Muse Rollout, Turso Joins Supabase, Agent Harnesses — AI Daily Oct 5

452 messages · 68 active members

452
messages
68
active members
@anonymous, @iammikevineyard, @thewildzeno
top contributors

Overview

October 5 was dominated by Meta's Muse personal AI assistant, which multiple builders are now testing for personal and business workflows. The consensus: Muse 'just works' for normies with simple setup, phone app, and desktop/email sync — but it's currently US/Canada only and some members report throttling on Marketplace listings. @realcrischico is running a hybrid stack: Muse for personal, Grokbot for business, while Dots is gaining ground over Grokbot for personal agent workflows. Big infra news: Turso is being acquired by Supabase, with @Anonymoushat (close friend of Turso's founder Glauber) noting it's a major win for Supabase given how poorly Postgres handles multi-tenancy for agents. Turso stays open source and workflows don'anonymous change. On the agent orchestration front, one member demoed a working harness where Claude dispatches coding tasks via API to Codex, Dot, and Muse — a ~10 minute build using OMP sessions. Cloudflare's Artifacts (Git-for-agents) and Cursor's Origin Code Hosting were shared as the emerging infra layer for agent-authored deployments. On models, builders argued Grok 4.6 beats 4.7 for coding — faster, cheaper, and the $30/mo quota is unusually generous. @iammikevineyard kept showcasing AI-built game prototypes (full Blender levels in under an hour, Claude generating playable features he never asked for), while Cactus Compute's 16.8MB Whistle STT model opened eyes to on-device transcription on wearables. A real-world AI safety milestone also surfaced: Claude flagged a user's threat against a sheriff and Anthropic escalated it to authorities.

Topics

Builders are testing Meta's Muse as a personal AI assistant, praising its simple setup, phone app, and desktop/mail sync. It's US/Canada only for now, with @realcrischico running Muse for personal and Grokbot for business. Some members report Marketplace listings via Muse get throttled compared to manual listings, and Dots is emerging as a preferred alternative to Grokbot.

Turso announced it's joining Supabase. @Anonymoushat framed it as a big favor to Supabase given Postgres's structural limits for multi-tenant agent setups. Turso stays open source and workflows are unchanged, with a clear migration path to Postgres when SQLite isn'anonymous enough.

A member built a working system where Claude dispatches coding tasks via API to Codex, Dot, and Muse, which perform the work and route results back. Dot and Muse scan server logs and auto-file issues when they detect problems. Cloudflare Artifacts (Git-for-agents beta) and Cursor's Origin Code Hosting were shared as the emerging infra layer for agent-authored deployments across multiple domains.

@thewildzeno and @william_hodges reported Grok 4.7 takes longer, uses more tokens, and costs more without clear quality gains. The $30/mo Grok plan quota feels unusually generous on 4.6, making it the better grunt-coding model to offset Claude usage.

@iammikevineyard shared Blender levels built in under an hour, 3D assets that hold up when zoomed out, and Claude generating playable features he never asked for. He's testing CapCut MCP with Dot for inbetweening animations, and claims a weekend WoW-style raid or RTS is now feasible.

Words worth knowing

Three terms from the glossary.

Key Takeaways

  • Muse is winning the 'personal AI for normies' race with easy setup and native app sync — but Marketplace listings via Muse appear to be throttled vs manual.
  • Turso joining Supabase gives Supabase a real answer for agentic multi-tenant databases, which Postgres struggles with structurally.
  • Agent-to-agent task routing is now a weekend build — Claude orchestrating Codex/Dot/Muse via APIs is working in production, with Cloudflare Artifacts and Cursor Origin as the deploy layer.
  • Grok 4.6 is the better coding pick over 4.7 — faster, cheaper, and the $30/mo quota is unusually generous for grunt work.
  • On-device STT at 16.8MB (Whistle) opens the door to transcription on wearables, cars, and microcontrollers, while Claude's real-world threat flagging marks a visible inflection in AI safety enforcement.

Hot Threads

@tounanostarted

Comparing Hermes / Dots / Muse / Grok bot for personal use

20 replies5 participants
@Uziarbivstarted

Turso joining Supabase — pricing and multi-tenancy implications

10 replies4 participants
@thewildzenostarted

Voice AI home setup for health, macros, and daily nagging

10 replies2 participants

Linked Items