deor

past founder of module (acq), now building @RelayProtocol | doing nothing at @raynlabs |

[deor @Deor]

a reflection on a year of operating a filler for @RelayProtocol, 7.4 million settled intents and $500m+ in settled volume later.

a filler is a new-ish term, and it describes the actor we run at @reservoir0x to help settle cross chain intents.

i get into it in the link below :)

[deor @Deor]

that passed, this also may

[Peter | Relay @ptrwtts]

Everyone expects stablecoins to be 1:1 with USD and each other. Currently, the two biggest ones are trading with 12bps spread. Fun challenge :)

Checkout the various price stabilization options we offer with @RelayProtocol https://t.co/4BUO8DZqwr

[deor @Deor]

lol the "midnight" option in slack instead of 12am

[deor @Deor]

@jonathangrahl

[deor @Deor]

will just make one

[deor @Deor]

anyone use an agent orchestrator / manager to manage multiple @AmpCode sessions? something like agent-deck or claude squad (both work poorly or dont work at all with amp)

[deor @Deor]

https://t.co/nB8SfJ78al

[deor @Deor]

weekly recap #3 from the last week, link below 🫡

should start renaming these to weekly ai recap

[deor @Deor]

@haikane writing is thinking 🫡🫡🫡

[deor @Deor]

@tanayjain yeah i use it for longer prompts usually. but mostly use it when typing out messages in slack

[deor @Deor]

since i started using AI agents more, i started writing a lot more. i asked amp to analyze my obsidian vault (where I write) to see the change:

very little of this is larger prompts. half of the increase is planning, the other half is journaling about problems

[deor @Deor]

i have a larger writeup about why i think this is happening in the weekly recap on sunday :)

[deor @Deor]

The feature that has been the most beneficial to me is the auto-compacting / removal of handoff.

I frequently use amp to do deep investigations and code reviews that eat up alot of context, and sometimes I would run out of context mid-task and have to break things up into multiple threads. Now the auto compacting is sufficient to keep things within the same thread. It lets me create a deep thread for a task, forget about it and revisit it later. Vs before having to baby it and keep an eye on its progress/context.

Things sometimes do hang still, but I assume thats just stability issues with the new remote system that are still being worked out.

Whats amazing to me is that I don't feel that there is any degradation with the auto compacting compared to handoff. Great work!

[deor @Deor]

@isthinking thats cool

[deor @Deor]

yo guys who is using port 3000 and 3001

can you stop using them so i can use thanks

[deor @Deor]

this is cool

[deor @Deor]

"there are cathedrals everywhere for those with the eyes to see"

weekly recap #2 ↓

with 25+ random things i read this week with notes

[deor @Deor]

@5hayan done ty

[deor @Deor]

"there are cathedrals everywhere for those with the eyes to see"

weekly recap #2 ↓

with 25+ random things i read this week with notes

[deor @Deor]

@amorfati yeah i ueses to use cc, but was awhile ago tbh

[deor @Deor]

if you are getting better results with gpt-5.5 than with opus 4.7 the problem might be your prompting