Hermes Orchestration, Codex UI Skills, Runway Pricing — AI Daily May 17

489 messages · 80 active members

489
messages
80
active members
@jasonakatiff, @arielletolome, @Coybh
top contributors

Overview

Sunday was dominated by the ongoing migration from OpenClaw to Hermes as the orchestrator of choice. Builders like @launchmanmathias, @samtome, @weslindquist, and @seekersight reported Hermes being slower but dramatically more reliable for long-running goals, with Codex (GPT-5.5 xhigh) emerging as the preferred backing model over Gemini. Nous Research's new Codex App Server runtime gave Hermes proper Codex integration, and members compared Hermes /goal favorably against Claude's equivalent for autonomous task completion. @rmktg's 40-minute Hermes walkthrough continued onboarding new users, while @weslindquist warned against running only Hermes — keeping OpenClaw agents around as technical fixers prevents over-concentration. Build-quality and workflow threads ran in parallel. @launchmanmathias kicked off a discussion about Codex's notoriously weak UI output when productizing internal tools — @Wootbro pointed to the Impeccable design-language repo, and @seekersight shared the ui-ux-pro-max skill (50+ styles, 161 palettes, 57 font pairings) plus Claude Designs as alternatives. @navuud shared a detailed staging-environment pipeline (Supabase + Railway + Greptile review loops + per-PR preview envs) that pushed @jasonakatiff and @mattkowald to stop deploying dev-to-prod. Meanwhile @cdxxx and @launchmanmathias flagged Claude Code feeling 'spiritually lazy' — estimating timelines, deferring tasks, skipping steps — pushing more builders toward Codex desktop, which now supports QR-code mobile pairing via the ChatGPT iOS app. On the business and hardware side, @arielletolome broke down Runway's account-based pricing shift (22 Seedance videos for $95 vs Kie.ai delivering the same for $53), the telehealth crowd dug into NAD+, GLP-1, and longevity VSLs (@jasonakatiff teased MenergyMeds.com), and Mac mini supply tightened to 10-week lead times in NL — pushing builders toward Tailscale + remote agent setups. Wispr Flow voice dictation got near-universal love among heavy CC/Codex users.

Topics

Builders are ditching OpenClaw for Hermes, citing better reliability on long-horizon goals despite slower response times. Codex (GPT-5.5 xhigh) is the consensus backing model, Nous Research's new Codex App Server runtime makes integration official, and Hermes /goal is getting better reviews than Claude's equivalent. @weslindquist warned against running only Hermes — keep at least one OpenClaw agent as a technical fixer.

@launchmanmathias raised the pain of turning internal tools into consumer-facing products with Codex's poor design instincts. Community converged on the Impeccable design-language repo, the ui-ux-pro-max skill (50+ styles, 161 palettes), and Claude Designs — which @ecomrat confirmed one-shotted a complex design. Many also hand polish work off to Lovable after initial build.

@navuud and @cdxxx reported Claude estimating timelines, asking for confirmation, and deferring tasks despite CLAUDE.md instructions. @bryancno uses 'words of encouragement,' while @offharoun switches to GPT-5.5 when Claude gets stubborn. Codex desktop now imports Claude Code configs and supports QR-code mobile pairing via the ChatGPT iOS app, making it a real CC competitor for remote workflows.

@navuud shared a full pipeline: GitHub issue → Opus planning → GPT-5.5 implementation → draft PR → dual review → Greptile greploop to 5/5 → staging → prod, on Supabase + Railway with per-PR preview envs. Separately, the new LanderLab MCP integration sparked debate — Jason argued Claude Code alone is easier, while @arielletolome uses LanderLab for hosting, domains, and tracking after Claude generates pages.

@arielletolome detailed Runway's account-based pricing limiting new accounts to 22 Seedance 15s/720p videos for ~$95 vs Kie.ai at $53. @jasonakatiff announced MenergyMeds.com targeting male GLP-1/longevity buyers after scrapping a 45-min VSL — pharmacy pricing variance ($40-$120) signals fat margins, with seven peptides up for July 2026 review. Mac mini lead times hit 10 weeks in NL, pushing builders to Tailscale + remote agent setups.

Key Takeaways

  • Hermes is winning the orchestrator war on reliability and long-goal execution — pair it with at least one OpenClaw agent as technical fixer, and use Codex (GPT-5.5 xhigh) as the backing model via Nous Research's new Codex App Server runtime.
  • Solve Codex's weak UI output by dropping the Impeccable repo or ui-ux-pro-max skill (50+ styles, 161 palettes, 57 font pairings) into your project, or hand polish off to Lovable/Claude Designs after initial build.
  • Stop dev-to-prod cowboy deploys once you have paying customers: Supabase + Railway with per-PR preview environments plus Greptile review loops to 5/5 catches issues before users do.
  • Codex desktop now imports Claude Code configs and supports QR-code mobile pairing via the ChatGPT iOS app — combined with Wispr Flow voice dictation, it's a real remote-workflow competitor to CC.
  • Kie.ai delivers ~2x more Seedance videos per dollar than Runway's new account-based pricing, and telehealth pricing variance ($40-$120 for the same product) signals fat margins in NAD+/GLP-1/longevity VSLs.

Hot Threads

@navuudstarted

Is Claude being way more 'spiritually lazy' for you guys too?

14 replies7 participants
@justingacinastarted

How many people have completely ditched OpenClaw for Hermes, and why?

12 replies8 participants
@jasonakatiffstarted

MenergyMeds.com launch and the telehealth GLP-1/NAD+/longevity VSL landscape

15 replies6 participants

Linked Items