue; grow the corpus by crawling NEW quality sources instead. This makes the coverage audit more important: run it to find thin/missing elements, then point targeted crawling at those gaps. Haiku remains the structurer.

• NANO BANANA / GEMINI confirmed working: GEMINI_API_KEY in affirmology-agent/.env authenticates (gemini-2.5-flash), and gemini-2.5-flash-image generated a real on-brand 1024x1024 PNG for ~$0.04. So Claude Code now has a cheap on-brand image generator (character art, quick visuals) without spending Higgsfield credits. Cowork has no native Gemini tool but has Higgsfield connected for in-chat gen.
• RENDER cloud move status (end of 2026-06-17): cloud Studio live at https://affirmology-studio-api.onrender.com; /api/health shows require_auth/music_bed/corpus_db all true; /api/compose + /api/audio return 401 without a token (door locked). Step-4 machinery pushed (594283c): corpus boot-pull re-pulls when R2 differs from local, and /api/corpus-stats exposes the live record count for verification after a refresh. REMAINING: (1) test cloud render is blocked only on the Render STUDIO_AUTH_TOKEN (laptop .studio_token does not match it; it is a dashboard secret) - paste it to Claude Code to run the render; (2) repoint studio.affirmology.ai from the laptop Tailscale funnel to Render and set the worker's STUDIO_AUTH_TOKEN to match (done with Jeff, awake, so he can verify); (3) end-to-end verify + retire the laptop funnel. One canonical STUDIO_AUTH_TOKEN should be shared across Render + worker + laptop.
• CLOUD MOVE PROVEN (late 2026-06-17): the full pipeline ran end to end in the cloud, compose (council, corpus-grounded, ~884 words) + ElevenLabs audio + WeasyPrint PDF + chart-integrity guard, with BOTH the MP3 and the PDF confirmed landed in R2 via head_object (status: done, both URLs populated, guard passed). PDF root cause was the weasyprint PACKAGE missing from the deploy deps (not a native lib, as first guessed); fixed by declaring weasyprint + pypdf (the integrity guard's PDF reader) plus native libs (libpangoft2-1.0-0, libfontconfig1, libglib2.0-0, libharfbuzz0b). A full dependency audit (AST scan of engine imports vs declared deps) confirmed all render-path deps are now declared and corpus-build-only deps (bs4, trafilatura, playwright, google-genai/vertexai, etc.) are correctly excluded to keep the image lean. Auth locked (require_auth true, 401 without token). ONLY REMAINING ITEMS (Cowork + Jeff, next session): (1) enable R2 PUBLIC SERVING on media.affirmology.ai, public fetch currently returns 403 though the objects are confirmed in the bucket, so download URLs do not resolve yet; (2) REPOINT studio.affirmology.ai from the laptop Tailscale funnel to the Render service, setting the worker's STUDIO_AUTH_TOKEN to match Render's; (3) one confirming render through the live domain, then retire the laptop funnel. Also queued (not blocking): a dedicated techniques agent (council currently only has a technique-fit gate + Hypnos gate, no generative techniques specialist), the Atlas PWA build on top of the now-stable cloud, and corpus growth via NEW sources (the 707-doc backlog was junk).

2026-06-16

• CONNECTIVE IDEA (option, not locked) for the investor film: open on the comedic meet-cute, and the lost guy in it IS the customer-transformation protagonist (renamed). Meet-cute first (funny/relatable) -> his transformation story -> the brand thesis -> the rest, or weave his beats through the thesis. The lost stranger becomes the transformed hero; ties three standalone assets into one through-line. Test with footage.
• NARRATIVE ASSETS are a LIBRARY of standalone pieces, not just investor-film ingredients. Each matters on its own and gets reused/varied across placements: (a) the brand-story film (a film for clients), (b) the customer-transformation story (rename, not "Bobby"), (c) the comedic two-strangers meet-cute (also a candidate comedic interlude inside the investor film). Build the assets as assets. The investor film is its own production that borrows from them and also carries its own additional arcs (why now/market, why us/team, the plan + fall lineup, the ask, myth/mission/empowerment).

• DIRECTIVE on the investor film: do NOT lock a running order. Everything is a concept pool / menu of beats to assemble with the footage and the other narratives the video needs. Rough arc to aim for: start strong with emotion and feeling, then logical and substance, then back into our plan and the fall lineup, then myth, mission, empowerment. Master doc reframed as a menu, not a sequence.

