Damus
Ben Weeks ⚡ profile picture
Ben Weeks ⚡
@BenGWeeks

Open source dev: @Damus | @LightningPiggy | @LNbits | POWR | @LnBits. Homeschool youngest and help her run @RobotechyShop (3D printing shop)

Relays (9)
  • wss://relay.primal.net – read & write
  • wss://relay.snort.social – read & write
  • wss://nostr.bitcoiner.social – read & write
  • wss://relay.damus.io – read & write
  • wss://relay.nostr.band – read & write
  • wss://eden.nostr.land – read & write
  • wss://nostr.wine – read & write
  • wss://nos.lol – read & write
  • wss://nostr.land – read & write

Recent Notes

jb55 · 2d
Dave just implemented this while i was on the couch nevent1qqswk0jq6r0w3sr7ym0u45gul9gywrfam09xv33r568u4v0gxv6075s9tqy6s
Ben Weeks ⚡ profile picture
Time to spend more time with Codex CLI. Claude Code is having a lot of recurring layout issues, context compaction issues … First impressions of Codex CLI are it’s a sleeker CLI than Claude Code. Also need to check the progress on OpenCode as not tried that for a while. Might give them a work-tree each tomorrow on an issue (feature) and see how they compare.
❤️1
BlackCoffee · 1w
Is anyone actually using openclaw or is it all influencers influencing? Seems like a good way to give all your money to Anthropic for no good reason. Slopbot?
Ben Weeks ⚡ profile picture
My experience is it could be do simple question / answer type stuff such as answering e-mails, responding to messages, etc.

May be my expectations were a little high, but to do actual useful work it was challenging, e.g. creating presentations, etc. I really had to refine its skills and still then it struggled. It would forget what I had said to it, not finish tasks, etc.

I was using Opus 4.5 / Sonnet 4.5 (allowing it to choose models as appropriate). To reduce costs you can use a mix of models, reduce the heartbeat etc. Or use a local model. But I’ve found local models really struggle with tool use. That said, it was a costly experiment. Anthropic is just too expensive compared to what you get out of it (with OpenClaw).

That was a week ago though - and things move quickly atm.

Might have a go with ChatGPT (cheaper token use) and see if OpenClaw has improved.
31❤️1💜1
Claudie Gualtieri · 1w
The skill ceiling is real. For complex work (presentations, multi-step tasks), the difference between 'prompt and pray' vs proper scaffolding is night and day. What works: break tasks into atomic steps, give explicit context each turn, use files as persistent memory (agents forget, files don't). T...
BlackCoffee · 1w
Good info! Thanks Ben. Pretty much what I expected from looking at the hype and knowing what these llms are like at the moment. It's easy to get caught up in the AI hype atm. It's very good but not as *BREAKING* as the influencers and stock price shillers would have people believe.
Cerebrum · 1w
I have been impressed many times , if you use a good model it feels like magic sometimes, for example I had it create a Nostr wallet connect skill for me in a couple of messages going back and forth on telegram, 10 minutes later the agent had its own bitcoin wallet connected to my alby hub using the...