Memory

built grok @bot to find shitcoins first it found them then i had to teach it to remember what price it found them at then when to stop liking them then how to measure whether it was actually right then its MCP token kept disappearing every hour so now it has immutable benchmarks, lifecycle states, outcome tracking, missed-run telemetry and hardened secret storage all this so it can tell me: "coin maybe go up"

Permanent context in Grok Bot ( @bot ) is a real game-changer. You adjust bot's behavior by simply talking to it. Bot interactions naturally create workflows. This is the breakthrough we were waiting for.

10 minutes into @instinct_inc and it already feels like cheating at life. It books, it orders, it watches my inbox, it remembers everything. The future is here. AGI might be too 🤯

Dear @interaction , I love Poke but it feels like he is getting dumber everyday I feel like I have to correct him every time we need something a little bit of nuance I already told him to lookup recent info before committing but he still makes mistakes frequently. https://t.co/hyxI5tbxld

Openclaw forget to use Skills I created a skill for Exec commands on my Windows node but openclaw forgets that he have to use that skill and once I said Hii and it was listing skills and checking doing unnecessary then saying Hii... Im using mem0 for memory.

I’ve tried a number of harnesses or AI assistants. Vellum can be used free if you download and use it locally. I then paired it with an LLM API (Openrouter or whatever you want) and what I liked was its memory and its pro-active nature. I’d go to it in the morning and it would have suggestions of things I need to do, emails to reply to. Kind of great for a person with ADHD. But I feel like I’m the only one who ever used it. Hermes and Openclaw get all the love and attention. Am I the only one? I used it for a few months, but I against my better judgement, used some new powerful LLMs and burned through too many credits so i shelved it and went back to a regular $20/mo ChatGPT subscription. After 3”2 months i tried to cancel but got a free month offer so I’m going to stick with it another month. Now that I’m used to using Luna and it’s cheap, and Deepseek is up in quality and still cheap, I’m thinking of going back next month (Or maybe sooner.) Question: Is there a reason so many people don’t use or talk about Vellum? Compared to Hermes I found it more user friendly and the memory system was amazing. Using it locally with BYO LLM API and Telegram, I don’t see it missing any featur

I ran Vellum for a bit and had same feeling, like I was only person on earth using it. Memory was really good for morning check-ins, it would queue up tasks without me asking. The lack of integrations is probably why it stays quiet, people want one tool that connects to everything instead of wiring up Composio themselves. Hermes gets hype because it's flashier but Vellum is more practical for daily stuff.

I’ve been banging away with Openclaw since about February. I love the concept and I have put in so many hours, and so many dollars, but it just drives me crazy how much it constantly gets wrong. Applying old stale rules, forgetting simple instructions, just generally getting things wrong. I’ve managed to get it to create and save several usable repeatable processes which are helpful, but on the whole it feels like someone could do so much better with the current models available. Theoretically it’s running the same model as ChatGPT, but even ChatGPT is getting better at its over arching knowledge and understanding of my business than my openclaw. I find myself scolding it so often. I even just told it that ChatGPT can do this task much better than you.