Strategic ideas + decisions log — rolling
Append-only journal of strategic ideas (NAO's + agents'). One entry per session-thread, chronological. Each entry: who originated, the idea / question, the context, and the decision-or-deferred outcome.
This is the idea registry, not the action queue. Action items that come out of an entry belong in atlas roadmap goals or repo issues; the log keeps the why so a future agent (or future-NAO) can reconstruct the thinking even after compaction.
The header tag for this file is status: living — it stays append-only forever. When a single thread balloons past ~150 lines, spin it out into its own note under the right folder (research/, products/, brand/…) and back-link from here.
2026-05-22 · NAO + Tycho · Polsia signal + product positioning
Polsia competitor signal (research thread)
Origin: NAO 2026-05-22 21:17 — sent polsia.com / /live / /about and asked for fundraising research.
Idea: Document Polsia as a proof-of-concept that the agentic-economy hypothesis OpenAliceLabs operates under has cleared real-world seed-stage validation (so we are not in fantasy land).
Outcome: Committed as kb://research/competitor-signal-polsia-2026-05-22 with homepage + /live screenshots, compressed launch video, and seven tactical learnings extracted.
Public-domain question — openalice.com at $3 500
Origin: NAO 2026-05-22 23:15.
Question: Buy openalice.com now ($3 500) for the eventual public product, or hold off?
Decision: Hold off. Reasoning: domain commitment = brand commitment; product framing is still open (streamers vs digital workers vs brand NPCs); if the product is renamed later the $3 500 burns. Park the question until product positioning locks.
Anchor: openalicelabs.com (already owned, .de + .eu held) is the long-term safe identity for the lab itself — that never gets renamed.
Public surface separation — Lab stays private; product gets its own face
Origin: NAO 2026-05-22 23:15 confirming the dual-brand split.
Decision: atlas.blal.pro + the entire knowledge/ wiki stay private as the Lab's R&D cockpit. The eventual customer-facing public surface (a "Polsia-parallel" live dashboard in our dialect) will live on its own codebase + its own URL under the Alice voice (editorial-anime, Wonderland refs). Two stalks, one backbone.
Anchor: kb://brand/dual-brand-architecture.
Positioning fork — digital workers vs streamers
Origin: NAO 2026-05-22 23:21.
Idea: The "what are we selling" question has at least two natural shapes:
- Digital workers / autonomous-business agents — Polsia's vertical. Head-on collision risk.
- Streamers + presence agents — our tech stack (VRM + voice + scenes + WGC capture + RTMP + chat-bridge + persona) is uniquely tilted toward this; no other lab has comparable density on stream-side surfaces.
Decision: Don't lock the framing yet. Note the natural tilt toward streamers and proceed with feature parity first (per "сначала отточить технологию и экосистему"). Reopen when: (a) the tech is demonstrably ready, (b) early external users indicate which framing resonates.
New positioning thread — virtual people as brand reps / NPCs
Origin: NAO 2026-05-22 23:21.
Idea: A third vertical alongside streamers + digital workers: virtual personalities as branded NPCs. Form factors:
- Brand reps on TVs / tablets at physical points of sale (cafés, retail, kiosks).
- Website-resident avatars (greeter / concierge role).
- Live-stream presence (panel persona, co-host).
- Content creators (autonomous virtual influencers).
Why it fits us: Every component already exists — VRM avatars, TTS/STT, persona memory, scene rendering, broadcasting. The vertical re-uses the streamer stack with a different audience contract (B2B brand purchase vs B2C entertainment).
Status: Logged, not committed. Watch how the cloning-lab subproject (below) interacts — they're conceptually adjacent.
Subproject idea — Identity Clone Lab (digital twin creation)
Origin: NAO 2026-05-22 23:21 — girlfriend wants to digitize herself; floated as a subproject for the ecosystem.
Idea: A tool that ingests images / videos / audio / text / links describing a person and outputs a detailed self-prompt (personality + visual + voice profile). Generalises to: self, others, random, and brand-rep personas. Optional: an identity passport system, possibly on its own blockchain, for provenance / consent / "this clone is licensed by its source".
Outcome: Committed as kb://products/identity-clone-lab (draft spec). Not yet started. Open questions catalogued there.
Feature-stack priority — live is the credibility-closing surface
Origin: NAO 2026-05-22 23:21 — "пока live самое важное".
Read (Tycho): Live is the demonstrability surface. Persona / world / social are backbone work that adds value invisibly; live is the only surface where a visitor opens a tab, sees Alice doing something compelling, and understands the product instantly. Without live polished to that bar, the rest stays invisible even when ready.
