I've been been using OpenClaw for a bit now and the thing I'm missing is observability. What's this thing thinking/doing right now? Where's my audit log? Every rewrite I see fails to address this.
I feel Elixir and the BEAM would be a perfect language to write this in. Gateways hanging, context window failures exhaustion can be elegantly modeled and remedied with supervision trees. For tracking thoughts, I can dump a process' mailbox and see what it's working on.
Agree on the observability. Every time I've seen that mentioned on the many, many discussions on Xitter theres one of the usual clickbait youtube 'bros' telling you to go watch their video on how to make your own ui for it. Really shouldn't need to for such a fundamentally basic and crucial part of it. It's a bit of a hot mess.
I do! I have an M3 Ultra with 512GB. A couple of opencode sessions running work well. Currently running GML 4.7 but was on Kimi K2.5. Both great. Excited for more efficiencies to make their way to LLMs in general.
The prompt processing times I've heard about have put me off wanting to go that high with memory on the M series (hoping that changes for the M5 series though). What's the average and longest times you've had to wait when using opencode? Has any improvements to mlx helped in that regard?
The M5 ultra series is supposed to have some big gains around prompt processing - something like 3-4x from what I've read. I'm tempted to swap out my m4 mini that I'm using for this kind of stuff right now!
I tried Coder yesterday with OpenCode... didn't have a great experience. Got caught in a loop reading a single file over and over again until the context filled up. GLM 4.7 has been crushing it so far. One's thinking and other isn't so that's part of it I'm sure.
I'll give it a shot. For me it's (promise) is about removing friction. Using the Unix philosophy of small tools, you can send text, voice, image, video to an LLM and (the magic I think) it maintains context over time. So memory is the big part of this.
The next part that makes this compelling is the integration. Mind you, scary stuff, prompt injection, rogue commands, but (BIG BUT) once we figure this out it will provide real value.
Read email, add reminder to register dog with the township, or get an updated referral from your doctor for a therapist. All things that would normally fall through the cracks are organized and presented. I think about all the great projects we see on here, like https://unmute.sh/ and love the idea of having llms get closer to how we interact naturally. I think this gets us closer to that.
Once we've solved social engineering scams, we can iterate 10x as hard and solve LLM prompt injection. /s
It's like having 100 "naive/gullible people" who are good at some math/english but don't understand social context, all with your data available to anyone who requests it in the right way..
reply