Research automation loops are tough this lends to my next idea since I use 1 model, run it autonomously between cli and desktop gpt project for better analysis. The local file access MCP helps the desktop app tremendously, and sol Pro in gpt desktop/web is amazing, well worth the setup
2026-07-11 07:07:03
0
Don Harry :
If you really want to know each Single Task you need your own Blockchain To log
2026-07-11 06:33:57
0
shibari_san :
interesting stuff, I'll certainly have a look - up to now I've been doing the whole obsidian brain thing to keep everything in sync but there has to be a better way!
2026-07-11 08:24:52
0
J455 L33 :
is there a solution for chatgpt/codex and claude?
2026-07-11 08:10:52
0
Lux Evilian :
yes, "whenever you see an issue or bug, that is out of context, file a ticket and continue". Those out-of-context issues would otherwise just be swallowed and never resolved.
2026-07-11 09:16:48
0
DF :
Been doing this for a year now. Love it
2026-07-11 03:04:16
4
Starbird :
I did the same, called it dispatch
2026-07-11 05:12:40
0
Curt Larson :
personal agents, local agents, work agents, server agents, they can all use inter agent communication.
2026-07-11 01:36:59
4
LearningAIwithLaureen :
The copy paste tax is so real. I run a whole team of AI agents for my billing business and the day I gave them a shared event system instead of me ferrying messages around was the day it stopped feeling like babysitting. Ticketing system for agents is exactly the right mental model. Great explanation (shark included).
2026-07-11 03:09:27
2
Chris Chism :
So did I through telegram on two different machines.
2026-07-11 03:27:55
1
QuantumWeaver :
I developed a workspace ecosystem for my agents to communicate with each other and themselves.
2026-07-11 01:33:27
1
im574677 :
guide
2026-07-11 03:58:24
1
steelno8585 :
I built something like this months ago. I track every action my agent does in a log, it gives a confidence score. Greater than 90% confidence is granted others I have to manually approve. We do sprint retros and retraining per agent and also track true api cost per project and session in the event we can’t be subsidized by Claude subscriptions anymore. Currently working on an HR platform for my agents so o can manage their employee profile outside of markdown
2026-07-11 01:11:51
1
Alyssa with AI | CPA :
Not me struggling with this exact build right now 😳
2026-07-11 02:49:47
2
Bryant :
I built a very basic AgentCollab.md file and instructed them to use it. They do pretty well with it.
2026-07-11 02:07:20
2
fr3klz :
Working on something similar but different to solve the same problem, but it’s more like a blockchain, so coordination without a coordinator.
2026-07-11 01:17:42
1
aubiematt :
I feel like everyone’s agents are writing this. Mine wrote a whole memory stack for MCP
2026-07-11 01:56:22
1
Nate Kelly :
100💯
2026-07-11 02:48:31
1
Lois :
I really want an agent that has the memory that the chat threads do.
2026-07-11 01:22:40
1
Elideus :
2026-07-11 01:22:43
1
Stephen Salone :
I built a trading bot that creates sub agents to trade stocks. I didn’t know this was possible
2026-07-11 01:16:51
1
Arty :
have you heard about the file system?
2026-07-11 06:58:47
0
⛰ Utah MTB :
this is nothing new. real devs been doing this for a while
2026-07-11 09:11:40
0
XtraOrdnary🇺🇸Dev🇸🇪Rick🇵🇭 :
using ado + mcp 🤷♂️
2026-07-11 10:20:54
0
HearSeeFeel :
In the Substack post are the links live?
2026-07-11 09:55:07
0
To see more videos from user @nate.b.jones, please go to the Tikwm
homepage.