• TWO-NARRATIVE concept (from Colin's "Bobby" customer story + our thesis brand story): they are complementary halves, the lived transformation/movement/FOMO (customer story) and the why/how (thesis). Captured Colin's full arc in Affirmology_BobbyStory_CustomerNarrative_v1.md. Decisions: do NOT use the exact Bobby arc; rename the character (Jeff dislikes "Bobby"; consider a female protagonist or a couple given the dual audience). WEAVE both in the main investor film (lead with the customer story to make them feel/believe, then the thesis to justify, mirroring our own mechanism). VERSION the two threads per placement: landing page, sales page, funnel, post-audio, email nurture, social, investor, each pulls the fitting slice. Reinforce that MANY PEOPLE already speak the cosmic-blueprint language, so the story lands on recognition, not education. Narrative-strategy note added to the master doc.

• New cultural-proof scene (corrected, in master doc): a flirty MEET-CUTE, not a cold reading. Two people who just met; she is curious and fluent, asks his rising and sun (Aries sun, no idea on rising), playfully interprets it (he is charmed: "that's actually accurate"), escalates to Human Design ("I bet you're a Manifestor" / he half-keeps-up: "Manifestor or Manifesting Generator"), shares hers ("I'm a Projector"), then asks his Gene Keys / Life's Work and he is fully lost. Arc: her fluency rising, him sliding from charmed to out of his depth. Turn: people already try to read each other through this but it gets deep fast and they are guessing; Affirmology is the way to actually know yourself and whether you fit. Ties to relationships/synastry (Eros) + the dual audience; likely an early cultural-proof beat.

• Brand story is the OPEN of the investor manifesto film (narrative first, then figures/graphs/charts and the backend visualization). Website video alignment (today = animation-based change): both existing films already tell most of the brand story. Pre-audio Primer trailer: leave nearly as-is (optional: add the "programming the world layered over you" motif + name the SOS frame/payoff once). Post-audio What's In Store: keep structure (it is the suite + membership + guide), upgrade by landing the explicit 3-step payoff where "Not just understanding. Becoming." sits (Know yourself / Align to your blueprint / Become who you were written to be) + optional "Join the movement" under the CTA. Specifics in Affirmology_WebVideo_BrandStoryAlignment_v1.md. Higgsfield/Remotion versions + the investor film come over the next few days.

• BRAND STORY written and captured in Affirmology_BrandStory_v1.md (the spine narrative for the manifesto film, the pre-audio film, the site, the pitch). Core arc: the blueprint is profound but too complex for the linear mind (a headache, gets lost) -> we bypass the mind and deliver the truth to the subconscious where it is believed and accepted -> then we layer the transformational neuroscience tools that only work once the identity base is set -> the Subconscious Operating System. The trademarked 3 steps, said many ways: "Know Thyself. Attune to Your Blueprint. Become Who You Were Written to Be." / "Identify. Align. Transcend." / "Recognize. Realign. Rise." Taglines locked-ish: "Become who you were written to be" (heroic/destiny) and "Become Magnificent" (bold marketing) both have a home. The journey = Starter Kit (know + align) -> membership (tailored transits + guide) -> growth programs (context-based: money, work, love, breakthrough). Going to Sol for feedback.

• PRODUCT LADDER + 3-STEP FRAMEWORK captured in full in Affirmology_ProductLadder_Framework_v1.md. Headlines: the framework is Identify (know thyself) -> Align (attune to your blueprint) -> Transcend (growth). Ladder: starter kit ~$50-88 (10-12 audios, steps 1-2) -> base membership ~$15-22/mo (tailored transits + chart-grounded chatbot + a few journeys) -> premium/deep (deep journeys, depth psychology, deep readings, custom audios, robust chatbot) -> Affirmologists (pro/community creators in the lab, their members convert to our list) -> future in-person retreats via an offshoot business. Draft cosmic tier names: Founding Constellation (free beta) / First Light (starter) / Inner Orbit (base) / The Constellation (premium) / Affirmologists (pro). Two beta tracks to delineate: APP track (test the experience and our audios) and LAB track (co-create in the Studio; the astrology-strong folks like Lauren Martinez and Monika Arenas); some do both. Open questions: starter price/contents, journeys-included-vs-upsell, final names.

