Slowdown.AI draft-site review — meeting notes

Date: 2026-08-17 (afternoon) Context: Live review of the three password-gated drafts at slowdownai.vercel.app — /1 The Movement, /2 The Evidence Brief, /3 The Statement (policy register). Two further drafts (the Letter, the Fork) exist locally in variants/.

Decisions

Process: Arunim makes calls and logs each design decision (so any can be backed out later); the group meets TOMORROW (Mon 8/18) to walk through the marked decisions together. Fast async feedback via a new Slack channel; Dan's input arrives async (recording), not via Claude Code.

Feedback by site

/1 The Movement

/2 The Evidence Brief

/3 The Statement

Action items

Consult Rochelle(?) on the brand question (separate identity vs. lightly-branded campaign) — named as having the best branding judgment of the group.
Brainstorm the call-to-action set — the biggest open content question. Some candidates live in Adrian's existing document (not exhaustive); Adrian to spend ~30 min generating more.
Make a Slack channel for the group's fast async feedback on the site.
Arunim: bring the group a large explicit list of open design decisions for the site, ideally with small templates/examples showing how each option would look — that beats circulating whole-site drafts. Arunim is free to just make choices where he has a view. (Dan won't use Claude Code directly; his feedback will be async — voice recording or similar.) Feedback rounds targeted Tue/Wed and possibly the weekend.
Assemble a 99designs playbook from last time's learnings before kicking off the slowdown.ai competition (some notes were written down after the previous run — find and consolidate them).
Ask Dan the reasoning behind the 99designs request (logo vs. broader visual identity) before kicking it off — the room questioned whether a new standalone identity is even wanted. Agreed path: draft the 99designs brief anyway, run it past Dan for feedback (and ask the why-question) before submitting.
Adrian to connect after the meeting and share lessons from the social-movement literature — the funnel-engagement / ladder-of-engagement model (get people in the funnel, move them up the ladder of escalating actions).

