Dot Orchestrator, Gemini 4, Opus Video Editing — AI Daily Sep 30
734 messages · 72 active members
Overview
Topics
Jarvis detailed using Dot as a filter layer over OMP/Orca with Slack channels per project, GitHub integration, and wake-up scripts for sleeping orchestrators — 3-click setup vs hours for Hermes. Others wired Dot into Spaces for auto-updating task sheets every 15 min. Downside: users reported progressive dumbing-down and 6-hour API waits, suggesting OpenAI is compute-constrained post-Dev Day.
Google announced Gemini 4 with DeepMind posting strong benchmarks, potentially ahead of Opus 5.5 and GPT Astra. Access is limited to Fairwind Program devs and enterprise beta testers, so builders are cautiously optimistic but wary of benchmaxxing until real-world usage lands.
A builder fully automated 10+ minute YouTube edits using Claude Opus and DaVinci Resolve Studio, matching paid-editor quality in ~4 hours on a 3090/5090 with 64GB RAM — handling HeyGen avatars, motion graphics, and screen recording. anonymous also kept shipping Punk77: Chrome Carnage end-to-end with Opus, swapping to a locally-connected Dot when Codex ran out of image fuel.
Full pipeline shared: iPhone 13 in Manila → OBS → SRT → AWS relay → WireGuard tunnel → office Mac mini → TikTok Shop. Filipina creators warm accounts cheaply, early seeded buys signal GMV to TikTok's algorithm, then US talent takes over at optimal hours. Target: 8 accounts live 24/7 by EOY.
WhiskeyATX shared a recurring pattern where IC-to-purchase ratios collapse from 33-35% to 25%, with new creative being the only reliable fix — pixel learns the control, which also explains why A/B tests on dialed accounts rarely win. For Meta's periodic garbage-traffic days, watch funnel-step percentages (especially initiate-checkout ratios) rather than raw costs, and consider training an agent on hourly metrics labeled with historical bad days.
Words worth knowing
Three terms from the glossary.
Key Takeaways
- Dot works best as a top-layer orchestrator over OMP/Orca — Slack channel per project mapped to a GitHub repo, plus a wake-up script triggered by GitHub events to resume sleeping agents.
- Gemini 4 looks frontier-tier on benchmarks but is Fairwind/enterprise-gated — wait for real-world usage before swapping stacks; OpenAI is visibly compute-constrained (Dot + GPT-6.1 Sol both slow).
- Opus + DaVinci Resolve Studio can produce broadcast-quality 10+ min YouTube edits in ~4 hours; ScrapeCreators + AnyAPI is replacing Apify with reported $6k/mo savings.
- FB ad fatigue has one reliable fix: brand-new creative. Monitor funnel-step percentages (initiate-checkout ratios) rather than raw cost to catch Meta garbage-traffic days before burning budget.
- ElevenLabs v4 + the Claude connector reportedly 10x's voice agent quality — let Claude pick and optimize the voice via the official MCP.
Hot Threads
Using Dot as orchestrator layer above OMP with Slack channels per project
Building Punk77: Chrome Carnage with Opus and locally-connected Dot
FB initiate-checkout CPA decay and creative refresh as the only fix