• Investor targeting + philosophy: we raise from mission-seeders and believers, not traction-perfectionists. An investor who requires revenue at this stage is not our investor. Downloaded VC checklists are additive only; take the gold, do not rework the story around them.

• Demand proof without our own revenue: use the market's willingness-to-pay as the evidence (people pay ~$44 for a single transit guide; CHANI ~$600k/month; the ~$15B-to-$20B market). Show it visually in the film as the demand signal; our own customer learnings come from the beta. If an investor needs revenue now, they are not our investor.
• Beta program (the "inner circle," name TBD; options: The Founding Constellation [pick], The First Forty, Inner Orbit): run a tight, committed FIRST WAVE of ~10-15 active testers (invite ~20-25 to allow for drop-off and the learning curve), then broaden in a second wave after folding in wave-one feedback. Depth over headcount given the product's learning curve.
• Gamified beta deal (expectations + reward ladder; self-selects the committed): ask = complete the journey, give feedback per audio (rating + voice note), try the second-level audios. Rewards = engage and give feedback -> Founding member status + free year; go deep -> lifetime free membership; community holders -> free affiliate membership + ability to make custom audios for their community. Recognition as a founding member is itself a reward.

• Timeline: start beta next week AND open investor conversations next week; the manifesto film serves both audiences; layer in real customer impressions as they arrive. (Jeff holds the full people list across categories: team, beta, investor, board, tech, music, future CEO, share-with. Not cataloged here per Jeff.)

• VC founder-evaluation checklist (source: "15 Things I Evaluate Before Investing in a Founder," Vinnie Lauria, Golden Gate Ventures, shared via Google Drive). Use it as the rubric the manifesto and pitch must pre-answer. The 15: (1) why this problem / personal connection, (2) why uniquely qualified to win, (3) can you be CEO of a $100M company (attract talent, align people, inspire, persist), (4) clear ambitious vision, (5) are you qualifying me as a partner, (6) do you know what great investors look like, (7) competition seen as the full battlefield, (8) how you handle feedback (coachable, not a follower), (9) what you learned from customers, (10) startups vs incumbents risk, (11) prove demand with evidence (revenue, retention, usage, pipeline), (12) where growth comes from and whether channels scale, (13) retention (do customers return on their own), (14) handling disagreement by finding alignment, (15) do people naturally want to follow you.

• Our readiness map: STRONG on 1, 2, 4, 7, 10, 15 (story, qualification, vision, competition, incumbent-awareness, and Jeff's gravity/community as the standout). HONEST GAPS on 9, 11, 12, 13 (customer-conversation evidence, proof of demand, growth channels, retention) because we are pre-revenue, exactly what the raise and the inner-circle/beta phase exist to close; name the early signal head-on rather than leaning on market size. BEHAVIORAL prep for 5, 6, 8, 14 (qualify investors, know what a great investor is, take feedback well, handle disagreement by finding alignment), prep Colin and Jeff on these for the room.
• Equity decision (long-term technical owner of the end-user app): offer about 2% vested (4-year vest, 1-year cliff) plus cash, not 5%. Start them as a contractor and convert only once they prove they can own it. Detail in Affirmology_BuildVsHire_UseOfFunds_v1.md.
• Build-vs-hire plan for the end-user app: Jeff vibe-codes the v1; a contractor hardens and scales it (security, payments, data, audio delivery, reliability, maintenance); optional long-term technical owner over time. Vibe-coding alone is risky for production scale, which is why a hire is funded.
• Use-of-funds working lines (added to the master doc after the PDF snapshot): technology and app development ~$50k; marketing and ad spend after inner-circle validation; September launch events and silent-disco headsets; beta-test refinement of the starter kit and membership tiers; affiliate-program prep. Finalize figures in the Executive Summary.
• Raise narrative on tech: we build the first version ourselves, but sustaining it at scale and reliability requires bringing someone on to own it, hence a funded technology line.
• Cost reference (2026): senior US full-stack freelancer ~$80 to $180/hr; offshore senior ~$25 to $60/hr; fixed-scope harden-to-production ~$15k to $50k; fractional/retainer ~$6k to $15k/month; agency MVP ~$30k to $150k.

(Earlier the same day, already folded into the master doc and its PDF: the full 15-point manifesto reframe, the founder-role correction, the 3-step secret sauce, the bypass-to-subconscious, the technique-and-context suite, why-now market data, the dedicated competitor analysis, and the "moat" scrub.)