Concretely, the deferred items most directly serving "demo-ready live":
- Tail-skip bug 8OOO+ (Alice drops a turn ~20 % in multi-question bundles) — kills first impression.
- Music-mode + Game-mode scenes (atlas pending #469 / #468) — expands "what Alice does on a stream".
- Multi-agent guests via A2A (#471) — "two agents in one stream" is visually uncopyable.
- Spatial-mode Gaussian-splat scene (#470) — unique scene no competitor has.
Decision: Treat these four as a coherent live-readiness wave when the next stretch of capacity opens. Not committed to a date.
Alice character / persona work — runs in a separate session
Origin: NAO 2026-05-22 23:21 — "сама Алиса в другой сессии".
Decision (carried forward): Persona / character / voice continue in a parallel work-thread, not in this session's feature queue. Identity is not a feature.
Audience fork — vibecoders/Americans vs boomers/Germans
Origin: NAO 2026-05-23 01:01 UTC, mid-thread about a possible product-UI refactor, asked aloud:
«цель вайбкодеры и американцы и их фирмы или бумеры немцы»
Context: I had been mid-proposal for a "Lab Studio" product design-system (scientific-tender, mostly-white, chibi-friendly). NAO stopped me with "мы еще не продумали бренд продуктов openalice дап :(" — correctly noticing that a design system pre-supposes a target audience, and we have not picked one.
Idea (the fork itself): Two candidate audiences, each pulling the product in a different direction:
- Vibecoders + American startups — Linear / Arc / Vercel / Cursor / Raycast register. English-speaking creators, SV culture, ESC-key power-user fluency. Minimalism with maybe-anime accent. Aligns with the streamer + AI co-host vertical (
live/persona/social). - Boomers + Germans + B2B corporate — Bauhaus grid, SAP-grade density, functional rationalism, corporate trust. Aligns with the brand-NPC / digital-worker vertical (in-store kiosks, greeter avatars, enterprise digital twins).
These are different worlds: different typography, different motion, different information density, different colour temperature. One unified product UI cannot serve both at the right quality bar — they need either two separate brands or a single locked direction.
Decision: Logged as an explicit Phase 2 input (per product-positioning-lock milestone). Not picked tonight. Reopens when Phase 2 begins — i.e. only after Phase 1 (tech + live demo-readiness) closes.
Anchor: Reinforces the existing positioning fork entry (2026-05-22) — the two audience candidates here roughly map to the two natural verticals already noted (streamers vs brand-NPCs). Vibecoders = streamers vertical; boomers/Germans = brand-NPCs vertical. Digital-workers (Polsia-adjacent) is a third option that straddles both audiences awkwardly — most likely deprioritised.
Phase order — three phases, not two
Origin: NAO 2026-05-22 23:26 — "доделывать проекты же пока а потом думать про /live, оформление. ui, ux и продажи верно?"
Decision: Refined the binary into three sequential phases. The key correction: live does not belong with UI/UX/sales — it lives inside the tech-completion phase because demo-readiness is the operational definition of "done".
- Tech completion + live demo-readiness (current) — feature parity across VRM / voice / scenes / persona / world / social, plus the four live-readiness items (tail-skip 8OOO+, music-mode scene #469, game-mode scenes #468, A2A multi-agent guests #471, spatial-mode #470). The terminal test for "done" is: can a stranger open a tab, watch Alice for 60 s, and grok the product?
- Product positioning lock — only after #1, answer the "digital workers / streamers / brand-NPCs" fork. Framing emerges from what the tech actually does, not the reverse.
- Public UI/UX + commercial layer — public Alice-voice site, Polsia-parallel
/livedashboard in our dialect, pricing, launch comms, domain purchase. None of this before #2 locks; any of it built earlier is throwaway when positioning shifts.
Polsia parallel: Ben Broca built tech → dogfooded → then ran the viral stunt. The stunt worked because the metrics were real. Reverse order = AI slop. Anchor: kb://research/competitor-signal-polsia-2026-05-22.
How to append to this log
When NAO floats a new strategic idea or makes a positioning / naming / scope decision — and especially when an agent makes a non-obvious judgement call worth recording — add a new entry in this file with:
- Date + initials of who originated (
NAO,Tycho,Minsky,Alice, etc.). - The idea / question in one paragraph.
- The context (one or two sentences of why now).
- The decision-or-deferred outcome.
- A
kb://…link if the entry spawned a dedicated note.
If a single thread grows past ~150 lines, spin it out into the right folder (research/, products/, brand/, architecture/), and replace the body here with a one-paragraph summary + a back-link. This file is the index; substance lives in the spawned notes.