Raw running notes

17:27 — Framing: Dan has been gravitating toward the broad slowdown as the solution for AI as a whole; simple enough that the public can get behind it. Dan wants to drive a lot of attention to it.
17:28 — Goals restated (matches SSOT): (1) the website that rallies the public behind the solution; (2) generically good; (3) people actually use it — the way activist types post Control AI's email-your-rep link on Slack/Twitter, slowdown.ai should have a use case people know how to reach for (exact use case not fully settled — likely a couple of CTAs + educational materials, cf. aisafety.com as aggregator); (4) communicates the idea effectively, converts to CTA or further reading; (5) super accessible to the general public.
17:29 — Other participant: goals make sense; open question whether CAIS is the right entity to hold/pursue this. Mass activism may be more the lane of Humans First / purpose-made orgs; CAIS Action Fund (C4) work is very different from what the C3 will do. Question raised: should this live under Humans First instead? → Parked, will return later (finish context first).
17:30 — (fragmented) Suggestion floated to amend goal #2: possibly "Humans First owns this instead" — the core of goal 2 is that "we" (CAIS broadly) own the slowdown idea and get credit/benefit; for now they'll bracket which entity (CAIS proper vs. Humans First) holds it.
17:31 — Goal 2 detail: some attribution so people know "we" run the site; CTAs structured to capture value — statement signing / email actions should feed a growing mailing list. Believed useful for the Team Human campaign (not its main CTA; possibly centerpiece of a follow-on campaign).
17:31 — Adrian: likes the idea; it overlaps a doc/action plan he has on why slowing down matters. Key point: Team Human has a petition, but the hard part is people vanishing into the ether after signing — post-sign retention/follow-up is the real problem to design for.
17:32 — Adrian's vision: the site as a standing action hub, not a one-shot petition — social initiatives (e.g. Team Human day), matching people by location to their local representatives with the exact petition/action, linking to open comment periods (e.g. FTC) as they arise. One place people go for all these actions. (Note: /3's write-your-rep letter generator is the closest existing piece.)
17:33 — Adrian: agrees this should be SEPARATE from CAIS's own website, because (1) it must be hyper-accessible, (2) it can provide things safe.ai doesn't — direct recommendations / political matching, (3) it isn't limited to educational materials; CAIS can still get credit on both platforms.
17:34 — Emerging third goal: the site should be ENGAGING — continually the centerpiece of a community, not a one-time visit. Team Human gets people to sign; what happens with them afterward is currently unclear — this site could be the "do something with the list" layer.
17:34 — Concrete position (Adrian): it needs to be a legally separate entity from CAIS — open whether that's Humans First or some association. Counter-consideration: associating it with just one or two orgs limits the aperture (and orgs are polarized on these issues).
17:35 — Movement theory: Team Human is a campaign that gets people to take ~1–3 actions; movements work by people taking action after action, building momentum toward larger actions over time. Implication: slowdown.ai should be the standing ladder of next actions, not another one-shot.
17:36 — Convergence: Team Human's main unsolved problem is getting people to keep acting after the petition/profile-picture moment, so slowdown.ai as the continuing-action layer is a natural fit. Campaign orgs (Humans First, Irreplaceable — apparently two distinct efforts) run campaigns only periodically, not weekly, so a standing source of action opportunities fills a real gap; general support for CAIS creating new ways for people to act.
17:38 — Site contents discussion: CTAs + educational materials + Dan's updating gallery of top #slowdown short-form videos (all three deployed drafts include this). Rochelle: need good tooling, especially for managing a large email list properly. Adrian & Rochelle already thinking about this for the big mailing list split with Irreplaceable — for slowdown.ai the requirement is tooling that scales.
17:40 — Stakeholders (garbled audio, verify): Dan is the primary requester/driver; Rochelle(?) secondary; Adrian consulted.
17:46 — CTA candidates: Emerson's slowdown site's set — sign the statement, tell your representatives, join the movement / spread the word — called "pretty good"; final CTA set is a decision still to make. Default primary CTA is sign-a-statement (easiest email capture), but it's the decision most dependent on broader strategic context — Adrian's input explicitly wanted here.
17:47 — Standing lens for Arunim: the user journey — keep asking "I'm a new person landing on this site: what does it feel like, what do I actually do here?"
17:47 — Two thoughts on framing/ownership: (1) "slowdown" may be a WEAKER ask than what a social movement coheres around — plausibly the movement ask is just "stop"; (2) even with CAIS owning the site, it can route people onward — e.g. a movements/get-involved section directing to Irreplaceable, Humans First, and others.
17:48 — Adrian (w/ Rochelle context): Team Human needs to start partnering with sister orgs anyway — both AI-safety-space and broader nonprofits who care about this — and slowdown.ai could be a good platform for those partnerships.
17:48 — Entity mechanics: C4s can make much stronger asks/pushes to signups than a C3 can; plan to share mailing lists across the entities (flagged as something to set up). Implementation sketch: a checked-by-default opt-in at signup sharing the email with Humans First / Irreplaceable lists.
17:50 — Branding debate: pushback on running yet another 99designs logo ("why create yet another visual identity?") — a new identity separates it in people's minds from CAIS / Team Human / Humans First / Irreplaceable. Counter: low-cost to try; and this site has one specific goal (slowdown) vs. CAIS's broader x-risk mission, which argues for its own identity. Reference model: FLI's superintelligence-statement site — deliberately barren, org branding light and at the edges, the statement is the star. Clarified: the point of the FLI example is that it was a campaign under FLI's existing brand — they did NOT create a whole new brand for it (an argument against a from-scratch identity). Near-consensus by 17:52: the site's mission (slowdown, a branch of CAIS's societal-scale-risk mission) warrants a separate aesthetic; the remaining open issue is narrowly the BRAND question (own logo/identity vs. lightly-branded campaign).
17:49 — Group looking at the Humans First website: strong reaction ("if I saw that I'd be like oh my god what's going on — but I think it's perfect") — intense aesthetic seen as effective for a movement.
17:41 — Guidance to Arunim: brief is vague, a million ways to do it — do what you think is right and get feedback. Acknowledged: generation is cheap, three designs already exist to look at; the open process question is how the FINAL website gets arrived at. Arunim raised: opinionated decisions vs. a fast Dan→Claude-Code feedback loop.
17:42 — Process: someone (Dan?) is out later this week, so move early. The three drafts are ready to review for aesthetics/direction now; the core trio will keep a tight feedback loop; Adrian's ongoing involvement TBD.
17:39 — 99designs competition to kick off (per SSOT next steps): ~$1,000 bounty, expect ~70–100 designers submitting ~2–7 logos each; last time required back-and-forth — make the brief more concise this time.